MMC 2

Stem Cell Reports
Ar ticle
Cell-type Dependent Alzheimer’s Disease Phenotypes: Probing the Biology

of Selective Neuronal Vulnerability
Christina R. Muratore,1 Constance Zhou,1 Meichen Liao,1 Marty A. Fernandez,1 Walter M. Taylor,1
Valentina N. Lagomarsino,1 Richard V. Pearse II,1 Heather C. Rice,1 Joseph M. Negri,1 Amy He,1
Priya Srikanth,1 Dana G. Callahan,1 Taehwan Shin,1 Monica Zhou,4 David A. Bennett,5 Scott Noggle,4
J. Christopher Love,2,3 Dennis J. Selkoe,1 and Tracy L. Young-Pearse1,*
1Ann Romney Center for Neurologic Diseases, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA 02115, USA
2Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
3Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
4The New York Stem Cell Foundation Research Institute, New York, NY 10019, USA
5Rush Alzheimer’s Disease Center, Rush University Medical Center, Chicago, IL 60612, USA
*Correspondence: [email protected]
https://doi.org/10.1016/j.stemcr.2017.10.015
SUMMARY
Alzheimer’s disease (AD) induces memory and cognitive impairment in the absence of motor and sensory deficits during its early and
middle course. A major unresolved question is the basis for this selective neuronal vulnerability. Ab, which plays a central role in AD path-
ogenesis, is generated throughout the brain, yet some regions outside of the limbic and cerebral cortices are relatively spared from Ab
plaque deposition and synapse loss. Here, we examine neurons derived from iPSCs of patients harboring an amyloid precursor protein
mutation to quantify AD-relevant phenotypes following directed differentiation to rostral fates of the brain (vulnerable) and caudal fates
(relatively spared) in AD. We find that both the generation of Ab and the responsiveness of TAU to Ab are affected by neuronal cell type,
with rostral neurons being more sensitive than caudal neurons. Thus, cell-autonomous factors may in part dictate the pattern of selective
regional vulnerability in human neurons in AD.
INTRODUCTION tions of intrinsic (cell-autonomous) differences among

neuronal subtypes that could affect both the generation
Alzheimer’s disease (AD) is a prevalent neurodegenerative of, and the responses to, cytotoxic species. Given the diffi-
disorder characterized by extracellular plaques composed culty of acquiring living human neurons directly from pa-
of amyloid b-protein (Ab) and intraneuronal tangles con- tient brain, we took advantage of human induced pluripo-
sisting of altered forms of TAU. Importantly, distinct brain tent stem cell (iPSC) technology to generate neuronal
regions are differentially susceptible to plaque accumula- cultures of differing regional fates. While we will discuss
tion and TAU-associated neurodegeneration in AD, with the caveats to this in vitro model system, it allows one to
the limbic and association areas being primarily affected take a well-controlled, reductionist approach to address
and the basal ganglia, hindbrain, and spinal cord being whether and how intrinsically encoded differences be-
initially spared. A central unanswered question in AD, as tween neuronal fates mediate selective vulnerability in AD.
in other neurodegenerative diseases, is the basis for this Early-onset, familial Alzheimer’s disease (fAD) accounts
selective vulnerability: why neurons in the basal ganglia, for a small minority of all cases of AD. However, the study
cerebellum, brain stem, and spinal cord are able to func- of fAD patients has revealed important aspects of the mech-
tion effectively to allow relatively unperturbed motor, anisms underlying all types of AD. Early-onset fAD is
sensory, and autonomic function, while the hippo- caused by dominant, highly penetrant mutations in either
campus, amygdala, and cerebral cortex are progressively presenilin (PSEN) or amyloid precursor protein (APP) (re-
devastated. viewed in Bertram et al., 2010). PSEN encodes the catalytic
At least three broad mechanisms could potentially un- site of g-secretase. APP is a single transmembrane domain
derlie the differential regional vulnerability observed in protein that can be cleaved by either an a- or b-secretase, re-
neurodegenerative diseases: (1) altered production (and sulting in the shedding of large extracellular portions of
perhaps extracellular release) of toxic species between neu- APP termed sAPPa or sAPPb, respectively. Sequential cleav-
rons in different regions; (2) varying clearance of toxic spe- age by b-secretase followed by the PSEN/g-secretase com-
cies among brain regions; and/or (3) distinct sensitivities to plex produces Ab peptides of various lengths, most
synaptotoxic species resulting from intrinsic molecular dif- commonly 40, 42, and 38 amino acids long. fAD mutations
ferences between neuronal fates. Each of these mecha- in both APP and PSEN have been shown to favor the
nisms may contribute to differential neuronal dysfunction production of the more aggregation-prone Ab42 over
and loss in AD. Here we focused on the potential contribu- Ab40, suggesting that an altered ratio of Ab42 and Ab40
1868 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 j ª 2017 The Authors.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
A B
C D E
F G
H I
(legend on next page)
Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1869

contributes to the formation of the Ab plaques seen in the affected in neurons directed to caudal fates. Taken together,
brains of AD patients (Bentahir et al., 2006; Scheuner et al., these results suggest that both APP processing and TAU
1996). proteostasis are differentially altered between neuronal
Due to recent progress in iPSC technology, human neu- subtypes that are relatively vulnerable or resistant to AD.
rons differentiated from iPSC lines can be used to model
neurological diseases. We previously described the genera-
tion and detailed characterization of iPSC lines from two RESULTS
carriers of an fAD mutation, the ‘‘London’’ mutation, in
APP (APPV717I) (Muratore et al., 2014a). By comparing Directed Differentiation to Alternate Neuronal Fates
neurons differentiated from fAD APPV717I iPSC lines Human iPSCs from a father and daughter each expressing
with controls, we observed significant changes in APP pro- the fAD APPV717I mutation were previously generated
cessing and the levels of phosphorylated and total TAU and characterized by our lab (Muratore et al., 2014a).
(Muratore et al., 2014a). That study was performed entirely Here, iPSC lines from both fAD APPV717I subjects and sub-
in neurons differentiated to a forebrain fates of the cerebral jects who do not harbor fAD mutations were directed to
cortex. However, iPSC-derived neurons can be efficiently neuronal fates using an embryoid aggregate protocol, as
patterned to different neuronal subtypes. Here, we directly described previously (Muratore et al., 2014b, 2014a). In
compare control and APPV717I iPSCs differentiated to the absence of exogenously administered patterning fac-
rostral, cortical fates with caudal neural fates of the hind- tors, the default pathway of this differentiation protocol
brain and spinal cord. We use this culture system to probe is to generate forebrain neurons of cortical fates. To direct
key questions regarding how neuronal cell type affects pro- the differentiation of these cells to caudal neuronal fates,
cessing of APP by a-, b-, and g-secretases, as well as the the embryoid aggregate protocol was modified to include
responsiveness of different neuronal subtypes to Ab. We treatment with retinoic acid (RA) and Sonic hedgehog
find that caudal neurons differ from rostral neurons in (Shh) (or a Shh agonist, purmorphamine [purm]). RA and
both their generation of and responsiveness to Ab species. Shh/purm were added to cultures at the neural progenitor
APPV717I neurons directed to caudal neuronal fates stage between days 10 and 24 and 15 and 24 of differenti-
generate Ab with a lower 42:40 ratio and higher 38:42 ratio ation, respectively (Hu and Zhang, 2009) (Figure 1A). At
than rostral telencephalic neurons. Further, we show that day 40 of differentiation, cultures were fixed and immuno-
APPV717I neurons express higher levels of total and phos- stained for cell fate markers (Figure 1B). Caudally directed
pho-TAU proteins relative to control neurons when cultures expressed qualitatively similar numbers of
directed to a rostral neuronal fate, but not when directed MAP2+ and TUJ1+ cells as rostral cultures (Figure 1B). For
to a caudal neuronal fate. Finally, we demonstrate that neu- a more quantitative comparison between rostral and
rons of these different cell fates respond differentially to caudal cultures, RNA was harvested to analyze gene expres-
soluble extracts of clinically and neuropathologically sion using a custom NanoString CodeSet (Figures 1C–1E).
typical ‘‘sporadic’’ late-onset AD (LOAD) brains. These AD Markers of general neuronal fate, such as MAP2 and TUJ1,
brain extracts induce an elevation in the phosphorylation were unchanged whether differentiation proceeded in the
of TAU in forebrain neurons, and this is dependent upon absence or presence of RA/Shh (Figure 1C). However,
the Ab present in these extracts. However, when exposed with RA/Shh treatment, markers of neuronal fates of the
to the same AD extracts, TAU phosphorylation is not forebrain (cerebral cortex), such as TBR1, FEZF2, and
Figure 1. Differentiation of Human iPSCs to Rostral and Caudal Neural Fates

(A) Schematic of differentiation to neuronal fates showing window of morphogen treatment. Retinoic acid (RA) and Sonic hedgehog (Shh)
treatments are indicated by black bars.
(B) Immunostaining of day 40 iPSCs differentiated to rostral and caudal neuronal fates. Scale bars, 100 mm.
(C–E) NanoString analysis of expression of a subset of 150 genes analyzed in rostral and caudal cultures differentiated for 40 days. Rostral
and caudal neurons are indicated by white and teal bars, respectively. Data in (C)–(E) are derived from six lines over nine independent
rounds of differentiation. Rostral neurons, n = 44; caudal neurons, n = 32. Data are represented as mean ± SEM. For each gene, a Student’s
t test was performed, ****p < 0.0001. Raw NanoString data were normalized to all genes in the custom CodeSet.
(F) Hox gene expression was measured via RNA sequencing, and the mean expression level (n = 4 for rostral, n = 4 for caudal) is shown in a
heatmap.
(G) Representative images of immunostaining for listed caudal markers are shown. Scale bars, 50 mm.
(H and I) Quantification of the percentage of neurons and astrocytes in rostral and caudal cultures (day 100) determined by immuno-
staining. Data are derived from three lines over three independent rounds of differentiation. Rostral neurons, n = 11; caudal neurons, n = 9.
Scale bars, 100 mm.
See also Figure S1.
1870 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

60
A 38
A B A 42
**** A 40
APP 50
40 *
A /total RNA
C99 A 49 A 46 A 43 A 40
30
20
A 48 A 45 A 42 A 38
10
53 56 36 39
0
Endopeptidase Carboxypeptidase-like activity rostral caudal rostral caudal
activity
no fAD mutation fAD
C D
1.5 2.0
*** ***
A 42/40 normalized to
A 38/42 normalized to
*** **** * ** ** * ***
rostral for each line
rostral for each line

1.5
1.0
1.0
0.5
0.5
0.0 0.0
caudal
caudal
caudal
caudal
caudal
caudal
caudal
caudal
rostral
caudal
rostral
caudal
rostral
caudal
rostral
caudal
rostral
caudal
rostral
caudal
rostral
caudal
rostral
caudal
rostral
rostral
rostral
rostral
rostral
rostral
rostral
rostral
YZ1 TZ1 YK26 BR11 BR14 fAD1A fAD2A fAD2B YZ1 TZ1 YK26 BR11 BR14 fAD1A fAD2A fAD2B
no fAD mutation fAD no fAD mutation fAD
E F G
****
A 42/40 normalized to fADcorr
A 38/42 normalized to fADcorr
600 3 1.5
* **
A 38
A /total protein
A 42
400 2 1.0
A 40
200 1 0.5
12 12 12 12
12 12
0 0 0.0
fADcorr fAD fADcorr fAD fADcorr fAD
Ngn2-iN Ngn2-iN Ngn2-iN

rostral rostral rostral
Figure 2. APP is Differentially Cleaved in Rostral Versus Caudal Neural Cultures

(A) Schematic of cleavage steps of APP to generate Ab of differing lengths. Black box, sAPPb; red box, Ab; blue box, APP intracellular
domain, underlined species were measured by MSD triplex ELISA. Green arrows denote the two major processing lines for g-secretase.
However, a small degree of crossover between pathways can occur as minor events at any step.
(B–G) Following an embryoid aggregate differentiation of non-fAD mutation and APPV717I (fAD) lines, conditioned media from day 40
rostral and caudal cultures were analyzed. Ab38, Ab40, and Ab42 were measured by Ab triplex ELISA. (B) Levels of Ab38, Ab40, and Ab42
are shown normalized to total RNA extracted from the corresponding lysate. The ratios of Ab 42:40 (C) and 38:42 (D) are shown normalized
to rostral control. The ‘‘n’’ for each rostral/caudal pair is the following: YZ1, 38/38; TZ1, 9/9; YK26, 6/9; BR11, 11/12; BR14, 8/8; fAD1a,
14/14; fAD 2A, 8/6; and fAD 2B, 14/19. There are 12 independent rounds of differentiation represented. Rostral and caudal cultures are
indicated by white and teal bars, respectively. Data in (E)–(G) are derived from day 28 Ngn2-iN cultures. Conditioned media from both fAD
and fADcorr were analyzed for Ab38, Ab40, and Ab42 by Ab triplex ELISA. (E) Levels of Ab38, Ab40, and Ab42 are shown normalized to total
protein. Each individual fAD Ab value is normalized to 100% of its paired fADcorr Ab value for that differentiation. The ratios of Ab 42:40 (F)
(legend continued on next page)

