Process Biochemistry 51 (2016) 509–516

Contents lists available at ScienceDirect

Process Biochemistry
journal homepage: www.elsevier.com/locate/procbio

Clavulanic acid separation on fixed bed columns of layered double

hydroxides: Optimization of operating parameters using
breakthrough curves
Marcus Bruno Soares Forte a,∗ , Christine Taviot-Guého b,c , Fabrice Leroux b,c ,
Maria Isabel Rodrigues a , Francisco Maugeri Filho a
a
Bioprocess and Metabolic Engineering Laboratory (LEMeB), Food Engineering Department (DEA), Faculty of Food Engineering (FEA), University of
Campinas (UNICAMP), R. Monteiro Lobato, 80, Campinas, SP 13083-862, Brazil
b
Clermont Université, Université Blaise Pascal, Institut de Chimie de Clermont-Ferrand, BP 10448, F-63000 Clermont-Ferrand, France
c
CNRS, UMR 6296, ICCF, Aubiere F-63171, France

a r t i c l e i n f o a b s t r a c t

Article history: The adsorption of clavulanic acid (CA) in a fixed-bed column of layered double hydroxides (LDHME ) was
Received 12 June 2015 investigated. Breakthrough curves were obtained experimentally and the system was evaluated with
Received in revised form 25 January 2016 regards to column operation time, efficiency and productivity as functions of simultaneous variations
Accepted 29 January 2016
of superficial velocity (vz ) and bed height (L) using a central composite rotatable design (CCRD). At the
Available online 1 February 2016
optimized condition (vz = 1.00 cm/min and L = 6.5 cm), the responses were: 46 min, 146 min, 100 min and
13 kg/h cm3 , for breakthrough time (tb ), exhaustion time (te ), the difference te − tb and productivity (P),
Keywords:
respectively. These results represent no change in tb , but a 50% decrease of te , 40% decrease of (te − tb ) and
Clavulanic acid
Layered double hydroxides
38% increase of P, which is advantageous for the process. The factorial design technique was shown to
Adsorption be an efficient tool for assessing factor influences and was effective in optimizing the column operating
Separation conditions. Good separation of CA from the amino acids tyrosine (TYR) and proline (PRO) was observed
Purification in fixed bed columns. The LDHME adsorbent was shown to be an excellent alternative for CA purification.
Fixed bed © 2016 Elsevier Ltd. All rights reserved.
Breakthrough curves
Factorial design

1. Introduction known that this step can be responsible for at least 50% of total pro-
duction costs of bioproducts, and may exceed 90% for products with
Biotechnology is one of the areas which has grown intensively high added value [1–3]. Despite these striking facts, comparatively
in recent years. Worldwide efforts have been made in research few studies have been published on downstream processes.
and industrial development to acquire more efficient use of natural Clavulanic acid (CA) is a beta-lactam compound widely used in
resources in several applications, including energy, food, pharma- the pharmaceutical industry due to its great inhibition potential of
ceutical and agricultural. Researchers have spent time and money beta-lactamases, enzymes that hydrolyze beta-lactam antibiotics
in the search of new technologies and strategies for the develop- such as penicillin and amoxicillin [4]. After fermentation, the fer-
ment of bioprocesses, especially in the fermentation stage, ranging mented broth containing CA is clarified by centrifugation and/or
from isolation, selection, identification and genetic improvement of filtration. The clarified broth is then subjected to primary extraction
microorganisms to optimize the production and scale-up. However, stages involving liquid-liquid extraction followed by adsorption
separation and purification of bioproducts (downstream) remain processes [5]. However, this step provides relatively low yields
the most costly step both economically and technologically. Irre- since CA lacks highly hydrophobic groups [6] and is highly instable
spective of the organism involved, the synthesized material is at unsuitable temperature and pH [1,7].
composed of a complex mixture containing the target biomolecule Bioseparation processes can be classified into four large groups:
and various types of molecules, depending on the case. It is well removal of insoluble molecules, product isolation, purification
and polishing [8]. Adsorption is characterized by the transport
of a solute (adsorbate) diluted in solution to a solid (adsor-
bent). This technique has received special attention in bioprocesses
∗ Corresponding author: Fax: +55 19 3521 4027.
since biomolecules may be selectively adsorbed within a range of
E-mail address: [email protected] (M.B.S. Forte).

