Characterization of solvent adsorption from single-component and multicomponent mixtures
To overcome the limitations caused by butanol accumulation, the development of an integrated bioprocess for in situ butanol removal via EBA was systematically explored. To reliably predict the adsorption behavior of butanol, acetone, and ethanol from fermentations, equilibrium adsorption isotherms on L-493 were characterized through a series of model studies. The adsorption behavior of both butanol and ethanol on L-493 has previously been the subject of extensive characterization,[43, 44] however, that of acetone and the solvent mixture has not yet been reported. Table 1 compares the resultant best-fit, Freundlich model parameter estimates for the individual solvents and the mixture. Despite variations between parameter estimates from this study and previous reports, the resultant predicted and experimental isotherms nevertheless strongly agreed over the concentration range of interest (see Supporting Information). We previously found that relative differences in octanol–water partitioning coefficients (logKO/W) could be used to predict relative differences in adsorption of butanol (logKO/W = 0.8) and ethanol (logKO/W = −0.26) on L-493, and from this work appears that acetone (logKO/W = −0.16) likewise follows a similar trend. Overall, butanol, acetone, and ethanol were effectively adsorbed from all solutions; however, competitive effects may have reduced the adsorption of each in multicomponent mixtures but the changes observed were not statistically significant. This lack of selectivity was not unexpected as nonspecific hydrophobic interactions are responsible for the adsorption all three solvents on L-493. Thus, although butanol recovery was the principal objective of this study, the simultaneous recovery of acetone and ethanol by EBA should likewise be expected (albeit to a lesser extent), and must accordingly be accounted for. It is worth noting that glucose, the most abundant species in the bioreactor, has previously been shown to not adsorb appreciably to L-493.44 Here, we further found its presence to insignificantly impact butanol adsorption (see Supporting Information). Similarly, as its surface is not functionalized (i.e., it is not an ion exchange resin), we have also previously found that L-493 does not significantly interact with any ionic species in the medium, including both acetate and butyrate, as well as inorganic nutrient salts (unpublished data). Thus, L-493 has been found to behave rather inertly with respect to all but the solvent products in the fermentation media, meaning that its presence should not be expected to inhibit growth or productivity.
Table 1. Freundlich Adsorption Isotherm Model Parameters for Acetone-Butanol-Ethanol (ABE) Fermentation Products and Dowex Optipore L-493
|Single component||n-Butanol||362 ± 25||2.00 ± 0.06||This study|
|n-Butanol||446 ± 115||2.22 ± 0.26||Nielsen et al.|
|Acetone||66 ± 11||1.46 ± 0.07||This study|
|Ethanol||18 ± 5||1.26 ± 0.09||This study|
|Ethanol||23 ± 12||1.25 ± 0.29||Nielsen et al.|
|Tertiary mixture||n-Butanol||328 ± 21||2.03 ± 0.06||This study|
|Acetone||55 ± 19||1.82 ± 0.26||This study|
|Ethanol||17 ± 5||1.60 ± 0.20||This study|
Development and characterization of an EBA process
A series of model characterizations were next performed to analyze the impact of select parameters of postulated importance to the EBA design. For instance, as seen in Figure 2A, butanol was removed from solution more rapidly as recirculation rates (in terms of superficial velocity) through the column were increased. Therefore, in practice, recirculation rates through the EBA column should be sufficiently high so as to suppress butanol accumulation during its peak production. Initial rates of butanol uptake were estimated as 5.6 ± 0.2, 12.6 ± 0.4, 19.9 ± 3.4, and 41.1 ± 4.4 g/L h for superficial velocities of 2.1, 5.7, 9.5, and 18.5, respectively, all of which are well above the typical volumetric productivity C. acetobutylicum ATCC 824. Moreover, as a strong linear correlation between superficial velocity and rates of butanol uptake was observed (R2 = 0.997; see Supporting Information), this implies that bulk transport and not intraparticle diffusion limited the kinetic performance of the EBA process over the range of conditions examined. Note, however, that at further higher recirculation rates this relationship is ultimately expected to fold as intraparticle diffusion then becomes rate limiting. Whereas higher recirculation enables faster butanol uptake, excessive pumping incurs higher utility costs and can impose detrimental shear stresses against the culture and should thus be avoided. Therefore, the midpoint circulation rate of 9.5 cm/min was selected for all subsequent studies.
Figure 2. Investigating relevant operating parameters and performance of the EBA design through a series of cell-free experiments.
(A) Butanol removal from the bioreactor by circulation through an EBA column at superficial velocities of rates (in cm/min) of: 1.0 (solid squares), 2.7 (solid circles), 4.7 (open squares), and 8.9 (open circles). (B) Butanol removal from the bioreactor using EBA columns containing L-493 at an adsorbent fraction (in g/L) of: 26.7 (solid circles), 53.3 (open squares), and 80 (open diamonds). (C) Bed expansion as a function of flow rate for EBA columns containing L-493 at an adsorbent fraction (in g/L) of: 26.7 (solid circles), 53.3 (open squares), and 80 (open diamonds). (D) Accumulation of butanol in a bioreactor when added exogenously at a constant rate of 0.37 g/L h without (solid line) and with EBA operation (solid line with solid circles). Error bars represent standard error from triplicate measurements.