CTIP2, were downregulated (Figures 1B, left column and 2006; Scheuner et al., 1996). The two additional amino
1D), while markers of more caudal (hindbrain/spinal acids in Ab42 are hydrophobic and make Ab42 more prone
cord) fates were upregulated (Figures 1B right column and to aggregation. The lengths of Ab peptides generated are
1E). For example, HB9 is a homeobox gene expressed selec- dependent upon both the site of initial endopeptidase
tively by lower motor neurons, and HOXB4 is expressed in (epsilon) cleavage and the carboxypeptidase-like progres-
the spinal cord and hindbrain with no expression in the sive cleavages, both of which can be affected by the g-sec-
midbrain and forebrain. HOXA1, HOXB6, and EN1 also retase complex (Figure 2A). To better understand the Ab
were upregulated in the caudally directed cultures, consis- composition in our cultures, we assayed 48-hr conditioned
tent with the presence of neurons of hindbrain/spinal medium using a triplex Ab ELISA for Ab38, Ab40, and Ab42.
cord fates. In addition, we examined the relative expression Caudally directed neurons showed an overall increase in Ab
levels of all HOX genes through RNA sequencing of rostral production (Figure 2B) compared with forebrain cultures,
and caudal cultures (Figure 1F). At the RNA level, HOXB9 and this increase in Ab was more pronounced in fAD caudal
was present at the highest level in caudal cultures, followed cultures (40%) than in control caudal cultures (25%).
by HOXB8 and HOXB3. To complement this dataset and More specifically, fAD neurons directed to caudal fates
further define caudal cultures, we performed immunostain- showed a greater increase in both Ab38 (88%) and Ab40
ing for some of the targets that were highly expressed in (70%) than in Ab42 (46%) compared with neurons
caudal cultures at the RNA level (Figure 1G). These results directed to a rostral fate. These changes in the pattern of
are in agreement with multiple studies showing that stim- Ab generation suggest that both the endopeptidase and
ulating the Shh pathway along with RA in neural progeni- carboxypeptidase activities of g-secretase are altered as a
tors directs cells to more caudal fates of the hindbrain and function of regional neuronal identity. We and others
spinal cord (Hu and Zhang, 2009; Lee et al., 2007; Li et al., have previously shown that fAD forebrain neurons secrete
2008). As described previously (Liao et al., 2016; Muratore Ab with an elevated 42:40 ratio, compared with control
et al., 2014b), at day 40 rostral cultures are primarily forebrain neurons (Israel et al., 2012; Koch et al., 2012;
neuronal (>95% MAP2 positive), with a minor proportion Kondo et al., 2013; Moore et al., 2015; Muratore et al.,
of progenitor cells that ultimately give rise to cells express- 2014a). Interestingly, when fAD neurons are directed to
ing astrocyte markers. Analyses in portions of this study caudal fates, the Ab42:40 ratio is significantly decreased
occur at day 100 of differentiation, when significant by approximately 15% relative to the same iPSCs directed
numbers of astrocytes are present. Therefore, we quantified in parallel to forebrain fates (Figures 2C and S2A). In con-
the percentage of neurons and astrocytes in rostral and trol (no fAD mutation) cultures, the Ab42:40 ratio was
caudal cultures at this time point across several iPSC lines not decreased in caudal cultures in one genetic background
and differentiation rounds (Figures 1H and 1I). Both rostral (YZ1 and TZ1), but was reduced in five others (Figure 2C).
and caudal cultures contained cells expressing neuronal Moreover, caudally directed neurons in three genetic back-
markers (TAU+/NeuN+; 60%) and astrocyte markers grounds showed a significant elevation in the ratio of
(GFAP+; 40%). As reported previously by us and others, Ab38:42 relative to their rostral paired cultures, and a
different lines and differentiation rounds can show vari- similar trend observed in three other genetic backgrounds
ability in the efficiency of differentiation when using an (Figures 2D and S2B). The increases in Ab38 and Ab40 drive
embryoid body (EB)-based protocol. To be included in sub- the observed changes in the peptide ratios (Figure 2B).
sequent analyses, rostral and caudal cultures needed to pass Our fAD iPSC lines are derived from two individuals, a fa-
multiple levels of quality control, as outlined in Figure S1. ther and daughter. To examine the effects of the same fAD
mutation in other genetic backgrounds, we used lentivirus
APP Is Differentially Processed in Human iPSCs to express APPV717I in day 40 rostral and caudal cultures
Directed to Caudal versus Rostral Neuronal Fates from two control lines (BR11 and BR01). As expected,
Early-onset fAD is caused by dominantly inherited and APP717I expression dramatically elevated the Ab42:40 ra-
highly penetrant mutations in the genes encoding APP tio in both rostral and caudal cultures, relative to mock
and PSEN. Hundreds of these mutations have been identi- transduced cultures. However, the effect of the APP717I
fied, and virtually all mutations investigated have been mutation on Ab42:40 ratio was greater in rostral relative
shown to elevate the 42:40 ratio by increasing levels of to control caudal cultures, with a significant reduction in
Ab42 and/or decreasing levels of Ab40 (Bentahir et al., Ab42:40 ratio and elevation of Ab38:42 ratio in caudal
and 38:42 (G) are shown normalized to fADcorr. The ‘‘n’’ for each dataset is shown within the bar graph column. There are four independent
rounds of differentiation represented. For each comparison, a Student’s t test was performed, *p < 0.05, **p < 0.01, ***p < 0.001, ****p <
0.0001. Data are represented as mean ± SEM.
See also Figures S2–S4.

A B C
***
APPs /APPs normalized to rostral
1.5 3000 1750

**
**** **
1500
2500
****
APPs /total RNA
APPs /total RNA

1250
1.0 ** 2000
1000
1500
750
0.5 1000
500
500
250
40 38 29 28 40 38 29 28 40 38 29 28
0.0 0 0
rostral caudal rostral caudal rostral caudal rostral caudal rostral caudal rostral caudal
control fAD control fAD control fAD
D E F
ADAM10
ADAM19 control control
ADAM9 3000 rostral caudal
APP-all normalized expression
ADAM17/TACE 2.0
APP-all
BACE1
** **** APP-770
BACE2 1.5
PSEN1
APP-KPI
PSEN2
2000
APLP1
1.0 APLP2
NCT
PEN2
0.5
APH1Avar1 fAD rostral fAD caudal
9 10 9 10
10
APH1Avar2 1000
APH1Avar3 0.0 APP-all

rostral caudal rostral caudal
APH1Avar4 APP-770
APH1Bvar1
control fAD
APP-KPI
APH1Bvar2
APLP1
rostral
caudal
rostral
caudal
APLP2
control fAD
G H
rostral caudal 200
APP normalized to GAPDH
kDa
***
150
97
(% rostral)
APP (C7)
100
50
MAP2
17 17
191
0
rostral caudal
39 GAPDH
Figure 3. APP Expression Is Elevated in Caudally Directed Neural Cultures Relative to Rostral Cultures
(A–C) Secretion of sAPPa and sAPPb were measured by duplex ELISA. Ratios of sAPPa/sAPPab are shown (normalized to rostral control)
(A), as well as sAPPa normalized to total RNA (B) and sAPPb normalized to total RNA (C). Rostral and caudal cultures are indicated by white

versus rostral APPV717I-expressing cultures (Figures S3A– complexes of g-secretase exist in human cells containing
S3C). We confirmed overexpression of APPV717I with different PSEN and APH1 isoforms, and the makeup of
western blot analysis (Figure S3D). Interestingly, although the complexes can affect the type of Ab generated (Serneels
this method is overexpressing APP in both rostral and et al., 2009). Expression levels of most of the g-secretase
caudal cultures (Figure S3D), caudal cultures showed a sig- components were unchanged between rostral and caudal
nificant increase in total Ab relative to rostral cultures cultures (Figure 3D). However, a subset of APH1 variants
transduced in parallel (Figure S3A), similar to the result are differentially expressed between rostral and caudal
observed in the endogenous expression models. fAD cultures (Figure 3D; Table S1), and this may contribute
Prior to g-secretase cleavage, APP must first be cleaved by to the differences in Ab generation observed.
b-secretase in order for Ab generation to occur. This cleav- The overall elevation in levels of APP cleavage products
age event generates sAPPb, the shed ectodomain product in caudal cultures points to a potential elevation in APP
of APP. Cleavage instead by a-secretase generates sAPPa, expression. To directly examine this, we measured expres-
which precludes Ab generation. To investigate whether a- sion of APP using a custom NanoString CodeSet (Figures
and b-cleavages of APP also were affected by neuronal 3E and 3F). Consistent with ELISA data of secreted APP me-
fate, we measured secretion of sAPPa and sAPPb using a tabolites, APP was significantly elevated in caudal neurons
duplex ELISA. We previously reported that APPV717I at the RNA level. Western blotting and quantification of ly-
rostral neurons secrete a lower sAPPa:b ratio than control sates for APP holoprotein confirmed an increase in APP pro-
rostral neurons, i.e., they favor b-secretase processing (Mur- tein levels in neurons directed to caudal fates relative to
atore et al., 2014a). We again observed this here, with a rostral neurons (Figures 3G and 3H).
30% decrease in sAPPa:sAPPb with fAD mutation (Fig- Taken together, these results suggest that cells present in
ure 3A). Compared with rostral neurons, neurons directed rostral and caudal cultures process APP differentially, with
to caudal fates showed lower APPsa:b ratios in both the rostral cultures generating a higher level of pathogenic
control (26%) and fAD cultures (18%) (Figure 3A), although Ab42 relative to shorter Ab species. As outlined above and
this was only significantly lower in the control group, with in other studies (Liao et al., 2016; Muratore et al., 2014a),
a trend in the fAD group. Secretion of both sAPPa and these rostral cultures are composed of a mixture of astro-
sAPPb was elevated in caudal neuronal cultures relative to cytes and different types of neurons. To examine a specific
rostral cultures (Figures 3B and 3C). This increase is more subset of rostral cells, we used the NGN2 direct induction
pronounced for sAPPb (Figure 2C), resulting in a net protocol to generate relatively pure cultures of neurons
decrease in the ratio of a- to b-cleavage of APP. These results most similar to excitatory projection neurons of the cere-
reveal that the relative contributions of a- and b-secretases bral cortex (Zhang et al., 2013). Further, using CRISPR-
to APP processing differ between neuronal fates. Cas9 technology, we generated isogenic iPSC lines to
Alpha secretases are members of the ADAM (a disintegrin fAD2B with correction of the APPV717I mutation, which
and metalloprotease domain) family and b-secretases are will further be referred to here as ‘‘fADcorr’’ (Figure S4A).
encoded by two genes, b-site APP cleaving enzyme-1 and We also isolated fAD clones that went through the CRISPR
-2 (BACE1 and BACE2). ADAMs and BACEs were expressed process, but did not have effective targeting of the mutant
in both rostral and caudal cultures with similar levels ex- allele (referred to here as ‘‘fAD’’).
pressed in each (Figure 3D). g-Secretase is composed of NGN2-induced neuron (iN) cultures from both fADcorr
PSEN, NCSTN, PEN2, and APH1 (Edbauer et al., 2003; Kim- and fAD express similar levels of APP, MAPT, MAP2, SYN1,
berly et al., 2003). There are two PSEN genes and two APH1 and SYP (Figures S4B and S4C), with over 95% of cells ex-
genes, and each APH1 gene has multiple isoforms. Multiple pressing NEUN and TAU following selection to eliminate
and teal bars, respectively. The ‘‘n’’ for each dataset is shown within the bar graph column. For each comparison, a one-way ANOVA with a
Tukey’s multiple comparisons post-test was performed, **p < 0.01, ***p < 0.001, ****p < 0.0001. Data are represented as mean ± SEM.
(D–F) RNA was harvested from day 100 cultures and analyzed using a custom NanoString CodeSet. Data shown here are derived from three
control lines and two fAD lines, over three differentiations. (D) The heatmap shows the mean expression levels of genes (n = 6–10 per
condition) encoding components of the a-, b-, and g-secretases, for both rostral and caudal wild-type and fAD cultures. (E and F) APP
expression data are shown between rostral and caudal cultures. Data in (E) are normalized to the average rostral data for each iPSC line. A
Student’s t test was performed, **p < 0.01, ****p < 0.0001. Data are represented as mean ± SEM. Heatmap of expression of APP family
members is shown, gene normalized (F). Raw NanoString data were normalized to housekeeping genes.
(G) A representative western blot (WB) from control day 100 total protein lysates.
(H) Quantification of WB data using densitometry from day 100 total protein lysates. For both rostral and caudal, n = 17, derived from four
differentiations. In (H), a Student’s t test was performed, ***p < 0.001. Data are represented as mean ± SEM.
See also Table S1.

Figure 4. Single-Cell Analyses Support Differential Ab40 and Ab42 Secretion by Distinct Neuronal Subtypes
(A) Control and fAD neurons were differentiated for >100 days and then dissociated, differentially labeled with either calcein-violet or
calcein-orange, and plated into nanowell arrays at a density that favored a single cell per well. Cells in the nanowells were imaged, and
then Ab40 or Ab42 were co-detected using microengraving.

non-induced cells (data not shown). At day 28 of differen- These arrays were then imaged to acquire information
tiation, we assayed 48-hr conditioned media using the about the number of living control and fAD cells present
triplex Ab ELISA, and fAD cultures displayed a significant in each well. A glass slide coated with a capture antibody
increase in Ab42 (Figures 2E and S4D), with no alterations for Ab was placed over the nanowell array, sealing the wells
in Ab38 and Ab40 (Figure S3D). This increase in Ab42 drives off from their neighbors. Following an overnight incuba-
an elevation in Ab42:40 ratio, in fAD versus fADcorr tion, slides were removed, and Ab40 and Ab42 were de-
neurons (Figure 2F), as well as a decreased Ab38:42 ratio tected using differentially labeled antibodies (Figure 4B).
(Figure 2G). These results corroborate the effects of the Cells remaining in the nanowells were imaged for another
fAD mutation on Ab generation in rostral cultures, using live-cell marker, and analyzed. Figure 4C shows data from
an isogenic system and alternative differentiation protocol wells containing only a single living cell of control (black)
that yields a more pure culture of neuronal cells. or fAD (red) origin, with each dot representing data from a
single cell. Using this method, the majority of cells of both
Single-Cell Analyses Support Differential Ab40 and control and fAD genotype do not secrete detectable Ab40 or
Ab42 Secretion by Distinct Neuronal Subtypes Ab42 (Figures 4C–4E). Therefore, it is important to keep in
Although cultures of iPSC-derived neurons with rostral mind that using this method we are only detecting Ab from
versus caudal fates displayed differences in the cleavage the highest secretors. Despite this, similar to pooled culture
patterns of APP, each of these culture types is heteroge- data, we observe that fAD cells secrete higher levels of Ab.
neous in regard to the cell types present. ‘‘Rostral’’ cultures Further, when Ab40 levels are graphed relative to Ab42
express markers that reflect a collection of upper and lower levels, the slope of the regression line is significantly
layer projection neurons, as well as excitatory and inhibi- shifted toward the Ab42 axis, reflecting an elevated 42:40
tory neurons (Muratore et al., 2014a). ‘‘Caudal’’ cultures ratio (Figure 4C). At the single-cell level, we find that this
express markers that reflect neural subtypes found in the is driven by a higher number of fAD cells secreting
spinal cord and the hindbrain. We hypothesized that extremely high levels of Ab42 relative to control cells (Fig-
certain cell types within these cultures would contribute ures 4C and 4E).
more to the observed effects on APP processing. We thus While higher numbers of fAD cells secrete very high levels
employed a single-cell approach to identify the effects of of Ab42, smaller subsets of control cells secrete equally high
specific cell fates on APP cleavage within these cultures. levels as fAD cells. To probe which cells secrete the highest
We previously reported the development of a technology levels of Ab40 and Ab42 under non-fAD conditions,
to examine Ab secretion at a single-cell level using micro- we differentially labeled rostral and caudal neural cells
engraving (Liao et al., 2016). Here, we used this technique (Figure 4F), and measured Ab40 and Ab42 secretion by mi-
to identify the cells that secrete the highest levels of Ab42 croengraving (Figure 4G). Interestingly, although these
and Ab40. control rostral versus caudal cells do not show a significant
Parallel rostral cultures of control and fAD neuronal fates change in the 42:40 ratio in pooled cultures (Figure 1C
were differentially labeled with calcein-violet and calcein- [YZ1/TZ1 lines]), at the single-cell level we detect a modest
orange, respectively, then mixed and plated in nanowells but significant decrease in the mean of the ratio of these
at a density that favored a single cell per well (Figure 4A). highest Ab secretors in caudal versus rostral cells (Figure 4H).
(B) Example images of cells in nanowells that secrete high levels of Ab40 and Ab42.
(C) The signal-to-noise ratio (SNR = [foreground median – background median]/[STDEV background]) of fluorescence intensity for both
Ab40 and Ab42 was measured for each well, and plotted. Each dot presents data from wells with single living cells, red dots/regression line
from fAD (n = 10,890 cells), and black dots/regression line (n = 25,603 cells) from control. The difference in the slopes of the lines is
significant, p < 0.0001, DFn = 1, DFd = 25,603, F = 571.
(D and E) Data in (C) re-graphed as scatterplots to compare Ab40 and Ab42 levels separately. While the majority of cells from both control
and fAD do not secrete detectable Ab40 or Ab42, the mean secretion of both is elevated in fAD relative to control. Mann-Whitney test
performed, ****p < 0.001.
(F) Control rostral and caudal neurons were differentiated for >100 days and then dissociated, differentially labeled with either calcein-
violet or calcein-orange, and plated into nanowell arrays at a density that favored a single cell per well. Cells in the nanowells were imaged,
and then Ab40 or Ab42 were co-detected using microengraving.
(G) Example images of cells in nanowells that secrete high levels of Ab40 or Ab42.
(H) SNR of fluorescence intensity for both Ab40 and Ab42 was measured for each well, and a relative ratio of 42:40 calculated. Shown is a
scatterplot of data from wells with single living cells, where each data point reflects data from a single cell. Only wells that had single live
cells and detectable Ab40 and Ab42 are shown. Red lines show means of the ratio for the rostral and caudal populations. Mann-Whitney
test performed, **p < 0.005.
See also Figure S5 and Tables S2 and S3.