http://dx.doi.org/10.1016/j.procbio.2016.01.011
1359-5113/© 2016 Elsevier Ltd. All rights reserved.
510 M.B.S. Forte et al. / Process Biochemistry 51 (2016) 509–516

solid materials, allowing the exploitation of these properties in The concentrations of proline (PRO) were determined by reac-
extraction steps of the target molecule [5,9]. The use of adsorp- tion with ninhydrin in the presence of glacial acetic acid and
tion columns, operating either in continuous or batch mode, is phosphoric acid. An adaptation was made to the methodology
considered a low-cost process for concentration, separation and reported by Bates et al. [20] and Torello and Rice [21] for plant
purification of bioproducts [9]. Some authors have used adsorp- extracts: 0.5 mL of the sample was reacted with 0.5 mL of acid-
tion columns (fixed or expanded bed) with different adsorbents ninhydrin (1.25 g of ninhydrin diluted in 30 mL of glacial acetic
for studying the separation of various biomolecules, such as sugars acid and 20 mL of 6 M phosphoric acid; when stored at 4 ◦ C this
(fructose and glucose), antibiotics (cephalosporin), enzymes (inuli- reagent is stable for 24 h) and 0.5 mL of glacial acetic acid. The
nase) and bio-dyes (C-phycocyanin) using breakthrough curves as mixture was submitted to a temperature of 100 ◦ C for 1 h; the reac-
a methodology of injecting samples [2,9–12]. tion was stopped by immersing the samples in an ice bath and the
Layered double hydroxides (LDH) are mineral oxides and total concentration of PRO was determined by spectrophotometry
hydroxides of the class of non-silicates [13,14], composed of at  = 520 nm.
brucite-like layers of positively charged metal hydroxides. The use The analyses of tyrosine (TYR) were performed by chromatog-
of LDH as adsorbents of CA may be a good alternative, since under raphy, using an ultra-high performance liquid chromatography
suitable conditions for greater stability, the CA molecule has an device (UHPLC), model Accela (Thermo ScientificTM ), equipped
anionic nature [7,15]. Additionally, the major contaminants of the with a C-18 column (HypercarbTM , 2.1 × 30 mm, particle diam-
pre-purified fermentation broth from previous steps are amino eter of 3 ␮m) with mobile phase consisting of a mixture of
acids, such as tyrosine and proline, which are cationic under such KH2 PO4 (50 mM) at pH 4.5 (90% v/v) and acetonitrile (10% v/v),
conditions [5,16], and this increases the potential of LDH as a at the following conditions: column temperature of 25 ◦ C, sam-
selective adsorbent for CA in relation to these contaminants. In pre- ple temperature of 15 ◦ C, feed flow rate of 800 ␮L/min (maximum
vious works CA adsorption has been evaluated using a specific LDH pressure maintained at 500 bar), with detection by scanning in
(hydrotalcites) [1]. the UV spectrum (190 <  < 400 nm) at 3 fixed reading channels
The factorial design methodology is a statistical tool that can (1 = 193 nm, 2 = 227 nm and 3 = 280 nm). This method, with a
be very useful to select and/or optimize the most relevant vari- run time of 8 min, resulted in separation of three components (CA,
ables, while decreasing the number of assays in comparison to the TYR and PRO), where TYR was detected with best resolution at
trial and error method. It also permits for the evaluation of exper- 1 = 193 nm, with an average retention time of 4.5 min. The soft-
imental errors, which is an extremely important tool to obtain ware ChromoquestTM was used to integrate the results.
reliable results; it further combines greater speed and reduced
number of assays for researchers with specific goals [17,18]. Some
2.2. Syntheses of layered double hydroxides (LDH)
experimental design studies have shown the effectiveness of using
such a technique, including the central composite rotatable design
LDH were obtained by using the coprecipitation method, per-
(CCRD), for optimization of operating conditions of biomolecule
formed with the addition of di and trivalent cation salts in a solution
adsorption processes in both fixed and expanded bed columns
of the anion to be intercalated, maintained at a constant pH by
[2,11].
adding alkaline solution (NaOH). The experimental apparatus was
In this context, the present study sought to evaluate the adsorp-
similar to that used by Troutier-Thuilliez et al. [22]. on a pilot scale.
tion of CA in fixed bed columns of LDH using the method of
The LDH phases were synthesized by the addition of salts contain-
breakthrough curves. Influences of the superficial velocity (vz ) and
ing the di and trivalent cations, as well as the respective anions
bed height (L) were investigated by means of a CCRD, where the
in the reactor (volume up to 10 L) containing decarbonated water
responses [breakthrough time (tb ), exhaustion time (te ), the dif-
at room temperature. Thereafter, to obtain the Zn2 Cr–NO3 phases
ference (te − tb ), CA adsorption efficiency (ads ), column utilization
solutions of Zn(NO3 )2 and Cr(NO3 )3 (Zn/Cr ratio = 2) were used, all
efficiency (col ) and volumetric productivity (P)] were evaluated as
resulting in a solution with final concentration of 0.1 M, with pH
functions of simultaneous variations of experimental conditions.
controlled by the addition of NaOH (0.2 M) and inert atmosphere
Moreover, the column effectiveness was verified using assays car-
(free of CO2 , controlled with a continuous flux of nitrogen). After
ried out for the separation of CA from amino acids.
approximately 24 h, the suspension was centrifuged, washed three
times with decarbonized water, dried in an oven at 30 ◦ C and then
analyzed by powder X-ray diffraction (XRD) and Fourier transform
infrared spectroscopy (FTIR).
2. Methodology