Download figure to PowerPoint
The extent of butanol removal, meanwhile, is controlled by the net adsorption capacity of the adsorbent or in this case the mass of the adsorbent bed used in the EBA column. As seen in Figure 2B, increasing the adsorbent fraction from 26.6 to 53.3, and then to 80 g/L, the residual butanol concentration remaining in bioreactor at steady-state dropped from 10.8 ± 0.1 to 9.4 ± 0.1, and finally to 8.0 ± 0.2 g/L. Meanwhile, specific loadings of butanol on L-493 of approximately 4,340, 4,060, and 3,780 mmol/kg, respectively, were achieved at steady state; which strongly agree with the predicted and experimental equilibrium isotherms (see Table 1 and Supporting Information). Note that in all cases the butanol concentration in the bioreactor was effectively reduced from a titer above its inhibitory threshold (∼14 g/L) to a level below this mark. Whereas using a larger adsorbent bed increases the extent by which the aqueous accumulation of inhibitory products can be reduced, practical trade-offs also exist and should be considered. For example, in addition to increased cost, increasing the adsorbent fraction reduces the headspace available for expansion (assuming a fixed column volume) and thus suitably limits the maximum possible superficial velocity through the column (i.e., circulation rate). As seen in Figure 2C, bed height expansion as a function of superficial velocity of the circulating fluid was found to be well-modeled in this study using a form of the Richardson-Zaki equation (Eq. (3)), with best-fit estimates of the terminal settling velocity (ut) and the Richardson-Zaki exponent (n) found to be 56.7 cm/min and 4.38, respectively. Relative bed expansion was largely insensitive to the absolute mass of adsorbent used, indicating that the entire bed was effectively fluidized under all conditions studied. With respect to the EBA process developed in this study, at adsorbent fractions of 26.6, 40, and 53.3 g/L, for example, the headspace available in the EBA column represented 86.7%, 80%, and 73.4% of its total volume and would accommodate maximum height increases of 7.5-, 5-, and 3.8-fold, respectively. Thus, whereas the threefold maximum increase in expanded bed height (H) experienced in Figure 2C was readily accommodated in the present EBA design, careful and concurrent consideration of bed size and recirculation rate must be made to ensure that excessive expansion does not result in bed packing at the effluent (i.e., top) of the column.
The ability of the proposed design to simultaneously remove butanol from the bioreactor as it is produced was at last assessed through simulation. Butanol was continuously added to a well-mixed bioreactor (originally containing water) at a constant rate of 0.37 g/L h to approximate the wild-type production rate of C. acetobutylicum ATCC 824.22 The bioreactor contents were continuously recirculated through an EBA column containing 33 g/L L-493 at a constant superficial velocity of 9.5 cm/min. As seen in Figure 2D, butanol accumulation in the aqueous phase was effectively reduced via rapid adsorption onto L-493. In the control, the would-be inhibitory butanol threshold of 13 ± 0.1 g/L was reached in approximately 36 h, whereas with EBA the aqueous butanol levels reached a maximum of only 5.2 ± 0.1 g/L.
Of course, for EBA to constitute an effective strategy it must also be possible to efficiently desorb butanol from the resin and ultimately recover it. This process simultaneously results in regeneration of the resin and EBA column for their subsequent reuse, minimizing process costs. We have previously shown that regeneration of L-493 is possible by heating it to 100°C and collecting the resultant vapors. Recovery of up to 85% of adsorbed butanol was possible by this approach with almost no deviation in adsorption behavior. However, there was concern that the repeated heating and cooling of a polymeric resin could prove detrimental in the long-term. Thus, here we have alternatively investigated the prospect of vacuum regeneration of L-493 at ambient temperatures, as described above. Table 2 presents the results from a series of three consecutive adsorption–desorption cycles, wherein it can be seen that butanol was effectively desorbed and collected. In most cases, two liquid phases were present in the condensed fraction as the butanol content exceeded its solubility limit in water (∼73 g/L at 25oC), a feature that may further aid in downstream processing. In total, the butanol recovered by vacuum regeneration averaged about 81 ± 10% of that, which was originally adsorbed, a mark that compares well with thermal regeneration. However, noting that all final fractions collected still contained appreciable levels of butanol, it is expected that further optimization of this regeneration protocol to include extended periods of vacuum application will only add to this mark. Finally, an additional benefit of note here is that the resin was never required to be removed from the EBA column throughout the experiment, a feature that bodes well for overall process modularity.