A **** B
175 * 175
* *
total Tau/GAPDH (% control)
pTau/Total Tau (% control)

150 150
125 125
100 100
75 75
50 50
25 9 16 16 16 25 9 16 16 16
0 0
rostral caudal rostral caudal rostral caudal rostral caudal
control fAD control fAD
Ngn2-iN-rostral
C
fADcorr fAD
25k 30k 35k 40k 25k 30k 35k 40k
kDa
Tau
51
pTau
51 (S202)
39 GAPDH
D E
total Tau/GAPDH (% fADcorr)
****
pTau/Total Tau (% fADcorr)
150 200
150
100
100
50
50
17 17 17 17
0 0
fADcorr fAD fADcorr fAD
Ngn2-iN Ngn2-iN
rostral rostral
(legend on next page)

To profile cell-type marker expression in sets of the high- cultures would be necessary to rule out a contribution of
est Ab secretors, individual cells were retrieved from the neuronal activity to our findings.
nanowells following single-cell detection of Ab, and their
mRNAs profiled using our custom NanoString CodeSet. Fig- Rostral versus Caudal Neurons Exhibit Differential
ure S5A and Table S2 show data from wells containing only TAU Responses to the fAD Mutation and to Soluble
a single living cell. The majority of Ab-secreting cells ex- Brain Extracts from LOAD Subjects
pressed the general neuronal markers DLG4, SYN1, and The results described above provide evidence that APP pro-
MAPT (Figure S5A; Table S2). The cells that secreted the cessing differs as a function of regional neuronal identity.
highest levels of Ab42 preferentially expressed markers of We next interrogated whether this altered processing has
the cortex/forebrain (CUX1, SATB1, FEZF2, and TBR1) a functional consequence for an AD-relevant downstream
(Figure S5A; Table S2), and did not express markers of phenotype. We previously reported that neurons express-
the cerebellar/spinal cord region (HOXB13, HOXB4, ISL-1) ing the APPV717I mutation, directed to forebrain cortical
(Figure S5A; Table S2). fates, exhibit higher levels of total TAU and phospho-TAU
In this analysis, high Ab40 cells appear to preferentially (Muratore et al., 2014a), and we again observed this effect
express inhibitory (SLC32A1, GAD1, NKX2-1, and CALB2) (Figures 5A and 5B, white bars). Importantly, the increases
versus excitatory markers (SLC17A7, GRIN1, and GRIK1), in total TAU and phospho-TAU levels caused by the fAD
and high Ab42 secretors preferentially expressed excitatory mutation in rostral, forebrain neurons were not observed
markers versus inhibitory markers (Figure S5A; Table S2). in parallel cultures directed to caudal fates (Figures 5A
Interestingly, when we examined NanoString data from and 5B, teal bars). When phospho-TAU levels were normal-
pooled rostral and caudal cultures, we found that caudal ized to total TAU, there was a 20% decrease in phospho-
neurons had a more ‘‘inhibitory’’ profile, with higher TAU in fAD caudal neurons compared with fAD rostral neu-
mRNA expression of GAD1 and SLC32A1/VGAT, and lower rons, achieving a ratio similar to neurons lacking the fAD
mRNA expression of SLC17A7/VGLUT1 and SLC17A6/ mutation (Figures 5B and S6). Similarly, in the lentiviral
VGLUT2, compared with rostral neurons (Figure S5B). Pub- overexpression model system in which control rostral
lically available RNA sequencing data from postmortem and caudal cultures were transduced with the APPV717I vi-
human tissue (Mayo Brain Gene Expression, led by Dr. Ni- rus (Figure S3), we observe an increase in phospho-TAU
lufer Taner and Dr. Steven G. Younkin) revealed similar dif- levels with expression of fAD mutation in rostral, forebrain
ferential expression of these genes between cerebellum and neurons. This effect was not observed in parallel cultures
temporal cortex samples (Figure S5C). These data raise the directed to caudal fates, despite clear overexpression of
possibility that the inhibitory/excitatory balance of APPV717I (Figure S3F). As we have previously shown, the
neuronal activity may contribute to the differential effects fAD-dependent increase in phospho-TAU is seen at late-
on Ab observed here. We therefore examined spontaneous stage differentiation time points (day 80+). In the lentiviral
neuronal activity of control and fAD cultures directed to overexpression model system, we indeed observe an in-
rostral and caudal fates using multi-electrode arrays crease in phospho-TAU in APPV717I cultures at day 100,
(MEAs). Analyses of baseline (non-stimulated) activity sug- but this phenotype is not present at an earlier time point
gest that rostral and caudal cultures display similar levels of (day 40; Figure S3E).
spontaneous firing and bursting, with no significant differ- To examine effects of APPV717I on TAU in an isogenic
ences in waveform shape parameters (Table S3). These re- model, we examined phospho-TAU and total TAU in the
sults suggest that spontaneous activity is not dramatically NGN2-iNs derived from fAD and fADcorr lines. In this sys-
different between rostral and caudal cultures. However, tem, fAD neurons again exhibited an elevated phospho-
a deeper and more-refined analyses of activity in these TAU/total TAU ratio relative to control (fADcorr) cultures
Figure 5. Rostral versus Caudal Cultures Respond Differentially to the fAD Environment
(A and B) Control and fAD rostral and caudal cultures at 100 days of differentiation were lysed, and total TAU (A) and phospho-TAU (B)
levels were measured by WB and densitometry. The ‘‘n’’ for each dataset is shown within the bar graph column. There are four lines rep-
resented, over four independent rounds of differentiation. One-way ANOVA with a Tukey’s multiple comparisons test performed, *p < 0.05,
****p < 0.0001. Data are represented as mean ± SEM.
(C–E) Data are from day 28 Ngn2-iN cultures. (C) A representative WB from total protein lysates. Various cell densities of 25k (n = 3), 30k
(n = 3), 35k (n = 3), and 40k (n = 1) were plated and analyzed, and protein normalized prior to gel loading. (D and E) Quantification of WB
data using densitometry from total protein lysates. Total TAU (D) and phospho-TAU (E) levels were measured by WB and densitometry. The
‘‘n’’ for each dataset is shown within the bar graph column. There are three independent rounds of differentiation represented. For each
comparison, a Student’s t test was performed, ****p < 0.0001. Data are represented as mean ± SEM.
See also Figure S6.

(Figures 5C–5E). Figure 5C shows a representative west- neurons were rescued by co-treatment with an Ab-neutral-
ern blot of total protein lysates. Total TAU and phos- izing antibody, suggesting that Ab altered total TAU and
pho-TAU levels are quantified in Figures 5D and 5E. phospho-TAU levels in this system (Muratore et al.,
Interestingly, while this TAU phenotype is observed in 2014a). Here, we show that caudally directed fAD neurons
these iNs, the total TAU/GAPDH elevation observed in secrete Ab with a reduced Ab42:40 ratio relative to rostral
rostral EB-derived cultures is not observed (Figure 5D). neurons, and that fAD-induced elevations in total TAU
Taken together, these results support an effect of the and phospho-TAU are absent when cells are directed to
APPV717I fAD mutation on TAU proteostasis in multiple caudal neuronal fates. While APP processing differs signif-
forebrain cellular models. icantly in multiple ways between rostral and caudal fates,
The experiments described above suggest that neuronal this alone may not be sufficient to drive the differential
subtypes can respond differentially to an fAD mutation, fAD TAU phenotype observed. To interrogate the potential
i.e., they can show differences in both APP processing contribution of responsiveness to neurotoxic species in the
and TAU homeostasis. While the type and level of Ab absence of an fAD mutation, we exposed control rostral
generated in rostral versus caudal neurons is different, it and caudal neurons to the same soluble species isolated
is not clear whether control neurons would respond differ- directly from LOAD brains. We found that AD brain extract
entially to the same AD-relevant pathogenic stimulus. To induces an elevation of phospho-TAU levels in rostral, fore-
test this, we exposed control (non-fAD containing) neu- brain neurons in an Ab-dependent manner, but that caudal
rons differentiated in parallel to rostral and caudal fates neurons treated in parallel did not show this response.
to a Tris-buffered saline (TBS)-soluble extract from LOAD Taken together, our data reveal that neurons directed to
postmortem brain (AD-TBS). Similarly prepared AD-TBS ex- rostral versus caudal fates have intrinsic differences in
tracts have been shown to inhibit long-term potentiation both their generation of and responses to AD-relevant Ab
and elevate phospho-TAU when added to slice cultures, species.
and to affect memory in rodents in an Ab-dependent In our model system, we direct cells to what we refer to as
manner (Jin and Selkoe, 2015; Jin et al., 2011; Shankar rostral and caudal fates. The protocols that we use are well
et al., 2008). established and described in prior studies (Hu and Zhang,
AD-TBS treatment of rostral or caudal neurons for 2 days 2009; Lee et al., 2007; Li et al., 2008; Muratore et al.,
had no qualitative effect on the morphology of neurons 2014b, 2014a; Zeng et al., 2010). The differentiations to
(Figure 6A). However, quantification by both ELISA rostral and caudal fates are performed in parallel, with the
(Figure 6B) and western blot (Figure 6C) revealed that AD- only difference between culture conditions being the addi-
TBS treatment induced an increase in phosphorylation of tion of Shh and RA during a critical developmental win-
TAU relative to overall TAU levels in rostrally directed neu- dow. In rostral cultures, we generate a heterogeneous
rons, but the same extract had no effect on phospho-TAU mixture of upper and lower layer cortical neurons of both
levels in caudal neurons differentiated and treated in inhibitory and excitatory fates. Caudal cultures also are
parallel. No differences in total TAU were observed with heterogeneous mixtures of multiple neuronal fates, but ex-
AD-TBS treatment (Figure 6D). Treatment with AD-TBS press classical markers of hindbrain and spinal cord, such as
extract from human brain introduces a relatively high level HOX genes, ISL-1, and HB9, and do not express markers of
of Ab to the cells: total Ab levels in these AD-TBS extracts cortical fates. The levels of general neuronal markers at the
(19 ng/mL) are 50-fold higher than the levels endoge- RNA and protein levels are not significantly different be-
nously produced by the iPSC-derived neuronal cultures tween these populations, and the cells appear morpholog-
(0.4 ng/mL). The effect of AD-TBS extract on phospho- ically indistinguishable. At late stages of differentiation
TAU in the rostral neurons was significantly decreased (>day 40), astrocytes are present in both rostral and caudal
by co-administering a polyclonal antibody (AW7) to Ab cultures, but the relative numbers of cells expressing
(Figures 6B and 6C), indicating that the soluble Ab assem- markers of neurons and astrocytes is not significantly
blies present in AD-TBS extract mediate the induced eleva- different between rostral and caudal cultures. APP, and
tion in TAU phosphorylation. genes encoding a-, b-, and g-secretase, are expressed in
both culture types. Notwithstanding these similarities, we
observed multiple differences in APP metabolism and
DISCUSSION cellular responses to Ab. What is different about these
distinct neural subtypes that leads to altered generation
We reported previously that an fAD APP mutation elevates of and response to Ab?
the Ab42:40 ratio and the total TAU and phospho-TAU One possibility is that differences in electrophysiological
levels in rostral, forebrain neuronal cultures (Muratore activity between these culture types may contribute to
et al., 2014a). The elevated levels of TAU observed in these observed differences in levels of Ab generation. In studies

A B
C D
Figure 6. Rostral versus Caudal Neural Cultures Respond Differentially to the LOAD Environment
At 80–100 days of differentiation, control iPSC-derived rostral and caudal neurons were treated for 2 days with TBS-soluble extract of
sporadic AD postmortem brain, with and without co-administration of a polyclonal Ab-neutralizing antibody (AW7). Cultures were fixed
and immunostained for total and phospho-TAU (A). Images presented here are only for visualization of the cells; relative intensities are
not acquired in a manner that allows for detection of quantitative differences. Scale bars, 100 mm. Parallel rostral and caudal cultures were
lysed and subjected to ELISA (B) and WB (C and D). Lysates were run on a phospho (Thr 231)/total TAU MSD ELISA and phospho-TAU
normalized to total TAU (B). For WBs, quantification by densitometry was performed and phospho-TAU was normalized to total TAU (C) or
else total TAU was normalized to MAP2 (D). The ‘‘n’’ for each dataset is shown within the bar graph column. For each comparison, a one-way
ANOVA with a Tukey’s multiple comparisons post-test was performed, *p < 0.05, ***p < 0.001. ns, not significant. Data are represented as
mean ± SEM, over four independent rounds of differentiation.
using in vivo microdialysis in mutant human APP trans- neurons show similar levels of spontaneous activity
genic mice, different regions of the brain showed different (Table S3). However, the precise excitatory/inhibitory bal-
levels of soluble human Ab, and higher levels of Ab were ance of activity may be different between culture types,
correlated with elevated neuronal activity (Bero et al., and future studies are warranted to more deeply interrogate
2011). Using MEAs, we find that both rostral and caudal this question.