2.1. Analytical methods 2.3. Adsorbent microparticles

CA concentrations were determined by spectrophotometric The layered double hydroxides (LDH) were microencapsulated
analysis of the product derived from the reaction of CA with imida- in alginate gels and used in packed columns. The microcapsules
zole. The classic method by Bird et al [19] was used, with Clavulin® particles (LDHME ) were prepared by ionic gelling of the sodium
(GlaxoSmithKline, United Kingdom) as a standard. Two different alginate biopolymer in the presence of CaCl2 at room tempera-
solutions were prepared (A and B). Solution A was composed of 5 mL ture. The microparticles were characterized in terms of density, p
of imidazole (60 g/L, pH 6.8) with 1 mL of sample, where the blank (gparticles /cm3 particles ), average particle diameter, dp (␮m), average
was the solution formed by mixing 5 mL of imidazole and 1 mL of pore diameter, dpore (nm), bulk density, bulk (gsolids /cm3 particles ),
distilled water. In solution B, 5 mL of distilled water was added to particle porosity, εp , and bed porosity, εB (Table 1). The density
1 mL of sample, where the blank was distilled water. The solutions was determined by pycnometry using density bottles of 25 mL.
were heated to 30 ◦ C for 15 min. Then, the absorbance of solu- The average diameter of the particles was determined in a laser
tions A and B were measured at  = 312 nm in a Nicolet (Evolution) diffraction Mastersizer 2000 equipment (Malvern Instruments Ltd.,
500 spectrophotometer. The difference in absorbance of the two UK). The pore structure was measured by the mercury porosimetry
substances (A and B) was used for calibration against the concen- technique, using a Micromeritics 9400 porosimeter (Micromeritics
tration. This relationship is linear until reaching a CA concentration Instrument Corporation, USA). The bed porosity was determined by
of 50 mg/L; thus, the samples were appropriately diluted. the method initially described by Arnold et al. [23]. The prepara-
 M.B.S. Forte et al. / Process Biochemistry 51 (2016) 509–516 511

Table 1
Properties of the LDHME particles: average particle diameter (dp ), average pore diameter (dpore ), bulk density (bulk ), real density (p ), particle porosity (εp ) and bed porosity
(εB ).
bulk
dp (␮m) dpore (nm) bulk (g/cm3 ) p (g/cm3 ) εp = 1 − p
εB

171.4 ± 0.7 154.1 0.63 2.17 0.71 0.40

the column height, since the column diameter was the same in all
assays. Eqs. (1)–(6) [25,26] were used to calculate these parameters.

tb
tb − C/C0 dt
Mb 0
ads = = (1)
Mb + Meff tb

tb
tb − C/C0 dt
Mb 0
col = = (2)
Me te
te − C/C0 dt
Fig. 1. Breakthrough curves of a component in fixed bed adsorption columns. 0

tb
tion of LDHME particles and their respective characterizations were C0 (vz A)tb − C0 (vz A) C
reported in detail in previous studies [16,24].
0
Mb = (3)
C0 dt
2.4. Experimental assays and process evaluation parameters
te
The assays were performed using a chromatography system. A C0 (vz A)te − C0 (vz A) C
jacketed column (T = 25 ◦ C) was used, with an internal diameter
0
d = 1.0 cm and different bed heights, L (cm), containing the LDHME Me = (4)
adsorbent particles. Solutions of CA (C0 = 2500 mg/L), obtained by C0 dt
diluting the pharmaceutical product Clavulin® in distilled water tb
and adjusting the solution to pH 6.5, were continuously injected
C0 (vz A) C
with different superficial velocities (flow rates), vz (cm/min). Sam-
ples were collected at the outlet of the column at defined time Meff =
0
(5)
intervals and the CA concentrations were determined. The perfor- C0 dt
mance of the process was evaluated in terms of vz and L by means of M  1
e
a factorial design. In the assays for validation and separation of CA P= × (6)
te Vc
in the presence of contaminants, the experiments were performed
according to the same procedure, with injection of multicomponent where C0 and C are the CA concentrations in the initial solution and
solutions of CA and the amino acids TYR and PRO at the concentra- at the column exit, respectively; vz is the fluid superficial velocity;
tion of 200 mg/L, since this is a concentration commonly found in A is the cross section area of the column; Vc is the volume of the
fermentation broths [5]. column; tb and te are the breakthrough and exhaustion times at
The operating features of the column and process evaluation which the exit concentrations reach 10 and 90% of the feed con-
is carried out by comparing the amounts of the target molecule centration, respectively; Mb , Me and Meff are the amounts of CA
adsorbed and leaving the column in the effluent, as well as the time adsorbed at tb , te , and that leaving the column in the effluent at
required for the entire operation. Fig. 1 shows a breakthrough curve time tb , respectively. The volumetric flow rate was kept constant
of one component, where the areas indicated are related to the used during experiments.
capacity of the column (S1 ), adsorbate leaving in the effluent (S2 ) As already explained above, breakthrough and exhaustion times
and the unused capacity of the column (S3 ) [25,26]. are important parameters for the chromatographic optimization
A practical rule of chromatography processes indicates that col- process, and therefore their difference, herein referred to as
umn feeding should stop when the compound concentration in the tads , will be considered as the process optimization parameter
effluent is about 10% of the inlet concentration, so as to minimize (tads = te − tb ). Low tads values tend to improve the process, since
losses. Moreover, based on Fig. 1 there is a large portion of the col- tads = 0 means the column is operating as an ideal tubular reactor.
umn that may remain unused (S3 ), which is undesirable although
unavoidable, and the larger the difference between te and tb , the 2.5. Process evaluation by factorial design (CCRD)
larger the unused portion of the column. Therefore, all process
optimization techniques must consider on one hand the minimiza- The breakthrough curves of CA through the LDHME bed were
tion of S3 , a predictable consequence of minimizing the difference evaluated with regards to breakthrough time [tb (min)], exhaus-
te − tb , and on the other hand the obvious maximization of S1 . In an tion time [te (min)], tads (min), solute adsorption efficiency [ads
attempt to fulfill this task in this present work, the following pro- (%)], column utilization efficiency [col (%)] and volumetric pro-
cess parameters were studied: solute adsorption efficiency [ads ductivity [P (kg/h cm3 )], according to a central composite rotatable
(%)], column utilization efficiency [col (%)] and volumetric pro- design (CCRD). This technique has been widely used in process opti-
ductivity [P (kg/h cm3 )], which in this case is only a function of mization with the evaluation of multiple responses as functions of
512 M.B.S. Forte et al. / Process Biochemistry 51 (2016) 509–516

Table 2
Real and coded values for the CCRD.

Parameter −1.41 −1 0 +1 +1.41

vz (cm/min) 0.25 0.37 0.65 0.93 1.05

L (cm) 3.0 3.6 5.0 6.4 7.0

multiple factors, which are varied simultaneously [17,18]. For this

[a] 1.0
purpose, a CCRD was performed consisting of 11 assays which were
functions of vz (cm/min) and L (cm) (independent variables), with
0.8
the responses (dependent variables) tb , te , tads , ads , col and P as
defined by Eqs. (1)–(6). In general, individual tb and te values are of 0.6 R1

0
C/C
minor importance in the process as a whole, however a high tb time R2
may be inconvenient due to the possibility of product degradation, 0.4
R4
especially when working with unstable molecules as is the case 0.2
R6
of CA. On the other hand, it is desirable to have te values as close
as possible to tb , as previously explained. The statistical analysis 0.0
was carried out by using the online software Protimiza Experimental 0 100 200 300 400 500 600 700
Design (http://experimental-design.protimiza.com.br/). The exper- t (min)
imental conditions and their coded levels are shown in Table 2.
[b]
1.0
3. Results and discussion
0.8
3.1. Experimental breakthrough curves