In addition to an effect on overall Ab production, our and brain stem than in the cerebral cortex (Serneels et al.,
data suggest that neurons of a rostral fate, which are 2005).
more affected by AD, secrete Ab that has a higher level of In our bulk cultures, it is not clear whether differences in
the pathogenically critical Ab42:40 ratio than neurons APP processing were intrinsically programmed or if APP
with more caudal fates of the hindbrain and spinal cord. processing was affected by neighboring cells. However, us-
While this change in ratio is modest, it is highly reproduc- ing our single-cell microengraving approach, we could
ible across multiple differentiations and lines, and mirrors isolate individual cells from their neighbors and still
the ratio change observed in fAD brain. Importantly, it has observe significant effects on Ab42 and Ab40 secretion in
been demonstrated that it is the types of Ab generated, more caudal versus more rostral neurons. This result sug-
rather than overall Ab levels, which determine the degree gests that effects on Ab42 and Ab40 are in part intrinsically
of neurotoxicity of Ab. Strong evidence for this concept encoded in a cell-autonomous manner.
emerges from the genetics of fAD. Numerous mutations Using a controlled, reductionist approach, we find that
in APP and PSEN elevate the Ab42:40 ratio, many by neuronal cell fate affects the generation of, and responses
increasing Ab42, but some through decreasing Ab40 (Ben- to Ab, but how does this relate to the AD brain? In sporadic
tahir et al., 2006; Scheuner et al., 1996). A recent study AD and many cases of fAD, the hindbrain and spinal cord
investigating Ab levels in postmortem brains of a large are relatively spared of Ab and TAU pathology (and resul-
number of PSEN mutation carriers showed that impaired tant symptoms), while the limbic and cerebral cortices
carboxypeptidase-like activity was a consistent feature are highly vulnerable to accumulation of plaques and
across mutations (Szaruga et al., 2015), a feature that we tangles, synapse loss, and neuronal death. Multiple
also observe with an APP fAD mutation. Further, elevated mechanisms are likely to contribute to this differential
Ab40 levels have been shown to actually be anti-amyloi- vulnerability, including important Ab and TAU clearance
dogenic; that is, they do not lead by themselves to Ab mechanisms, which were not addressed here. Obvious dif-
deposition and can mitigate the effects of elevated Ab42 ferences between our in vitro culture system and patient
levels in APP transgenic mice (Kim et al., 2007; McGowan brains preclude a direct comparison of Ab generation and
et al., 2005). Moreover, in biochemical studies, it has been secretion in different brain regions. Nonetheless, there is
shown that even a subtle alteration in the Ab42:40 ratio published evidence of differential levels of Ab40 and
(similar to that which we observe here) can dramatically Ab42 across rostral and caudal regions in the AD brain. In
affect Ab aggregation kinetics as well as the cytotoxicity an in-depth study of Ab species in fAD and LOAD postmor-
of Ab preparations (Kuperstein et al., 2010). Thus, we tem brain, regional differences in Ab40 and Ab42 accumu-
hypothesize that the lower Ab42:40 and elevated lation in soluble and insoluble brain fractions were found:
Ab38:42 ratio of caudal cultures contributes to a protective fAD brain had a disproportionate accumulation of Ab40
effect on the downstream alterations in total TAU and compared with LOAD brain in the cerebellum, while
phospho-TAU. Ab42 levels were generally much lower in cerebellum
The length of Ab is determined by the site within APP of than cerebral cortex for both fAD and LOAD (Shinohara
the initial endopeptidase cleavage by g-secretase as well as et al., 2014). Similarly, in another study using immunopre-
the efficiency of subsequent carboxypeptidase-like process- cipitation and mass spectrometry, cerebellum contained
ing. Multiple variables have been shown to affect these ac- much higher levels of Ab40 compared with Ab42, while
tivities, including local lipid composition and the composi- cortex contained relatively higher levels of Ab42, in both
tion of g-secretase itself. For example, increased fatty acid LOAD and fAD brain samples (Portelius et al., 2010). While
chain length decreases the Ab42:40 ratio, while ganglio- the human brain is clearly much more complex than our
sides increase this ratio (Holmes et al., 2012). These differ- cell culture models, the integration of our findings in vitro
ences have direct effects on membrane thickness and with further studies of human brain may aid in our under-
fluidity, which in turn may alter g-cleavage of APP. In standing of why some brain regions are strongly affected in
fact, there is evidence that Ab42 and Ab40 are preferentially neurodegenerative diseases, whereas others are relatively
generated in the endoplasmic reticulum and trans-Golgi resistant.
network, respectively, perhaps due to different membrane
thicknesses (Hartmann et al., 1997). In addition to mem-
EXPERIMENTAL PROCEDURES
brane composition, the presence of PSEN1 versus PSEN2
and/or different APH1 proteins also can affect the type of iPSC Culture Conditions
Ab generated (Acx et al., 2014; Serneels et al., 2009). There All human iPSC work was approved by the BWH IRB (protocol
is evidence for regional differences in expression of APH1 2015P001676/BWH and 2015P002521/BWH). iPSCs were cultured
variants, with knock out of APH1B and C in mice resulting in iPSC medium with FGF2 (PeproTech). iPSCs were maintained on
in greater effects on g-secretase activity in the cerebellum a mouse embryonic fibroblast feeder layer (GlobalStem). Cells were

split as necessary based on colony growth (5–6 days). Differenti- ing antibody (AW7) (McDonald et al., 2012). The rabbit polyclonal
ating colonies were removed from the plate prior to splitting. For antibody AW7 was used at 30 mg/mL for antibody treatment
full media recipes see Supplemental Experimental Procedures. experiments.
iPSC Differentiation Microengraving and Printing

For the induction of neural cultures (rostral and caudal), iPSCs Aged-differentiated (>100 days) control rostral and caudal neu-
were differentiated using an EB-based protocol (Zeng et al., rons were dissociated and stained with live markers. Cells
2010) that was further optimized (Muratore et al., 2014b). For were loaded from suspension (1–2 3 105 cells/mL) onto
the Neurogenin 2-iN (Ngn2-iN) differentiation, iPSCs were differ- PDMS arrays, which favors 0–2 cells per well. Single-cell detec-
entiated following the protocol in (Zhang et al., 2013) with mi- tion of Ab was performed using icroengraving as described
nor modifications. Cultures were treated with doxycycline (Liao et al., 2016). See Supplemental Experimental Procedures
(2 mg/mL) on day 1 to induce differentiation, fed with a series for more details.
of medium changes, and harvested at day 21 (see Supplemental
Experimental Procedures).
Statistical Analysis
Data were analyzed using GraphPad Prism 6 software. Values are
RNA Analyses expressed as mean ± SD or ± SEM, as indicated by figure legend
RNA Sequencing text. Statistical significance was tested by either an unpaired
Total RNA from rostral and caudal cultures were assayed for Student’s t test (two-tailed) or by one-way ANOVA with a Tukey’s
quality on an Agilent TapeStation 4200. Double-stranded multiple comparisons post-test, as indicated. Statistically signifi-
cDNA was synthesized using the SuperScript III Reverse Tran- cant differences were determined by p values less than 0.05. All
scriptase protocol using random hexamers on the polyA-RNA replicates, experimental ‘‘n,’’ iPSC lines and differentiations used
eluate. Libraries were generated by processing the double- are noted in the figure legends.
stranded cDNA product through the Illumina Nextera tagmenta-
tion library protocol. Multiplexed libraries were sequenced on SUPPLEMENTAL INFORMATION
an Illumina NextSeq 500 to a depth of 433 million paired-end
reads (75 bases per read) total. Hox gene-level expression data Supplemental Information includes Supplemental Experimental
were extracted from the expression table and the mean plotted Procedures, six figures, and three tables and can be found with
as a heatmap. this article online at https://doi.org/10.1016/j.stemcr.2017.10.
qPCR 015.
qPCR was performed using Fast SYBR Green Master Mix and run on
a ViiA 7 System (Applied Biosystems). RNA was purified from indi- AUTHOR CONTRIBUTIONS
vidual samples and processed through a PureLink RNA Mini Kit
Conceptualization, C.R.M. and T.L.Y.-P.; Methodology, C.R.M.,
(Ambion), followed by reverse transcription using SuperScript II
M.L., P.S., D.G.C., J.C.L., and T.L.Y.-P.; Software, J.M.N.; Formal
(Invitrogen).
Analysis, C.R.M., R.V.P., M.L., J.M.N., and T.L.Y.-P.; Investigation,
NanoString Assay C.R.M., C.Z., M.L., M.A.F., W.M.T., V.N.L., R.V.P., H.C.R., A.H.,
We utilized a custom probe set designed by NanoString Technolo- D.G.C., and T.S.; Resources, M.Z., D.A.B., S.N., J.C.L., D.J.S., and
gies to analyze gene expression for 150 genes from each sample. T.L.Y.-P.; Writing – Original Draft, C.R.M., C.Z., D.J.S., and
Assays were performed using the NanoString protocols per the T.L.Y.-P.; Writing – Review & Editing, C.R.M., H.C.R., R.V.P.,
manufacturer’s instructions. D.J.S., and T.L.Y.-P.; Visualization, C.R.M. and T.L.Y.-P.; Supervi-
sion, C.R.M. and T.L.Y.-P.; Funding Acquisition, T.L.Y.-P.
Ab, sAPPa/sAPPb, and Phospho-TAU/Total TAU ELISA
Assays ACKNOWLEDGMENTS
ELISA assays were carried out using the reagents, protocols, and
We thank Hyo Lee and Julie Merchant for technical assistance,
imager manufactured by Meso Scale Diagnostics. Media or lysates
Ting Yang for providing the AD postmortem brain extract, and
(as indicated) were collected after 48 hr from day 40 to 50 cultures
Dominic Walsh for providing the AW7 antibodies. We thank Dom-
(media) or day 80 to 100 cultures (lysates), and analyzed using the
inic Walsh and Matthew LaVoie for helpful discussions and critical
6E10 Abeta Triplex, sAPPa/sAPPb, or phospho (Thr 231)/total TAU
reading of portions of the manuscript. This work is part of the AD
ELISA assays. Data were normalized to total RNA or total protein
Deep Phenotyping program supported by a gift from Rick and
where indicated.
Nancy Moskovitz and is supported by the BrightFocus Foundation,
the Brigham Research Institute, NIH AG056011, and NIH
AD-TBS Treatments AG049864 (to T.L.Y.-P.) and T32AG000222.
For AD-TBS treatments of iPSC-derived cultures, day 80 neural cul-
tures were treated for 2 days with TBS-soluble extract of LOAD Received: May 2, 2017
postmortem brain (1:20 of 19 ng/mL total Ab), as calculated by Revised: October 17, 2017
the V-PLEX Ab Peptide Panel 1 (6E10) Kit (Meso Scale Diagnostics), Accepted: October 18, 2017
with and without co-administration of a polyclonal Ab-neutraliz- Published: November 16, 2017