0
0.6 R3
C/C
R5
Table 3 and Fig. 2 present the results for the CA breakthrough 0.4
R7
curves. In order to obtain more consistency in data analysis, it is
important to analyze breakthrough and exhaustion times simulta- 0.2 R8

neously with the respective efficiencies and productivities, since

0.0
there may be combinations that hamper good development of the
0 100 200 300 400 500 600 700
process. Flow rate and bed height are usually considered the most
important variables as far as breakthrough curves are concerned t (min)
[9,11]. However, in this work insignificant variations were observed [c]
1.0
in solute recovery efficiency for the change in these variables, with
all results near 98% which are excellent CA recoveries. 0.8
Column efficiency (col ), however, varied considerably from 16% R9
0

to 62% for relatively similar operating conditions. Similar behav- 0.6

C/C

R 10
ior was observed by Burkert et al. [11] in breakthrough curves of R 11
0.4
cephalosporin-C in fixed bed columns with Amberlite XAD-2 resin,
where the authors reported that column efficiency was mostly 0.2
influenced by bed height. Column efficiency is directly related to
the contact time of the solute with the adsorbent, where the higher 0.0
the flow rates the lower the contact time. On the other hand, contact 0 100 200 300 400 500 600 700
time is also closely related to the bed height [11]. This is an exam- t (min)
ple of how, especially in this case, the process variables should be
Fig. 2. Breakthrough curves for CA adsorption in the LDHME fixed bed column:
evaluated simultaneously.
experiments according to the CCRD matrix.
The breakthrough and exhaustion times ranged from 15 to
104 min and 111 to 593 min, respectively. As might be expected,
the highest P values were associated with the shortest times, Other authors have found similar results, i.e. higher flow rates
where these values were between 3.5 and 15.3 kg/h cm3 . The assay and larger bed lengths led to better results for adsorption of
R2 stresses the importance of simultaneous evaluation of the cephalosporin-C in fixed bed columns composed of the resin
responses, since in this run the lowest tads (88 min) and highest Amberlite XAD-2 [11]. Process feasibility depends on the compro-
col (62%) were obtained, however P only reached an intermediate mise between low tads and high efficiencies and productivities.
value (11.6 kg/h cm3 ). From Fig. 3, it can be seen that high flows and high bed heights are
Effects of the operational parameters were only evaluated for required regarding tads . However, high heights reduce produc-
responses statistically relevant at 5% significance (confidence level tivity. In conditions of maximum productivity there is a significant
of 95%). Reparameterizations were carried out, and empirical coded increase in tads (from less than 100 up to about 200 min), and
equations for tb , te , tads and P as functions of vz and L were a reduction in breakthrough time. On the other hand, very high
obtained, as described in Table 4. The analysis of variance (ANOVA) breakthrough times are not convenient for unstable molecules
was performed to validate the empirical models at the assumed 5% because of possible degradations. In such cases, the analysis of
significance, and the results can be seen in Table 4. From this table, it response surfaces allows for a more reliable choice of the best oper-
could be concluded that the empirical models were validated since ating conditions, based on the simultaneous responses. Decisions
the coefficients of determination were greater than 90%, except for should be made based on the response surfaces as well as additional
P which was near 85%, and the values of Fcalc were higher than those constraints of the process, where in this case is the degradability
listed [17]. Thus, it is possible to generate the response surface for of the product. For example, the advantage of using systems with
each response studied, as shown in Fig. 3. flow vz = 1 cm/min and bed height L = 6.5 cm can be observed, since
 M.B.S. Forte et al. / Process Biochemistry 51 (2016) 509–516 513

Table 3
CCRD matrix for the CA breakthrough curves: breakthrough time (tb ), exhaustion time (te ), tads , adsorption efficiency (ads ), column use efficiency (col ) and productivity
(P) as functions of superficial velocity (vz ) and bed height (L). * Coded and real (in brackets) values.

Assay vz * (cm/min) L* (cm) tb (min) te (min) tads (min) ads (%) col (%) P (kg/h cm3 )

R1 −1 (0.37) −1 (3.6) 60.3 236.3 176.0 98.7 48.9 7.6

R2 1 (0.93) −1 (3.6) 22.3 111.0 88.7 97.8 61.9 11.6
R3 −1 (0.37) 1 (6.4) 103.7 593.0 489.3 98.3 39.5 3.5
R4 1 (0.93) 1 (6.4) 38.7 156.0 117.3 98.4 44.6 10.6
R5 −1.41 (0.25) 0 (5.0) 95.0 470.0 375.0 98.4 38.5 3.8
R6 1.41 (1.05) 0 (5.0) 27.0 150.9 123.9 97.4 37.5 14.0
R7 0 (0.65) −1.41 (3.0) 14.7 175.0 160.3 96.6 16.0 15.3
R8 0 (0.65) 1.41 (7.0) 70.0 261.3 191.3 98.9 49.7 7.1
R9 0 (0.65) 0 (5.0) 50.0 205.0 155.0 97.4 36.0 10.2
R10 0 (0.65) 0 (5.0) 54.0 215.6 161.6 98.4 48.3 9.0
R11 0 (0.65) 0 (5.0) 35.6 210.0 174.4 97.8 32.6 9.0

Table 4
Empirical coded models (statistically significant parameters) and the related ANOVA for breakthrough time (tb ), exhaustion time (te ), tads and productivity (P).

Response Model R2 Fcalc Flist (DFReg , DFRes )

2
tb (min) tb = 45.6 − 25.0vz + 8.7vz + 17.3L 0.944 39.2 4.35 (3, 7)
te (min) te = 216.2 − 126.7vz + 50.7vz 2 + 65.5L − 77.9vz L 0.945 25.8 4.53 (4, 6)
tads (min) tads = 170.6 − 101.8vz + 42.0vz 2 + 48.2L − 71.2vz L 0.911 15.3 4.53 (4, 6)
P (kg/h cm3 ) P = 9.2 + 3.2vz − 2.1L 0.845 21.8 4.46 (2, 8)

R2 , coefficient of determination; Fcalc , calculated F-factor; Flist , listed F-factor at 5% significance; DFReg , Degree of Freedom (Regression); DFRes , Degree of Freedom (Residues).

Fig. 3. Response surfaces (contour curves) for clavulanic acid adsorption: [a] breakthrough time (tb ); [b] exhaustion time (te ); [c] tads ; [d] productivity (P).
514 M.B.S. Forte et al. / Process Biochemistry 51 (2016) 509–516

Fig. 4. Observed and predicted values for the empirical models of breakthrough time (tb ), exhaustion time (te ), tads and productivity (P).

the increase in breakthrough time and the decrease in delta time is

1.0
sufficient to offset the decrease in productivity.
The obtaining of empirical models enables the construction of 0.8
response surfaces, allowing for process evaluation, as well as com-
puter simulations within the experimental data range. Besides, it 0.6
C/C0

is possible to determine the optimal operating conditions, since CA

there is good agreement between predicted and observed values, 0.4
TYR
as illustrated in Fig. 4.
With the aid of Fig. 3 and process constraints, such as pressure 0.2 PRO
drop inside the column, which can cause flow obstruction along the
0.0
bed in extreme conditions of flow, the most suitable operational
condition considered that: vz = 1 cm/min and L = 6.5 cm. The theo- 0 50 100 150 200 250
retical values of the responses, according to the coded empirical
t (min)
models (Table 5) for the column operating under these conditions,
are: tb = 46 min, te = 105 min, tads = 67 min and P = 11 kg/h cm3 . Fig. 5. Separation of CA in the presence of tyrosine (TYR) and proline (PRO): break-
through curves in fixed bed columns of LDHME at the optimized conditions.

3.2. Process validation and CA separation

ing the optimized results with the nominal values (central point of
Assays were performed at the optimized conditions and with the CCRD), there was no change in tb , on the other hand there were
multicomponent solutions, containing CA and the main contam- 50% decrease in te , 40% decrease in tads and 38% increase in P.
inants found in fermentation broths, which are the amino acids Differences found in breakthrough curves for CA, TYR and PRO
TYR and PRO [5]. The results from these assays are shown in Fig. 5. confirm the potential of LDHME columns for the separation of CA
Experimental data are the average of three responses with stan- from those two amino acids (Table 6). It can also be observed that
dard deviation of less than 0.05. Regarding the CA adsorption, the quite small amounts of amino acids were adsorbed in the column
empirical models were validated, since the responses were consis- which is also a positive result. CA purity could be calculated from
tent with the predicted values, with slight variations among the Eq. (3), considering the contaminants (TYR and PRO) at the CA
theoretical values of the coded empirical models (Table 5). The breakthrough point (46 min), assumed as the end of the adsorp-
responses showed more favorable results when compared to all tion step. Thus, CA purity increased from 85% to 95%. Moreover, a
assays performed previously in the CCRD (Table 3). When compar- gain in the process may be observed in the CA responses calculated
 M.B.S. Forte et al. / Process Biochemistry 51 (2016) 509–516 515

Table 5
Predicted and observed values for CA adsorption: breakthrough time (tb ), exhaustion time (te ), tads , and productivity (P) at the optimized conditions.a

Response tb (min) te (min) tads (min) P (kg/h cm3 )

Predicted (Model) 46 105 67 11

Observed 46 146 100 13
a
vz = 1 cm/min; L = 6.5 cm.

Table 6
Separation of CA at the optimized conditionsa : breakthrough time (tb ), exhaustion time (te ), tads , adsorption efficiency (ads ), column utilization efficiency (col ) and
productivity (P).