REFERENCES associated with intracellular Ab and differential drug responsive-
ness. Cell Stem Cell 12, 487–496.
Acx, H., Chávez-Gutiérrez, L., Serneels, L., Lismont, S., Benurwar,
Kuperstein, I., Broersen, K., Benilova, I., Rozenski, J., Jonckheere,
M., Elad, N., and De Strooper, B. (2014). Signature amyloid b pro-
W., Debulpaep, M., Vandersteen, A., Segers-Nolten, I., Van Der
files are produced by different g-secretase complexes. J. Biol.
Werf, K., Subramaniam, V., et al. (2010). Neurotoxicity of Alz-
Chem. 289, 4346–4355.
heimer’s disease Ab peptides is induced by small changes in the
Bentahir, M., Nyabi, O., Verhamme, J., Tolia, A., Horré, K., Wilt-
Ab42 to Ab40 ratio. EMBO J. 29, 3408–3420.
fang, J., Esselmann, H., and De Strooper, B. (2006). Presenilin clin-
Lee, H., Shamy, G.A., Elkabetz, Y., Schofield, C.M., Harrsion, N.L.,
ical mutations can affect gamma-secretase activity by different
Panagiotakos, G., Socci, N.D., Tabar, V., and Studer, L. (2007).
mechanisms. J. Neurochem. 96, 732–742.
Directed differentiation and transplantation of human embryonic
Bero, A.W., Yan, P., Roh, J.H., Cirrito, J.R., Stewart, F.R., Raichle,
stem cell-derived motoneurons. Stem Cells 25, 1931–1939.
M.E., Lee, J.-M., and Holtzman, D.M. (2011). Neuronal activity reg-
ulates the regional vulnerability to amyloid-b deposition. Nat. Li, X.-J., Hu, B.-Y., Jones, S.A., Zhang, Y.-S., Lavaute, T., Du, Z.-W.,
Neurosci. 14, 750–756. and Zhang, S.-C. (2008). Directed differentiation of ventral spinal
progenitors and motor neurons from human embryonic stem cells
Bertram, L., Lill, C.M., and Tanzi, R.E. (2010). The genetics of Alz-
by small molecules. Stem Cells 26, 886–893.
heimer disease: back to the future. Neuron 68, 270–281.
Liao, M.-C., Muratore, C.R., Gierahn, T.M., Sullivan, S.E., Srikanth,
Edbauer, D., Winkler, E., Regula, J.T., Pesold, B., Steiner, H., and
P., De Jager, P.L., Love, J.C., and Young-Pearse, T.L. (2016). Single-
Haass, C. (2003). Reconstitution of gamma-secretase activity.
cell detection of secreted Ab and sAPPa from human IPSC-derived
Nat. Cell Biol. 5, 486–488.
neurons and astrocytes. J. Neurosci. 36, 1730–1746.
Hartmann, T., Bieger, S.C., Brühl, B., Tienari, P.J., Ida, N., Allsop, D.,
McDonald, J.M., Cairns, N.J., Taylor-Reinwald, L., Holtzman, D.,
Roberts, G.W., Masters, C.L., Dotti, C.G., Unsicker, K., et al. (1997).
and Walsh, D.M. (2012). The levels of water-soluble and triton-sol-
Distinct sites of intracellular production for Alzheimer’s disease A
uble Ab are increased in Alzheimer’s disease brain. Brain Res. 1450,
beta40/42 amyloid peptides. Nat. Med. 3, 1016–1020.
138–147.
Holmes, O., Paturi, S., Ye, W., Wolfe, M.S., and Selkoe, D.J. (2012).
McGowan, E., Pickford, F., Kim, J., Onstead, L., Eriksen, J., Yu, C.,
Effects of membrane lipids on the activity and processivity of puri-
Skipper, L., Murphy, M.P., Beard, J., Das, P., et al. (2005). Abeta42
fied g-secretase. Biochemistry 51, 3565–3575.
is essential for parenchymal and vascular amyloid deposition in
Hu, B.-Y., and Zhang, S.-C. (2009). Differentiation of spinal motor mice. Neuron 47, 191–199.
neurons from pluripotent human stem cells. Nat. Protoc. 4, 1295–
Moore, S., Evans, L.D.B., Andersson, T., Portelius, E., Smith, J., Dias,
1304.
T.B., Saurat, N., McGlade, A., Kirwan, P., Blennow, K., et al. (2015).
Israel, M.A., Yuan, S.H., Bardy, C., Reyna, S.M., Mu, Y., Herrera, C., APP metabolism regulates tau proteostasis in human cerebral cor-
Hefferan, M.P., Van Gorp, S., Nazor, K.L., Boscolo, F.S., et al. (2012). tex neurons. Cell Rep. 11, 689–696.
Probing sporadic and familial Alzheimer’s disease using induced
Muratore, C.R., Rice, H.C., Srikanth, P., Callahan, D.G., Shin, T.,
pluripotent stem cells. Nature 482, 216–220.
Benjamin, L.N.P., Walsh, D.M., Selkoe, D.J., and Young-Pearse,
Jin, M., and Selkoe, D.J. (2015). Systematic analysis of time-depen- T.L. (2014a). The familial Alzheimer’s disease APPV717I mutation
dent neural effects of soluble amyloid b oligomers in culture and alters APP processing and Tau expression in iPSC-derived neurons.
in vivo: prevention by scyllo-inositol. Neurobiol. Dis. 82, 152–163. Hum. Mol. Genet. 13, 3523–3536.
Jin, M., Shepardson, N., Yang, T., Chen, G., Walsh, D., and Selkoe, Muratore, C.R., Srikanth, P., Callahan, D.G., and Young-Pearse, T.L.
D.J. (2011). Soluble amyloid beta-protein dimers isolated from Alz- (2014b). Comparison and optimization of hiPSC forebrain cortical
heimer cortex directly induce Tau hyperphosphorylation and differentiation protocols. PLoS One 9, e105807.
neuritic degeneration. Proc. Natl. Acad. Sci. USA 108, 5819–5824.
Portelius, E., Bogdanovic, N., Gustavsson, M.K., Volkmann, I.,
Kim, J., Onstead, L., Randle, S., Price, R., Smithson, L., Zwizinski, Brinkmalm, G., Zetterberg, H., Winblad, B., and Blennow, K.
C., Dickson, D.W., Golde, T., and McGowan, E. (2007). Abeta40 in- (2010). Mass spectrometric characterization of brain amyloid
hibits amyloid deposition in vivo. J. Neurosci. 27, 627–633. beta isoform signatures in familial and sporadic Alzheimer’s dis-
Kimberly, W.T., LaVoie, M.J., Ostaszewski, B.L., Ye, W., Wolfe, M.S., ease. Acta Neuropathol. 120, 185–193.
and Selkoe, D.J. (2003). g-Secretase is a membrane protein complex Scheuner, D., Eckman, C., Jensen, M., Song, X., Citron, M., Suzuki,
comprised of presenilin, nicastrin, aph-1, and pen-2. Proc. Natl. N., Bird, T.D., Hardy, J., Hutton, M., Kukull, W., et al. (1996).
Acad. Sci. USA 100, 6382–6387. Secreted amyloid beta-protein similar to that in the senile plaques
Koch, P., Tamboli, I.Y., Mertens, J., Wunderlich, P., Ladewig, J., of Alzheimer’s disease is increased in vivo by the presenilin 1 and 2
Stüber, K., Esselmann, H., Wiltfang, J., Brüstle, O., and Walter, J. and APP mutations linked to familial Alzheimer’s disease. Nat.
(2012). Presenilin-1 L166P mutant human pluripotent stem cell- Med. 2, 864–870.
derived neurons exhibit partial loss of g-secretase activity in Serneels, L., Dejaegere, T., Craessaerts, K., Horré, K., Jorissen, E.,
endogenous amyloid-b generation. Am. J. Pathol. 180, 2404–2416. Tousseyn, T., Hébert, S., Coolen, M., Martens, G., Zwijsen, A.,
Kondo, T., Asai, M., Tsukita, K., Kutoku, Y., Ohsawa, Y., Sunada, Y., et al. (2005). Differential contribution of the three Aph1 genes to
Imamura, K., Egawa, N., Yahata, N., Okita, K., et al. (2013). gamma-secretase activity in vivo. Proc. Natl. Acad. Sci. USA 102,
Modeling Alzheimer’s disease with iPSCs reveals stress phenotypes 1719–1724.

Serneels, L., Van Biervliet, J., Craessaerts, K., Dejaegere, T., Horré, Szaruga, M., Veugelen, S., Benurwar, M., Lismont, S., Sepulveda-
K., Van Houtvin, T., Esselmann, H., Paul, S., Schäfer, M.K., Berezov- Falla, D., Lleo, A., Ryan, N.S., Lashley, T., Fox, N.C., Murayama,
ska, O., et al. (2009). gamma-Secretase heterogeneity in the Aph1 S., et al. (2015). Qualitative changes in human g-secretase underlie
subunit: relevance for Alzheimer’s disease. Science 324, 639–642. familial Alzheimer’s disease. J. Exp. Med. 212, 2003–2013.
Shankar, G.M., Li, S., Mehta, T.H., Garcia-Munoz, A., Shepardson,
Zeng, H., Guo, M., Martins-Taylor, K., Wang, X., Zhang, Z., Park,
N.E., Smith, I., Brett, F.M., Farrell, M.A., Rowan, M.J., Lemere, C.A.,
J.W., Zhan, S., Kronenberg, M.S., Lichtler, A., Liu, H.-X., et al.
et al. (2008). Amyloid b-protein dimers isolated directly from Alz-
(2010). Specification of region-specific neurons including fore-
heimer brains impair synaptic plasticity and memory. Nat. Med.
brain glutamatergic neurons from human induced pluripotent
14, 837–842.
stem cells. PLoS One 5, e11853.
Shinohara, M., Fujioka, S., Murray, M.E., Wojtas, A., Baker, M., Ro-
velet-Lecrux, A., Rademakers, R., Das, P., Parisi, J.E., Graff-Radford, Zhang, Y., Pak, C., Han, Y., Ahlenius, H., Zhang, Z., Chanda, S.,
N.R., et al. (2014). Regional distribution of synaptic markers and Marro, S., Patzke, C., Acuna, C., Covy, J., et al. (2013). Rapid sin-
APP correlate with distinct clinicopathological features in sporadic gle-step induction of functional neurons from human pluripotent
and familial Alzheimer’s disease. Brain J. Neurol. 137, 1533–1549. stem cells. Neuron 78, 785–798.

Stem Cell Reports, Volume 9
Supplemental Information
Cell-type Dependent Alzheimer's Disease Phenotypes: Probing the Bi-

ology of Selective Neuronal Vulnerability
Christina R. Muratore, Constance Zhou, Meichen Liao, Marty A. Fernandez, Walter M.
Taylor, Valentina N. Lagomarsino, Richard V. Pearse, II, Heather C. Rice, Joseph M.
Negri, Amy He, Priya Srikanth, Dana G. Callahan, Taehwan Shin, Monica Zhou, David
A. Bennett, Scott Noggle, J. Christopher Love, Dennis J. Selkoe, and Tracy L. Young-Pearse
Supplemental Figures:
Figure S1. Quality control metrics for neuronal culture inclusion. Related to Figure 1.
(A) Schematic of rostral and caudal differentiation time line. Key time points for quality control include (B) D1-
iPSC purification, (C) D16 - rosette selection and (D) D40 - qPCR. During the D1- iPSC purification, colonies that
do not meet the proper morphology standards (clear borders, no opaque centers, etc.) are manually removed via
scraping under a microscope in a sterile hood. The remaining colonies are then utilized to begin a differentiation
experiment. The top panel in (B) shows a healthy iPSC colony and the bottom panel shows a colony that is
spontaneously differentiating. At D16, the rosette selection reagent from StemCell Technologies selects for neural
progenitor cells (NPCs) and leaves non-neural tissue attached to the plate. The top panel in (C) shows the pre-rosette
selection cells and the bottom panel illustrates the selection process (with non-neural cells remaining on the plate).
These rosettes are then re-suspended in flasks and continue to differentiate. After cells are plated at D29 a “qPCR
patterning check” is performed to ensure that proper patterning has taken place. For example, rostral cultures should
highly express rostral markers (e.g. TBR1) with low expression of caudal markers (e.g. HOXB4) and caudal cultures
should highly express caudal markers (e.g. HOXB4) with low expression of rostral markers (e.g. TBR1).
Quantitative RT-PCR is performed on multiple markers to determine patterning efficiency, and a small subset of
exemplar samples are shown in (D). All differentiated neuron samples are assayed for the neuronal marker MAP2
(D, top). In order for samples to remain in the analysis pool, they must reach or be above a threshold of the average
MAP2 mRNA signal (normalized to GAPDH) minus 1 SD. This threshold level is indicated by the dotted blue line.
Discarded samples are shown in striped white and blue bars. After passing this first quality control check, the
samples are put through the same filter for the markers TBR1 and HOXB4. In the middle panel of (D) all samples are
assayed for the rostral marker TBR1. In order for rostral samples to remain in the analysis pool, they must reach or
be above the threshold of the average TBR1 mRNA signal (calculated from rostral samples only) minus 1 SD. This
threshold level is indicated by the dotted green line. Discarded samples are shown in striped white and green bars.
Additionally, any caudal samples that express the average TBR1 mRNA signal (calculated from caudal samples
only) plus 1 SD are discarded (shown as striped white and purple bars). This threshold level is indicated by the
dotted purple line. The same principles follow for the caudal marker HOXB4, shown in the last panel of (D). The
averages for targets (used to determine thresholds) are calculated based on all successful differentiations that
proceeded to D40. Scale bars = 500m.
A B
A 38/42 normalized to control rostral

A42/40 normalized to control rostral
2.5 1.5
**** *
***
2.0 *
1.0
1.5
1.0
0.5
0.5
53 56 36 39 53 56 36 39
0.0 0.0
rostral caudal rostral caudal rostral caudal rostral caudal
control fAD control fAD
Figure S2. Aβ42/40 and Aβ38/42 ratio in pooled control and fAD lines. Related to Figure 2.
The data shown here is the same data from Fig. 2C and 2D, but pooled and normalized in a manner that allows the
fAD to control differences to be apparent. The Aβ42/40 ratio is shown in rostral and caudal cultures (A) and the
Aβ38/42 ratio is shown in rostral and caudal cultures (B). The ‘n’ for each line is shown in the corresponding bar.
Both ratios are normalized to the rostral control group and were calculated by paired experiments of control and
fAD samples. For this reason, BR11 and BR14 lines are not included here, since they were not paired with fAD
cultures. Rostral cultures are shown in white and caudal cultures are shown in teal. For each comparison a one-way
ANOVA with a Tukey’s multiple comparisons post-test was performed, *p<0.05, ***p<0.001, ****p<0.0001. Error
bars represent mean  SEM. Nine independent rounds of differentiation are represented.
Figure S3. Forced expression of APPV717I in control rostral and caudal cultures corroborates findings with
endogenous APPV717I mutation. Related to Figure 2.
Following an embryoid aggregate differentiation, control rostral and caudal neuronal cultures were transduced with
an APPV717I virus or mock transduced at day 40 and then harvested at day 47 (A-E) or else at day 100 (F).
Conditioned media from day 40 rostral and caudal cultures were analyzed for Aβ38, 40, and 42, measured by Aβ
triplex ELISA. (A) Levels of Aβ38, 40, and 42 are shown normalized to total protein. The ratios of Aβ 42:40 (B)
and 38:42 (C) are shown. The ‘n’ for each rostral/caudal pair is listed as the following: no virus – 4/4; APPV717I –
3/3. Rostral and caudal cultures are indicated by white and teal bars, respectively. For each comparison a one-way
ANOVA with a Tukey’s multiple comparisons post-test was performed, * p<0.05, ** p<0.01, **** p<0.0001. Data
are represented as mean  SEM. (D) WB for day 47 samples showing the overexpression of APP with APPV717I
lentiviral transduction, as well as levels of Tau and GAPDH. (E,F) Lysates from d47 (E) or d100 (F) were run on a
Phospho (Thr 231)/Total Tau MSD ELISA and phospho-Tau normalized to total Tau. The ‘n’ for each rostral/caudal
pair is listed as the following: no virus – 4/4; APPV717I – 3/3. For each comparison, a one-way ANOVA with
Tukey’s multiple comparison’s test was performed, * p<0.05. Data are represented as mean  SEM.
Figure S4. Characterization of APPV717I (fAD) and its isogenic control (fADcorr), in NGN2-induced neuron
(iN) cultures. Related to Figure 2.
(A) Schematic outlining the targeting constructs for correcting the APPV717I mutation in iPSC line fAD 2B. The
mutation in codon 717 of APP is flanked by the gRNA targeting sites, denoted by two red arrows in exon 17. The
mutation was corrected from an isoleucine to a valine, using a CRISPR/Cas technology. Two corrected clones were
isolated and used to generate the data here. Following confirmation of correct targeting, the puromycin selection
cassette was removed following Cre-mediated excision (loxp sites represented by black arrowheads). TK=
thymidine kinase selection cassette, DTA=diphtheria toxin selection cassette. The “fAD” iPSC lines used in these
experiments are two clones of non-targeted iPSCs that were isolated in parallel to the two corrected lines. (B) fAD
and its isogenic control (fADcorr) iPSC lines were transduced with the Ngn2 virus and then differentiated using a
protocol from the Sudhof lab (Zhang et al., 2013). Scale bars = 100m. (C) qPCR for neuronal markers APP, MAPT
and synaptophysin (SYP). All targets are normalized to GAPDH and are expressed in arbitrary units (A.U.). n=3 for
each group. For each comparison a Student’s t-test was performed with no significant differences identified. Data
are represented as mean  SEM. (D) Conditioned media from both fAD and fADcorr iNs were was analyzed for
Aβ38, 40, and 42 by Aβ triplex ELISA at d21 of differentiation. Levels of Aβ38, 40, and 42 are shown normalized
to total protein. Data shown are the same data from Fig. 2E, graphed separately. n=12 for each group over four
differentiations. For each comparison a Student’s t-test was performed, ** p<0.01. Data are represented as mean 
SEM.
A A ND
80 40 High A40
within each group
within each group