Assay tb (min) te (min) tads (min) ads (%) col (%) P (kg/h cm3 )

CA 46 146 100 98 46 13.0

TYR 17 75 58 98 51 0.8
PRO 10 28 18 98 57 1.0
a
vz = 1 cm/min; L = 6.5 cm.

from these results: tb = 46 min, te = 146 min, tads = 100 min and References
P = 13 kg/h cm3 . Values of the adsorption efficiencies and column
use are quite similar to those previously determined (ads = 98% [1] M.B.S. Forte, E.C.L. Elias, H.O. Pastore, F. Maugeri, M.I. Rodrigues, Evaluation of
clavulanic acid adsorption in MgAl-layered double hydroxides: kinetic,
and col = 46%). The volumetric productivity (P) is an important equilibrium and thermodynamic studies, Adsorpt. Sci. Technol. 30 (2012)
response in the adsorption column study due to its standard 65–80.
approach (related to the column volume). However this response [2] C.C. Moraes, M.A. Mazutti, M.I. Rodrigues, F. Maugeri, S.J. Kalil, Mathematical
modeling and simulation of inulinase adsorption in expanded bed column, J.
has not been often evaluated in literature. The productivities of Chromatogr. A 1216 (2009) 4395–4401.
the present work were higher than those reported in studies using [3] S.J. Kalil, F. Maugeri-Filho, M.I. Rodrigues, Ion exchange expanded bed
similar systems (P = 3 kg/ h cm3 ) [11]. chromatography for the purification of an extracelular inulinase from
Kluyveromyces marxianus, Process Biochem. 40 (2005) 581–586.
[4] P.S. Saudagar, S.A. Survase, R.S. Singhal, Clavulanic acid: a review, Biotechnol.
Adv. 26 (2008) 335–351.
4. Conclusion [5] C.S. Silva, M.F. Cuel, V.O. Barreto, W.H. Kwong, C.O. Hokka, M. Barboza,
Separation of clavulanic acid from fermented broth of amino acids by an
aqueous two-phase system and ion-exchange adsorption, New Biotechnol. 29
CA adsorption on microencapsulated layered double hydroxides (2012) 428–431.
(LDHME ) particles was evaluated using experimental breakthrough [6] A.F. Mayer, F.B. Anspach, W.D. Deckwer, Purification of clavulanic acid by
curves. It was shown that application of the factorial design ion-pairing systems, Bioseparation 6 (1996) 25–39.
[7] P.A. Bersanetti, R.M.R.G. Almeida, M. Barboza, M.L.G. Araujo, C.O. Hokka,
technique is an efficient tool to obtain a fast, economically and Kinetic studies on clavulanic acid degradation, Biochem. Eng. J. 23 (2005)
statistically controlled evaluation and optimization of the process, 31–36.
which allowed a simultaneous evaluation of the responses (pro- [8] P.A. Belter, E.L. Cussler, W.-S. Hu, Bioseparations: Downstream Processing for
Biotechnology, Wiley-Interscience, New York, 1988.
cess times, efficiencies and productivity) via the response surface [9] D.A. Luz, A.K.O. Rodrigues, F.R.C. Silva, A.E.B. Torres, C.L. Cavalcante, E.S. Brito,
analysis methodology. The best results were observed with the D.C.S. Azevedo, Adsorptive separation of fructose and glucose from an
column operating at the following conditions: vz = 1 cm/min and agroindustrial waste of cashew industry, Bioresour. Technol. 99 (2008)
2455–2465.
L = 6.5 cm, whose responses were tb = 46 min (breakthrough time), [10] C.C. Moraes, M.A. Mazutti, F. Maugeri, S.J. Kalil, Modeling of ion exchange
te = 146 min (exhaustion time), ads = 98% (adsorption efficiency), expanded-bed chromatography for the purification of C-phycocyanin, J.
col = 46% (column utilization efficiency), tads = 100 min (differ- Chromatogr. A 1281 (2013) 73–78.
[11] C.A.V. Burkert, G.N.O. Barbosa, M.A. Mazutti, F. Maugeri, Mathematical
ence te − tb ) and P = 13 kg/h cm3 (productivity), which represents a
modeling and experimental breakthrough curves of cephalosporin C
50% decrease in te , 40% decrease in tads and 38% increase in P, with adsorption in a fixed-bed column, Process Biochem. 46 (2011) 1270–1277.
respect to the nominal conditions. The methodology of adsorption [12] A.K. Mostafazadeh, M. Sarshar, S. Javadian, M.R. Zarefard, Z.A. Haghighi,
Separation of fructose and glucose from date syrup using resin
in fixed bed columns by the breakthrough curves method proved
chromatographic method: experimental data and mathematical modeling,
to be a good option to study the separation and purification of CA. Sep. Purif. Technol. 79 (2011) 72–78.
High purity solutions (95%) were obtained with the column operat- [13] C. Forano, T. Hibino, F. Leroux, C. Taviot-Guého, Layered double hydrocides,
ing under the optimized conditions. The LDHME adsorbent showed in: F. Bergaya, B.K.G. Theng, G. Lagaly (Eds.), Handbook of Clay Science,
Elsevier, Amsterdam, 2006, pp. 1021–1095.
good ability to promote separation between CA and the amino acids [14] A.C.D. Newman, Chemistry of Clays and Clay Minerals, Wiley-Interscience,
TYR and PRO, emerging as an economical alternative for obtaining New York, 1987.
this biomolecule. [15] M. Barboza, R.M.R.G. Almeida, C.O. Hokka, Influence of temperature on the
kinetics of adsorption and desorption of clavulanic acid by ionic exchange,
Biochem. Eng. J. 14 (2003) 19–26.
[16] M.B.S. Forte, C. Taviot-Guého, F. Leroux, M.I. Rodrigues, F. Maugeri Filho,
Acknowledgements Development and characterization of a new adsorbent for biomolecule
separation: intercalation and adsorption of clavulanic acid in layered double
hydroxides, J. Chem. Technol. Biotechnol. (2015), http://dx.doi.org/10.1002/
The authors are grateful to Dr. Carlos Raimundo Ferreira Grosso jctb.4759 (in press).
of the Food Microstructure Laboratory (DEPAN/FEA/UNICAMP), Dr. [17] M.I. Rodrigues, A.F. Iemma, Experimental Design and Process Optimization,
CRC Press, New York, 2014.
Rosiane Lopes da Cunha of the Process Engineering Laboratory
[18] M.B.S. Forte, M.A. Mazutti, F. Maugeri, M.I. Rodrigues, Comparative studies on
(DEA/FEA/UNICAMP) and Dr. Meuris Gurgel Carlos da Silva of the parametric sensitivity analyses using conventional and factorial design
Environmental Engineering Laboratory (DDPP/FEQ/UNICAMP) for methodologies: mathematical modelling of clavulanic acid adsorption on
zeolites, J. Chem. Technol. Biotechnol. 87 (2012) 1715–1722.
allowing the use of equipment to obtain and characterize the
[19] A.E. Bird, J.M. Bellis, B.C. Gasson, Spectrophotometric assay of clavulanic acid
microparticles. This study was funded by the FAPESP (Founda- by reaction with imidazole, Analyst 107 (1982) 1241–1245.
tion for Research Support of the State of São Paulo) – process No. [20] L.S. Bates, R.P. Waldren, I.D. Teare, Rapid determination of free proline for
2008/57298-3 and by the CAPES (Coordination of Improvement of water-stress studies, Plant Soil 39 (1973) 205–207.