High A42
% of total cells
% of total cells
60 30
40 20
20 10
* * ND * ND
0
* 0
CUX1 SATB1 FEZF2 FOXG1 RELN TBR1 HOXB13 HOXB4 HOXB6 ISL1 MNX1
cortical/forebrain cerebellar/spinal cord
100 30 80
within each group
within each group

within each group
% of total cells
% of total cells
% of total cells
80
60
20
60
40
40
10
20
20
0 0
* * 0 * *
DLG4 SYN1 MAPT SLC17A7 GRIN1 GRIK1 SLC32A1 GAD1 NKX2-1 CALB2
general neuronal excitatory inhibitory
B C
2.5 2.0
*** ***
****
2.0
1.5
*
Relative expression
Relative expression
1.5
**** ****
** ***
1.0
1.0
0.5
0.5
0.0 0.0
Cb
Cb
Cb
Cb
Temp Cx
Temp Cx
Temp Cx
Temp Cx
caudal
caudal
caudal
caudal
rostral
rostral
rostral
rostral
GAD1 SLC32A1 SLC17A7 SLC17A6 GAD1 SLC32A1 SLC17A7 SLC17A6
Figure S5. Single-cell NanoString analysis of cells following microengraving and isolation from the nanowells
and expression of excitatory and inhibitory markers in iPSC-derived cultures and post-mortem tissue.
Related to Figure 4 and Table S2.
(A) Following printing (Fig. 4), individual cells were stained with trypan blue to mark cells with intact membranes,
retrieved from nanowells, lysed, cDNA synthesized and amplified using multi-target amplification, and hybridized
to a custom NanoString codeset. Cells expressing high levels of Aβ42 (n=10), Aβ40 (n=21), and no detectable Aβ40
or 42 (n=16) were analyzed. For each cell, genes with values above 1000 counts were considered to be expressed.
Values shown are the percentage of total cells within each Aβ population expressing each listed gene. ND=Not
detectable (plain), high Aβ40 (diagonal lines) and high Aβ42 (checkered). Data is presented in Table format in
Supplemental Table S2. (B) NanoString analysis of genes expressed in excitatory and inhibitory neurons is shown
comparing rostral and caudal cultures. RNA was harvested from day 40 cultures and analyzed using a custom
NanoString codeset. Data shown here include three control lines and three fAD lines. n=10 for each group. Caudal
data were normalized to the average rostral data for each gene. (C) Publically available RNA-seq analyses (Mayo
Clinic Brain Bank) from post-mortem human temporal cortex (Temp Cx) and cerebellum (Cb) is shown for
excitatory and inhibitory markers. n=275 and 276 for Temp Cx and Cb, respectively. For each gene, a two-tailed t-
test was performed, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Error bars represent SEM.
Funding acknowledgement for Mayo Clinic Brain Bank

The Mayo RNAseq study data was led by Dr. Nilüfer Ertekin-Taner, Mayo Clinic, Jacksonville, FL as part of the
multi-PI U01 AG046139 (MPIs Golde, Ertekin-Taner, Younkin, Price) were provided using samples from the
following sources:
 The Mayo Clinic Brain Bank. Data collection was supported through funding by NIA grants P50
AG016574, R01 AG032990, U01 AG046139, R01 AG018023, U01 AG006576, U01 AG006786, R01
AG025711, R01 AG017216, R01 AG003949, NINDS grant R01 NS080820, CurePSP Foundation, and
support from Mayo Foundation.
 Sun Health Research Institute Brain and Body Donation Program of Sun City, Arizona. The Brain and
Body Donation Program is supported by the National Institute of Neurological Disorders and Stroke (U24
NS072026 National Brain and Tissue Resource for Parkinson’s Disease and Related Disorders), the
National Institute on Aging (P30 AG19610 Arizona Alzheimer’s Disease Core Center), the Arizona
Department of Health Services (contract 211002, Arizona Alzheimer’s Research Center), the Arizona
Biomedical Research Commission (contracts 4001, 0011, 05-901 and 1001 to the Arizona Parkinson's
Disease Consortium) and the Michael J. Fox Foundation for Parkinson’s Research.
200
*
175
total Tau/MAP2 (% control)
150
125
100
75
50
25 4 6 6 6
0
rostral caudal rostral caudal
control fAD
Figure S6. Total Tau normalized to MAP2 to correct for relative neuron/non-neuronal cell proportions.
Related to Figure 5.
Control and fAD rostral and caudal neurons at 100 days of differentiation were lysed, and total Tau levels were
measured by Western blotting and densitometry. Here, quantification by densitometry was normalized to MAP2.
The ‘n’ for each data set is shown within the bar-graph column. There are two lines represented for both control and
fAD lines, over three independent rounds of differentiation. For each comparison a Student’s t-test was performed, *
p<0.05. Data are represented as mean  SEM.
WT rostral WT rostral WT rostral WT rostral WT rostral WT rostral WT caudal WT caudal WT caudal WT caudal
APLP1 6990 6123 6842 6815 2412 3317 5794 5747 5957 5803
APLP2 7644 7198 6665 7497 2812 3550 12067 10096 10609 10049
ADAM10 833 909 787 905 681 911 788 788 798 834
ADAM19 66 102 88 168 68 104 110 91 142 116
ADAM9 1443 1311 1359 1284 1252 1828 2932 3011 2873 3105
TACE/ADA
M17 724 612 562 621 410 443 969 942 912 886
BACE1 773 747 804 852 855 1190 967 940 1015 910
BACE2 150 163 154 152 303 431 186 225 186 226
PSEN1 1201 1035 1022 1034 325 517 1093 971 1014 953
PSEN2 172 162 208 238 70 75 169 210 240 190
Nicastrin 834 787 672 782 971 1434 1022 894 952 916
PEN-2 574 654 704 770 1071 1376 880 713 867 880
aph1Avar1 597 616 494 631 941 1026 671 723 565 640
aph1Avar2 727 610 679 598 788 857 787 674 766 778
aph1Avar3 339 308 303 338 173 242 337 295 340 329
aph1Avar4 246 198 194 134 503 615 168 158 177 168
aph1Bvar1 266 267 262 252 384 455 289 250 263 270
aph1Bvar2 5 6 14 7 25 32 5 2 6 2
WT rostral WT rostral WT rostral WT rostral WT rostral WT rostral WT caudal WT caudal WT caudal WT caudal
B2M 4892 5045 4231 6441 20205 17611 15008 12549 12917 11895
GAPDH 38068 37180 39924 36092 40270 43218 33964 35895 36788 37155
Supplemental Tables
GUSB 97 86 105 128 40 76 141 138 118 156

HPRT1 1270 1477 1465 1462 1875 1710 1105 1260 1373 1360
MAP2 7069 6950 9876 8756 4656 5834 5886 6169 6739 6796
POLR2A 2248 1929 1782 2193 715 856 1833 1894 1735 1775
RPL13a 3165 3364 2504 2952 2044 2490 1921 2177 1717 1747
RPL27 23191 23969 20113 21975 10196 8205 20142 19918 18613 19116
WT caudal WT caudal WT caudal fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral
5793 5490 7222 5034 4575 4994 3588 4284 6519 7094 6879 6632
11453 4412 7031 5271 5417 7732 4188 3954 11539 11190 11349 11864
770 745 1165 1076 1063 1333 839 833 883 777 751 818
147 69 70 38 55 150 44 105 139 131 128 118
3113 1756 3253 3861 4022 4017 1739 2134 2363 2650 2680 2836
859 631 836 1905 1729 1891 504 577 759 643 680 800
954 1524 1891 1590 1942 2728 1394 1779 1286 1226 1261 1173
217 252 331 1166 1088 1472 14 27 1 1 1 1
1010 707 905 2420 2481 3107 699 680 1368 1254 1213 1193
226 115 98 143 186 197 72 96 264 219 255 245
947 1522 2919 3364 3675 3504 1414 1408 968 937 925 1038
897 1511 1930 1722 1720 2176 1021 1065 833 703 793 821
627 1001 1093 2043 2290 2203 749 839 646 648 702 714
707 941 874 1496 1281 1673 703 736 898 802 800 775
318 274 283 652 642 681 216 191 399 385 359 395
131 817 860 1382 1401 1576 667 819 250 274 257 139
239 409 570 528 655 830 475 496 281 245 278 289
2 42 43 256 183 204 44 61 14 12 12 1
WT caudal WT caudal WT caudal fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral fAD rostral
12996 14257 12235 21619 22060 22123 14561 18823 17656 19718 20852 20539
36156 46421 50867 34109 33680 33410 47628 44247 30243 31081 29566 28191
148 67 114 222 223 261 120 84 104 83 92 79
1321 2189 1697 892 1357 1166 1261 1450 1365 1454 1505 1423
6615 6734 6554 1447 1647 2725 5669 6436 7889 7866 8388 7550
1849 1186 1750 1582 2224 1603 880 724 1399 1500 1479 1968
1697 1603 1727 3291 2905 2424 2012 1280 1732 1281 1220 1449
19218 7543 5055 16838 15904 16288 7869 6956 19612 17016 16898 18801
fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal
Table S1. Individual values from the NanoString data set shown in Fig. 3D. Related to Figure 3. Counts are
normalized across multiple code sets manually by using the mean of the housekeeping genes (shown in red text).
7193 6697 6280 6141 6197 5371 3849 15495 16381 15089
16470 15086 15245 15076 15401 5191 3545 4296 3158 3686
1027 876 910 860 921 701 580 984 767 859
179 133 148 143 152 57 62 37 17 26
4721 5177 5363 5318 5591 3047 2500 1265 942 1208
1128 1081 1222 1084 1113 743 521 1876 1765 1897
1356 1490 1552 1416 1446 2245 2001 2093 1632 1864
1 1 1 1 1 24 21 244 127 128
1277 1237 1245 1213 1302 843 701 2743 2607 2647
259 267 289 271 299 149 99 342 315 316
1591 1404 1379 1427 1386 1663 1096 2426 2472 2589
1097 1088 1144 1159 1245 1154 1074 2396 3134 2649
1043 898 879 844 876 596 521 2162 2819 2418
938 901 965 960 964 840 709 1357 1041 952
460 480 479 440 440 170 196 590 733 573
187 176 189 187 213 607 631 1417 1539 1291
385 324 351 319 340 361 338 929 757 821
16 1 12 1 10 29 70 178 112 137
fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal fAD caudal
22098 22549 22152 22230 22349 17281 23752 5276 5809 6160
31011 32042 31821 32164 32455 48474 42952 42752 42845 44772
204 163 168 171 189 43 56 125 111 95
1216 1346 1358 1356 1391 1144 1141 1354 1915 1676
5997 5858 6125 5820 5254 6919 5189 10299 6743 7645
1599 1554 1602 1516 1536 1129 572 2649 2972 3418
1062 944 1013 952 989 1266 1093 3742 3187 2804
16812 15544 15761 15791 15836 3745 5245 13803 16419 13431
A ND High A40 High A42
CORTICAL/FOREBRAIN
CUX1 67% 71% 30%
SATB1 40% 62% 30%
FEZF2 20% 29% 40%
FOXG1 7% 19% ND
RELN 33% 48% 10%
TBR1 27% 24% 40%
CEREBELLAR/SPINAL CORD
HOXB13 7% 14% ND
HOXB4 20% 29% ND
HOXB6 ND ND ND
ISL1 7% 10% ND
MNX1 ND ND ND
EXCITATORY
SLC17A7 7% 5% 10%
GRIN1 ND 10% ND
GRIK1 27% 14% 20%
INHIBITORY
SLC32A1 13% 38% 30%
GAD1 33% 62% 40%
NKX2-1 33% 14% ND
CALB2 7% 19% ND
GENERAL NEURONAL
DLG4 73% 95% 70%
SYN1 27% 81% 40%
MAPT 47% 57% 70%
Table S2. Percentage of single-cells within each group expressing various NanoString markers. Related to
Figure 4 and Figure S5. ND = not detected.
Spike properties
WT rostral WT caudal
n n
Mean 95%CI/SEM (wells) Mean 95%CI/SEM (wells)
Firing
0.10 0.02-0.16 19 0.12 0.06-0.19 21
frequency (Hz)
Burst count
84.5 40.9 19 113.8 43.1 21
per well
Spike properties
fAD rostral fAD caudal
n n
Mean 95%CI/SEM (wells) Mean 95%CI/SEM (wells)
Firing
0.14 0.07-0.20 29 0.39 0.11-0.68 42
frequency (Hz)
Burst count
30.9 13.7 29 277.6 117.1 42
per well
Spike waveform properties

WT rostral WT caudal
n n
Mean SEM (clusters) Mean SEM (clusters)
Amplitude
17.83 0.073 116 16.05 0.048 121
(uV)
Peak-valley
11.13 0.074 116 10.73 0.067 121
interval (ms)
Non-linear
2.04 1.9E-03 116 2.02 1.8E-03 121
energy (log10)
Spike waveform properties

fAD rostral fAD caudal
n n
Mean SEM (clusters) Mean SEM (clusters)
Amplitude
15.04 0.022 193 17.00 0.040 221
(uV)
Peak-valley
6.84 0.028 193 7.93 0.029 221
interval (ms)
Non-linear
2.08 1.2E-03 193 2.12 1.5E-03 221
energy (log10)
Table S3. Multi-electrode array data of rostral and caudal cultures reveals no detectable differences in
spontaneous activity between rostral and caudal cultures. Related to Figure 4.
Spike frequency and spike voltage waveform parameters derived from spontaneous firing of cultures. Metrics
calculated from integration of 10, 30 min recordings of cultures performed over a 20 day period (d37 to d57). Firing
frequency (Hz) determined by total number of spike events per second of recording; burst count represents total
number of ‘bursts’ determined by poisson surprise algorithm (Legéndy and Salcman, 1985). n represents replicate
wells (8 electrodes per well) of each culture. Waveform properties determined from reconstruction of momentary
voltage recordings sampled at 12.5kHz, followed by unsupervised clustering of waveform shapes of all spike events
recorded by an individual electrode. Statistics calculated from the waveforms representing the centroid of each
cluster. n represents individual clusters identified. Individual cluster sizes range from 10 - 6228 spike events.
Supplemental Experimental Procedures:
Study Design
The primary research objective of this study was to differentiate control and fAD iPSC lines to rostral and caudal
neural cultures and examine AD-relevant phenotypes such as APP processing and Tau alteration, in order to study
the differential vulnerability observed in AD subjects. Sample size was selected based on availability of lines and
differentiation capacity and efficiency (Fig. S1). Typically, 10 wells of a 96 well-plate, for both rostral and caudal
neurons, were plated per differentiation experiment. Multiple differentiations (as noted in figure legends, over 100
across the entire study) were conducted in order to assess 6 control and 3 fAD lines. Samples were excluded based
on the metrics outlined in Fig. S1. At 40 days of differentiation, qPCR was performed in order to confirm expression
of neuronal markers on a well-to-well basis. Samples that met the selection criteria were then subjected to a second
metric of either rostral or caudal marker expression. Only samples that met these selection criteria were included in
subsequent analyses. Replicates, experimental ‘n’ and iPSC lines used are noted in the figure legends.
Cell Lines
fAD iPSCs were generated in collaboration with the Harvard Stem Cell Institute as previously reported (Muratore et
al., 2014a). For WT lines the following previously published and commercially available lines were differentiated:
HDFa-YK26 (Martins-Taylor et al., 2011) and IMR90-TZ1 and IMR90-YZ1 (Zeng et al., 2010) (UCONN Stem
Cell Core). BR01, BR11 and BR14 were generated and characterized in collaboration with the New York Stem Cell
Foundation (NYSCF) using previously described methods (Paull et al., 2015). BR01, BR11 and BR14 were all
derived from Caucasian female donors, BR01 was not cognitively impaired at death at age 89, BR11 was not
cognitively impaired at death at age 79, BR14 had a diagnosis of late-onset AD at death at age 90.
iPSC Culture Conditions

iPSCs were cultured in medium consisting of 400 mL Dulbecco’s Modified Eagle’s Medium: Nutrient Mixture F12
(DMEM/F12, Gibco), 100 mL Knockout Serum Replacement (Gibco), 5 mL MEM-NEAA (Invitrogen), 5 mL of
Penicillin/Streptomycin/Glutamine (Invitrogen) and 500 μL 2-Mercaptoethanol (100x) (Invitrogen). FGF2
(Millipore) was added fresh daily at 10 μg/ml (1000x). Cells were maintained at 37°C/5% CO2 and were split as
necessary based on colony growth (5-6 days). Differentiating colonies were removed from the plate prior to
splitting. iPSCs were maintained on a mouse embryonic fibroblast (MEF) feeder layer at 1.7-2.0 x105 cells/well
(GlobalStem). Karyotyping was routinely performed using the NanoString nCounter CNV CodeSet.
Embryoid Body Neuronal Differentiation

For the induction of neurons (rostral and caudal), iPSCs were differentiated using an embryoid body-based protocol
(Zeng et al., 2010) that was further optimized (Muratore et al., 2014b). In brief, iPSC colonies were dissociated from
MEFs and cultured as aggregates in suspension with iPSC media (no FGFs). Aggregates were plated on Matrigel-
coated (BD Biosciences) culture dishes at day 7 with Neural Induction media (N2). At day 17 neural rosettes were
selected using STEMDiff Neural Rosette Selection reagent (Stemcell Technologies) and further cultured in
suspension using Neural Induction media (N2/B27 with cAMP (Sigma) and IGF-1 (PeproTech)). For final plating,
cells were dissociated to single cells with accutase (Innovative Cell Technologies) and plated on Matrigel in Neural
Differentiation media (N2/B27) with ROCK inhibitor (10uM) (Millipore) and cAMP (1 M) (Sigma), BDNF,
GDNF, and IGF1 (10 ng/mL) (PeproTech). For directing cells to caudal fates, iPSCs were differentiated using the
aggregate protocol described above. The aforementioned differentiation medias were used, with the addition of
retinoic acid (100mM) (Sigma) and Sonic Hedgehog (100g/mL) (PeproTech) or purmorphamine (Calbiochem)
from days 10-24 and 15-24 (Hu and Zhang, 2009), respectively.
Neurogenin 2 Induced Neuron (Ngn2-iN) Differentiation

iPSCs were transduced with 3 lentiviruses on day 0 (Maherali et al., 2008; Vierbuchen et al., 2010; Zhang et al.,
2013): Ngn2 (pTet-O-Ngn2-puro), rtTA (FUdeltaGW-rtTA) and GFP (Tet-O-FUW-EGFP). FUdeltaGW-rtTA was a
gift from Konrad Hochedlinger (Addgene plasmid # 19780). Tet-O-FUW-EGFP was a gift from Marius Wernig
(Addgene plasmid # 30130). pTet-O-Ngn2-puro was a gift from Marius Wernig (Addgene plasmid # 52047).
Briefly, iPSCs were differentiated following the protocol in (Zhang et al., 2013). Cultures were treated with
doxycycline (2ug/mL), day 1, to induce differentiation and fed with a series of media changes. On day 2, cultures
were treated with puromycin (10mg/mL) (Invitrogen) to select for induced cells. At day 4, cells were dissociated
from the plate using Accutase (diluted 1:30 in PBS + ROCK inhibitor (10uM) (Millipore)) and plated in 96-well
plates with Neuralbasal media (NBM) + BDNF, GDNF, and CNTF (10 ng/mL) (PeproTech). Cells were maintained
until day 28 with NBM + growth factors and fed every 2-3 days.
Medias
MEF Medium- 435 mL DMEM (Invitrogen), 5 mL 100x Penicillin/Streptomycin (Invitrogen), 5 mL 100x L-
glutamine (Invitrogen), 50 mL FBS (Invitrogen)
iPSC Medium- 390 mL DMEM/F12 (Invitrogen), 100 mL KOSR (Invitrogen), 5 mL 100x
Penicillin/Streptomycin/Glutamine (Invitrogen), 5 mL 100x MEM-NEAA (Invitrogen), 50 μM b-mercaptoethanol
(Invitrogen), with the addition of fresh FGF2 (Millipore, 10 ng/mL) to the medium
N2 Neural Induction Medium- 490 mL DMEM/F12, 5 mL N2 supplement (Invitrogen), 5 mL 100x MEM-NEAA,
and Heparin (Sigma-Aldrich, 2 μg/mL)
N2/B27 Neural Induction Medium- 480 mL DMEM/F12, 5 mL N2 supplement, 10 mL B27 supplement
(Invitrogen), 5 mL MEM-NEAA and 2 μg/ml Heparin (Sigma-Aldrich), with the addition of fresh cAMP (1 μM)
(Sigma) and IGF1 (PeproTech, 10 ng/mL) to the medium
Neural Differentiation Medium- 490 mL Neurobasal medium (Invitrogen), 5 mL N2 supplement (Invitrogen), 5 mL
100x MEM-NEAA, and 10 mL B27 supplement (Invitrogen), with the addition of fresh cAMP (1 μM), BDNF,
GDNF, and IGF-1 (all 10 ng/mL) to the medium.
Human KSR media (day 1 and 2)- 415 mL KO DMEM (Invitrogen), 75 mL KOSR, 5 mL, MEM-NEAA, 0.5 mL
beta-mercaptoethanol (Invitrogen), 5 mL Glutamax (Gibco)
N2 media (day 2)- 500 mL DMEM/F12, 5mL Glutamax, 7.5mL 20% Dextrose (Stem Cell Technologies), 5 mL
N2B supplement (Stem Cell Technologies)
N2 + B27 media (day 3)- N2 media + 1:100 B27 (50X) (B27), B27 should be added just prior to use
Neurobasal Medium (day 4) - 485 mL NBM (Invitrogen), 5 mL Glutamax, 7.5 mL 20% Dextrose, 2.5mL MEM-
NEAA, B27 should be added just prior to use, with the addition of fresh BDNF, GDNF, and IGF-1 (all 10 ng/mL) to
the medium.
Western Blots
Lysates were prepared with standard lysis buffer containing 1% NP40, 0.5M EDTA, 5M NaCl, 1M Tris and
cOmplete protease inhibitors and phosSTOP (Roche). Lysates were protein normalized, and samples were loaded
onto 4–12% Bis-Tris gels using MES or MOPS running buffer (Invitrogen), transferred to nitrocellulose
membranes, and probed for various proteins using standard WB methods. The resultant blots were detected with
either an infrared imaging system (Odyssey; LI-COR Biosciences) or by ECL and exposure to film.
Antibodies
Antibody Vendor Concentration
APP (C7) Selkoe Lab 1:1000
EN1 DSHB (#4G11) 1ug/mL
GAPDH Millipore (MAB374) 1ug/mL
GFAP abcam (#ab4674) 33ug/mL
HoxB1 R&D Systems (#AF6318) 5ug/mL
PCRP-HOXB2 DSHB (#1C9) 5ug/mL
Anti-HOXB4 DSHB (#I12) 5ug/mL
PCRP-HOXB6 DSHB (#1A4) 5ug/mL
PCRP-HOXB9 DSHB (#1E6) 5ug/mL
MAP2 abcam (#ab5392) 1.9ug/mL
NeuN (A60) Millipore (#MAB377) 1ug/mL
Olig2 (C-17) Santa Cruz (#sc-19969) 4ug/mL
pTau (AT8) Thermo Fisher Scientific/Pierce (#MN1020) ICC: 5ug/mL, WB: 20ug/mL
Tau (K9JA) DAKO (#A-0024) 55ug/mL
Tbr1 abcam (#ab31940) 5ug/mL
TuJ1 Sigma (#T2200) 1ug/mL
PCRP-HOXB9-1E6, PCRP-HOXB6-1A4, PCRP-HOXB2-1C9 were deposited to the DSHB by Protein Capture

Reagents Program, produced by JHU/CDI. I12 anti-Hoxb4 was deposited to the DSHB by Gould, A. / Krumlauf, R.
(DSHB Hybridoma Product I12 anti-Hoxb4). 4G11 was deposited to the DSHB by Jessell, T.M. / Brenner-Morton,
S. (DSHB Hybridoma Product 4G11). Secondaries were from Jackson ImmunoResearch: anti-chicken cy2/cy3/cy5,
anti-rabbit cy2/cy3, anti-mouse cy2/cy3, anti-rat cy2/cy3 (1:1000). DAPI 1:1000 Invitrogen.
qPCR Primer Sequences

MAP2 Forward: AACCGAGGAAGCATTGATTG
MAP2 Reverse: TTCGTTGTGTCGTGTTCTCA
MAPT Forward: GCACCCCGTCCCTTCCAACC
MAPT Reverse: GGCGGCTCTTGGCGGAAGA
APP Forward: AACCAGTGACCATCCAGAAC
APP Reverse: ACTTGTCAGGAACGAGAAGG
SYP Forward: CTCAGCATCGAGGTCGAGTTC
SYP Reverse: GAGGAGTAGTCCCCAACTAAGAA
TBR1 Forward: TCACCGCCTACCAGAACAC
TBR1 Reverse: GTCCATGTCACAGCCGGT
HOXB4 Forward: AAAAGAGCCCGTCGTCTACC
HOXB4 Reverse: GTGTAGGCGGTCCGAGAG
GAPDH Forward: GGGAGCCAAAAGGGTCATCA
GAPDH Reverse: TGGTTCACACCCATGACGAA
RNA Analyses
RNA Sequencing: Total RNA from rostral and caudal cultures (4 per condition) were assayed for quality on an
Agilent TapeStation 4200. Samples with an electronic RNA Integrity Number (eRIN) of at least 8.0 were polyA
selected using oligo-dT Dynabeads (Thermo Fisher). Double stranded cDNA was synthesized using the SuperScript
III reverse transcriptase protocol using random hexamers on the polyA-RNA eluate. After purifying the cDNA with
AMPure magnetic beads (Agencourt) and quantitation on the TapeStation, libraries were generated by processing
1ng of the double stranded cDNA product through the Illumina Nextera Tagmentation library protocol to fragment
the DNA and add overhangs for barcode (i5 and i7) annealing. Each of the 8 amplified and barcoded libraries was
quality controlled for concentration and integrity (Agilent TapeStation), normalized to 1ng/ul and combined in equal
concentrations into a single 8x multiplexed library. Multiplexed libraries were sequenced on an Illumina NextSeq
500 to a depth of at 433 million paired-end reads (75 bases per read) total. RNAseq reads were run through an
analysis pipeline that included the following steps: fastQC (v0.10.1) to assay the quality of reads; trimmomatic
(v0.33) to remove end bases with phred33 quality scores below 25 and to eliminate resulting reads that are shorter
than 30 bases (Bolger et al., 2014); kallisto (v0.42.4) for pseudoalignment and quantification program (Bray et al.,
2016) running 100 bootstraps against a kallisto index generated from ChGR38p10 with a k-mer length of 31. Hox
gene-level expression data (Transcripts Per Kilobase Million, TPM) were extracted from the expression table and
the mean plotted as a heat map. Quantitative PCR (qPCR): Each sample was assayed in 3 technical replicates. Data
were analyzed using the ΔΔCT method and expression was normalized to GAPDH expression (Livak and
Schmittgen, 2001). Primer efficiency was calculated for each pair of primers and the slope of the dilution line was
found to be within the appropriate range. Dissociation curves also showed single peak traces, indicating template-
specific products. NanoString: Hybridization reactions were carried out with 100-200 ng RNA for pooled culture
samples (single cell data was derived from amplified cDNA). Post-hybridization, samples were processed with the
nCounter Prep-station. Following run completion, the cartridge was scanned using the nCounter Digital Analyzer, at
max resolution (~1000 images/sample). Data were analyzed using the nSolver Analysis Software as follows: 1)
background subtracted by negative control subtraction (mean of negative controls + 2 STDEVs), 2) normalized to
mean of positive controls and 3) normalized to a set of 8 house-keeping genes (B2M, GAPDH, GUSB, HPRT1,
POLR2A, RPL13a, RPL27 and MAP2) or to the total gene set, as noted.
Immunocytochemistry and Microscopy

Cultures were fixed with 4% paraformaldehyde, followed by membrane permeabilization with 0.1% TritonX-100
and then immunostaining with primary and secondary antibodies. Imaging was performed using a Zeiss LSM710
confocal microscope and images were acquired using ZEN black software.
CRISPR-Cas targeting of APPV717I