Higher Education Personnel) – process No. BEX 6843/10-7.

516 M.B.S. Forte et al. / Process Biochemistry 51 (2016) 509–516

[21] W.A. Torello, L.A. Rice, Effects of NaCl stress on proline and cation [25] P. Sridhar, N.V.S. Sastri, J.M. Modak, A.K. Mukherjee, Mathematical simulation
accumulation in salt sensitive and tolerant turfgrasses, Plant Soil 93 (1986) of bioseparation in an affinity packed column, Chem. Eng. Technol. 17 (1994)
241–247. 422–429.
[22] A.-L. Troutier-Thuilliez, C. Taviot-Guého, J. Cellier, H. Hintze-Bruening, F. [26] H.A. Chase, Affinity separations utilising imobilised monoclonal antibodies—a
Leroux, Layered particle-based polymer composites for coatings: Part I. new tool for the biochemical engineer, Chem. Eng. Sci. 39 (1984) 1099–1125.
Evaluation of layered double hydroxides, Prog. Org. Coat. 64 (2009) 182–192.
[23] F.H. Arnold, H.W. Blanch, C.R. Wilke, Liquid-chromatography plate height
equations, J. Chromatogr. 330 (1985) 159–166.
[24] M.B.S. Forte, C. Taviot-Guého, F., Leroux, M.I. Rodrigues and F. Maugeri Filho,
Micropartículas de alginato com hidróxidos duplos lamelares encapsulados e
seu uso, National Institute of Industrial Property (INPI), Patent (2014)
BR1020140178430.