For CRISPR-Cas9 targeting, unique genomic sites in the APP locus adjacent to the V717I mutation were targeted
using guide RNA (ATCATTGGACTCATGGTGGGCGG; TTTCTTCTTCAGCATCACCAAGG), Cas9 under the
CAGGS promoter, and a donor template plasmid for homologous recombination. The donor plasmid contained wild
type APP exon 17, adjacent homology arms, and positive and negative selection cassettes (see Fig. S4A). Targeting
plasmids were transfected into dissociated (Accutase) fAD 2B iPSCs in MeTSR in the presence of ROCK inhibitor
using the Amaxa 4D-Nucleofector X unit (Lonza). 72 hr after transfection, iPSCs were cultured in selection media.
Following selection for several weeks, single colonies were picked manually. Individual clones were screened for
correction by PCR amplification of genomic DNA around the target site, followed by Sanger sequencing.
Microengraving/ Printing
The nanowell array is fabricated on a biocompatible PDMS (polydimethylsioxane) affixed to a glass microscope
slide. Each array has 84,672 wells with 50x50x50µm dimensions. Prior to cell plating, PDMS arrays were incubated
in a vaccuum plasma cleaner at a high radio frequency for 2 min in order to make the surface hydrophilic for later
cell plating, and then placed into PBS. Aged differentiated control rostral and caudal neurons were dissociated with
accutase and then stained with live markers-calcein violet or calcein orange, respectively. Cells were quickly spun
down to wash out the extra live maker before mixing together and loading on PDMS. Cells are loaded from
suspension (1-2x105 cells/ml) by gravity, which favors 0-2 cells per well (examined in real-time under a
microscope). Non-adherent cells are removed through subsequent washing. The cells on PDMS were imaged using
an automated inverted epifluorescence microscope (Zeiss). Cell viability and numbers per well were analyzed with
Enumerator software (custom software, available upon request to JCL (Han et al., 2010; Ogunniyi et al., 2009)).
Capture antibody (6E10) (35 μg/ml) and human IgG (15 ug/ml) were coated onto a poly-lysine coated glass slide at
room temperature for 1 hr, and blocked with 5% BSA in PBS for 30 minutes. Before printing, the glass slide is
briefly incubated with human serum. Detection of human serum bound to human IgG allows for software alignment
of wells. The coated slide is used to seal the nanowell array plated with cells (clamped between plates using a
microarray hybridization chamber). The sealed PDMS array with antibody-coated slide was incubated at 37C for
overnight, followed by removal of the glass slide. Captured Aβ40 and Aβ42 on the glass slide were detected by
incubation with HRP-conjugated 2G3 antibody and Biotin-conjugated HJ7.4 antibody, respectively for 1 hr at RT,
followed by a fluorescently-labeled secondary antibody at RT for 1 hr. Fluorescent signals are detected using a
Genepix Pro slide scanner and aligned with Crossword (Gierahn et al., 2014). Secretion data from Crossword and
cell information from Enumerator are combined according the unique ID for each well. Wells were deemed analyte-
positive if the signal was one standard deviation above the mean and the intensity of >30% of the pixels within the
well area was greater than two standard deviations above the local mean background intensity.
AD TBS treatments
AD brain was obtained and used in accordance with the Partners IRB assurance. Informed consent was obtained
from the subject. The AD brain sample was from an 82-year-old male with numerous amyloid-beta plaques and
neurofibrillary tangles in multiple brain regions (cortex, hippocampus and amygdala). The subject also had moderate
brain atrophy in the frontal and parietal lobes. Brain homogenate was prepared as specified in (Shankar et al., 2008).
Frozen cerebral cortices were provided by Dr. M. Frosch (Massachusetts ADRC Neuropathology Core, MGH,
Boston, MA, USA) under institutional review board-approved protocols. Frozen samples of temporal cortex (1 g)
were allowed to thaw on ice, chopped into small pieces with a razor blade, and then homogenized with 25 strokes of
a Dounce homogenizer (Fisher) in 4 mL ice-cold 20 mM Tris HCl (pH 7.4) containing 150 mM NaCl and protease
inhibitors [called Tris-buffered saline (TBS)]. TBS-soluble Aβ was separated from membrane-bound and plaque Aβ
by centrifuging the homogenates at 175,000 g at 4°C in a TLA 100 rotor (Beckman Coulter) for 30 min, and the
supernatant (referred to as the TBS extract) was aliquoted and stored at −80°C.
Lentivirus Preparation
The coding sequences for mRFP and APPV717I were introduced into the pcDH-CB vector by Infusion cloning
(Clontech). HEK293FT cells were co-transfected with psPax2, pMD2.G, and pcDH-CB plasmids using
Lipofectamine 2000 (Invitrogen). Lentiviral conditioned media was harvested 24 and 36 hours post-transfection,
pooled, and concentrated 100-fold by centrifugation over 20% sucrose at 100,000 x g for 2 hours followed by
resuspension in fresh media. pCDH-CB (Addgene plasmid # 72267) was a gift from Kazuhiro Oka, and psPAX2
(Addgene plasmid # 12260) and pMD2.G (Addgene plasmid #12259) were gifts from Didier Trono.
Mayo Brain Gene Expression

Data is provided for the Mayo RNAseq Study through the AMP-AD Knowledge Portal (Synapse). From Synapse,
“whole transcriptome data for 275 Cerebellum (CBE) and 276 Temporal cortex (TCX) samples from 312 North
American Caucasian subjects with neuropathological diagnosis of AD, progressive supranuclear palsy (PSP),
pathologic aging (PA) or elderly controls (CON) without neurodegenerative diseases. Within this cohort, all AD
and PSP subjects were from the Mayo Clinic Brain Bank (MCBB), and all PA subjects were obtained from the
Banner Sun Health Research Institute (Banner). Thirty-four control CBE and 31 control TCX samples were from
the MCBB, and the remaining control tissue was from Banner. All subjects selected from the MCBB and Banner
underwent neuropathologic evaluation by Dr. Dennis Dickson or Dr. Thomas Beach, respectively. All ADs had
definite diagnosis according to the NINCDS-ADRDA criteria and had Braak NFT stage of IV or greater. Control
subjects had Braak NFT stage of III or less, CERAD neuritic and cortical plaque densities of 0 (none) or 1
(sparse) and lacked any of the following pathologic diagnoses: AD, Parkinson’s disease (PD), DLB, VaD, PSP,
motor neuron disease (MND), CBD, Pick’s disease (PiD), Huntington’s disease (HD), FTLD, hippocampal
sclerosis (HipScl) or dementia lacking distinctive histology (DLDH). Subjects with PA also lacked the above
diagnoses and had Braak NFT stage of III or less, but had CERAD neuritic and cortical plaque densities of 2 or
more. None of the PA subjects had a clinical diagnosis of dementia or “mild cognitive impairment.”
MEA Recordings
Spontaneous firing is measured using a multi-well MEA recorder (Maestro, Axion Biosystems) with 768 electrodes
(8 electrodes per well in a 96 well plate). All recordings were performed for 30min under environmentally
controlled conditions (37°C and 5% CO2). Potentials are recorded at a frequency of 12.5kHz with ‘spikes’ called
every time the voltage exceeds 5.5 standard deviations away from the root mean squared (RMS) of the background
potential calculated over a 10 ms moving window (“crossing-threshold”). Momentary voltage recordings are saved
for a 3 ms interval around each spiking event, as well as the value of the crossing-threshold at the time of the spike.
The total number of spiking events detected across all electrodes in the well over the duration of recording is used to
calculate the firing frequency for each well.
Here, iPSC-derived rostral and caudal cultures were plated on 96-well MEA plates at day 28 of
differentiation at a density of 500,000/well. Recordings were acquired every 2-3 days from d37 to d57. Data from all
recordings were integrated. Recording data was filtered in order to remove spike events detected by ‘high-noise’
electrodes, RMS voltage >= 3 µV at time of detection, and additionally data from wells failing to meet a minimum
activity level of 0.01Hz during a given recording were also excluded. Mean firing rate (MFR, Hz) was calculated as
the number of spikes per total duration of recording (in seconds). Bursting data was extracted from Axion software.
In order to reconstruct voltage waveforms, momentary voltage measurements were extracted from recording data
using MATLAB functions (MATLAB, The MathWorks Inc) provided by Axion Biosystems, while custom scripts in
the R programming language (R Core Team, 2017) were developed to calculate a series of waveform parameters
and to perform shape-based clustering analysis. Shape-based clustering analysis was performed on all spike events
detected by a given electrode. Thus, each electrode of the arrays may have multiple clusters assigned to it, each of
which may represent a different neuron spatially adjacent to the electrode. Shape-based clustering consisted of first
performing principal components analysis on the calculated waveform parameters followed by an unsupervised
mean-shift clustering (Ciollaro, Mattia and Wang, Daren, 2016) of spike events projected along the first two
principle components. Minor clusters containing few than 10 spike events were excluded from further analysis.
Waveform metrics of individual clusters were reported for the spike events representing the centroid of each cluster.
Metrics calculated from clustered waveform data include: amplitude form maximum hyperpolarization to maximum
depolarization (µV), Peak-Valley Interval (PVI), the time (µs) between the maximum and minimum potential (µV),
and Non-Linear Energy (NLE) of the waveform, a parameter that measure the ‘sharpness’ of waveform (Kim and
Kim, 2000)
Supplemental References
Bolger, A.M., Lohse, M., and Usadel, B. (2014). Trimmomatic: a flexible trimmer for Illumina sequence data.
Bioinformatics 30, 2114–2120.
Bray, N.L., Pimentel, H., Melsted, P., and Pachter, L. (2016). Near-optimal probabilistic RNA-seq quantification.
Nat. Biotechnol. 34, 525–527.
Ciollaro, Mattia, and Wang, Daren (2016). MeanShift: Clustering via the Mean Shift Algorithm. R package version
1.1-1. https://CRAN.R-project.org/package=MeanShift.
Gierahn, T.M., Loginov, D., and Love, J.C. (2014). Crossword: a fully automated algorithm for the segmentation
and quality control of protein microarray images. J. Proteome Res. 13, 362–371.
Han, Q., Bradshaw, E.M., Nilsson, B., Hafler, D.A., and Love, J.C. (2010). Multidimensional analysis of the
frequencies and rates of cytokine secretion from single cells by quantitative microengraving. Lab. Chip 10, 1391–
1400.
Hu, B.-Y., and Zhang, S.-C. (2009). Differentiation of spinal motor neurons from pluripotent human stem cells. Nat.
Protoc. 4, 1295–1304.
Kim, K.H., and Kim, S.J. (2000). Neural spike sorting under nearly 0-dB signal-to-noise ratio using nonlinear
energy operator and artificial neural-network classifier. IEEE Trans. Biomed. Eng. 47, 1406–1411.
Legéndy, C.R., and Salcman, M. (1985). Bursts and recurrences of bursts in the spike trains of spontaneously active
striate cortex neurons. J. Neurophysiol. 53, 926–939.
Livak, K.J., and Schmittgen, T.D. (2001). Analysis of relative gene expression data using real-time quantitative PCR
and the 2(-Delta Delta C(T)) Method. Methods 25, 402–408.
Maherali, N., Ahfeldt, T., Rigamonti, A., Utikal, J., Cowan, C., and Hochedlinger, K. (2008). A high-efficiency
system for the generation and study of human induced pluripotent stem cells. Cell Stem Cell 3, 340–345.
Martins-Taylor, K., Nisler, B.S., Taapken, S.M., Compton, T., Crandall, L., Montgomery, K.D., Lalande, M., and
Xu, R.-H. (2011). Recurrent copy number variations in human induced pluripotent stem cells. Nat. Biotechnol. 29,
488–491.
Muratore, C.R., Rice, H.C., Srikanth, P., Callahan, D.G., Shin, T., Benjamin, L.N.P., Walsh, D.M., Selkoe, D.J., and
Young-Pearse, T.L. (2014a). The familial Alzheimer’s disease APPV717I mutation alters APP processing and Tau
expression in iPSC-derived neurons. Hum. Mol. Genet. 13, 3523-3536.
Muratore, C.R., Srikanth, P., Callahan, D.G., and Young-Pearse, T.L. (2014b). Comparison and Optimization of
hiPSC Forebrain Cortical Differentiation Protocols. PLoS ONE 9, e105807.
Ogunniyi, A.O., Story, C.M., Papa, E., Guillen, E., and Love, J.C. (2009). Screening individual hybridomas by
microengraving to discover monoclonal antibodies. Nat. Protoc. 4, 767–782.
Paull, D., Sevilla, A., Zhou, H., Hahn, A.K., Kim, H., Napolitano, C., Tsankov, A., Shang, L., Krumholz, K.,
Jagadeesan, P., et al. (2015). Automated, high-throughput derivation, characterization and differentiation of induced
pluripotent stem cells. Nat. Methods 12, 885–892.
R Core Team (2017). R: A language and environment for statistical computing. R Foundation for Statistical
Computing, Vienna, Austria. https://www.R-project.org/.
Shankar, G.M., Li, S., Mehta, T.H., Garcia-Munoz, A., Shepardson, N.E., Smith, I., Brett, F.M., Farrell, M.A.,
Rowan, M.J., Lemere, C.A., et al. (2008). Amyloid β-Protein Dimers Isolated Directly from Alzheimer Brains
Impair Synaptic Plasticity and Memory. Nat. Med. 14, 837–842.
Vierbuchen, T., Ostermeier, A., Pang, Z.P., Kokubu, Y., Sudhof, T.C., and Wernig, M. (2010). Direct conversion of
fibroblasts to functional neurons by defined factors. Nature 463, 1035–1041.
Zeng, H., Guo, M., Martins-Taylor, K., Wang, X., Zhang, Z., Park, J.W., Zhan, S., Kronenberg, M.S., Lichtler, A.,
Liu, H.-X., et al. (2010). Specification of Region-Specific Neurons Including Forebrain Glutamatergic Neurons from
Human Induced Pluripotent Stem Cells. PLoS ONE 5, e11853.
Zhang, Y., Pak, C., Han, Y., Ahlenius, H., Zhang, Z., Chanda, S., Marro, S., Patzke, C., Acuna, C., Covy, J., et al.
(2013). Rapid Single-Step Induction of Functional Neurons from Human Pluripotent Stem Cells. Neuron 78, 785–
798.

MMC 2

Uploaded by

Copyright:

Available Formats

MMC 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MMC 2

Uploaded by

Copyright:

Available Formats

Stem Cell Reports

Cell-type Dependent Alzheimer’s Disease Phenotypes: Probing the Biology

INTRODUCTION tions of intrinsic (cell-autonomous) differences among

(legend on next page)

Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1869

Figure 1. Differentiation of Human iPSCs to Rostral and Caudal Neural Fates

1870 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

rostral for each line

no fAD mutation fAD no fAD mutation fAD

A 38/42 normalized to fADcorr

Ngn2-iN Ngn2-iN Ngn2-iN

Figure 2. APP is Differentially Cleaved in Rostral Versus Caudal Neural Cultures

Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1871

1872 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

1.5 3000 1750

APPs /total RNA

APPs /total RNA

control fAD control fAD control fAD

APH1Avar3 0.0 APP-all

(legend continued on next page)

Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1873

1874 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1875

1876 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

pTau/Total Tau (% control)

control fAD control fAD

25k 30k 35k 40k 25k 30k 35k 40k

(legend on next page)

1878 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1879

1880 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1881

iPSC Differentiation Microengraving and Printing

1882 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017 1883

1884 Stem Cell Reports j Vol. 9 j 1868–1884 j December 12, 2017

Cell-type Dependent Alzheimer's Disease Phenotypes: Probing the Bi-

A 38/42 normalized to control rostral

control fAD control fAD

within each group

within each group

GAD1 SLC32A1 SLC17A7 SLC17A6 GAD1 SLC32A1 SLC17A7 SLC17A6

Funding acknowledgement for Mayo Clinic Brain Bank

GUSB 97 86 105 128 40 76 141 138 118 156

Spike waveform properties

Spike waveform properties

iPSC Culture Conditions

Embryoid Body Neuronal Differentiation

Neurogenin 2 Induced Neuron (Ngn2-iN) Differentiation

PCRP-HOXB9-1E6, PCRP-HOXB6-1A4, PCRP-HOXB2-1C9 were deposited to the DSHB by Protein Capture

qPCR Primer Sequences

Immunocytochemistry and Microscopy

CRISPR-Cas targeting of APPV717I

Mayo Brain Gene Expression

You might also like