Table 2. Comparing Butanol Recovery via Vacuum Regeneration of L-493 Following Three Repeated Adsorption–Desorption Cycles
|Cycle||C0 (g/L)||Ce (g/L)||Butanol Adsorbed (g)||Fraction||Ctrap (g/L)||Vtrap (mL)||Butanol Trap (g)||Butanol Recovered (%)|
|1||15.0 ± 0.1||10.6 ± 0.1||4.7 ± 0.2||1||97.7 ± 0.4||8.5 ± 0.2||0.83 ± 0.02||78.9 ± 1.3|
|2||93.7 ± 0.4||8.0 ± 0.2||0.75 ± 0.02|
|3||91.9 ± 0.4||10.0 ± 0.2||0.92 ± 0.02|
|4||77.8 ± 0.4||8.5 ± 0.2||0.66 ± 0.02|
|5||65.9 ± 0.4||8.0 ± 0.2||0.53 ± 0.01|
|Total||85.8 ± 0.5||43.0 ± 1.0||3.69 ± 0.09|
|2||14.2 ± 0.1||9.6 ± 0.1||4.6 ± 0.2||1||79.5 ± 0.4||8.5 ± 0.2||0.68 ± 0.02||72.5 ± 1.4|
|2||87.7 ± 0.4||8.0 ± 0.2||0.70 ± 0.02|
|3||82.8 ± 0.4||7.0 ± 0.2||0.58 ± 0.02|
|4||64.0 ± 0.4||7.0 ± 0.2||0.45 ± 0.01|
|5||83.2 ± 0.4||7.0 ± 0.2||0.58 ± 0.02|
|6||63.2 ± 0.4||5.5 ± 0.2||0.35 ± 0.01|
|Total||77.5 ± 0.6||43.0 ± 1.2||3.33 ± 0.10|
|3||13.2 ± 0.1||9.6 ± 0.1||3.6 ± 0.2||1||89.0 ± 0.4||8.0 ± 0.2||0.71 ± 0.02||92.4 ± 1.4|
|2||78.8 ± 0.4||8.0 ± 0.2||0.63 ± 0.02|
|3||84.4 ± 0.4||7.0 ± 0.2||0.59 ± 0.02|
|4||68.8 ± 0.4||7.0 ± 0.2||0.48 ± 0.01|
|5||79.6 ± 0.4||7.2 ± 0.2||0.57 ± 0.02|
|6||60.5 ± 0.4||5.5 ± 0.2||0.33 ± 0.01|
|Total||77.8 ± 0.6||42.7 ± 1.2||3.32 ± 0.10|
| || || || || || || || || |
Traditional batch fermentation of C. acetobutylicum ATCC 824
To establish a baseline for comparison, a traditional batch fermentation of C. acetobutylicum ATCC 824 was first performed, with products allowed to accumulate to their maximum achievable levels. The collective results from the 72-h fermentation are shown in Figure 3 and summarized in Table 3. Following a brief (6–8 h) lag phase, the culture entered acidogenesis wherein production of acetic and butyric acids was predominant, reaching maximum titers of 3.4 ± 0.1 and 4.9 ± 0.1 g/L by 28.6 h, respectively. Solvent production began at approximately 24 h and peaked after 52 h as biomass accumulation began to slow (note: DCW reached a maximum concentration of 4.99 ± 0.05 g/L at an overall glucose yield of 0.045 ± 0.001 g/g). The butanol titer reached a maximum of 11.5 ± 0.2 g/L, approaching the reported toxicity limit of approximately 13 g/L. The average (estimated between 24 and 52 h) volumetric productivity for butanol was 0.41 ± 0.02 g/L h, which compares well with prior reports of 0.32, 0.35, 0.37, and 0.48 g/L h. Acetone and ethanol titers, meanwhile, reached up to 3.9 ± 0.4 and 2.7 ± 0.1 g/L, respectively, giving a total solvent titer 17.7 ± 0.7 g/L, which agrees well with prior reports.[54, 55] Note that whereas ethanol and acetone can contribute to overall solvent stress, as is typical of C. acetobutylicum ATCC 824 fermentations, their accumulation did not approach their reported inhibitory thresholds (50–60 and 70 g/L, respectively), leaving butanol as the primary stressor in this case. Thus, of the 140 g/L glucose initially added, only 77% was ultimately consumed, likely as a result of butanol inhibition which prevented its further and complete consumption. The yield of butanol in this case was found to be 0.11 ± 0.01 g/g with that of total solvents being 0.17 ± 0.01 g/g. These values notably compare well with those of Mermelstein et al. (calculated as 0.12 and 0.18 g/g, respectively), a study performed using the same media and otherwise similar fermentation conditions (i.e., temperature, pH, and agitation). However, relative to the established literature as a whole, these yields are generally low. For example, for batch fermentations of C. acetobutylicum ATCC 824 at pH 4.5, Li et al. reported butanol and total solvent yields of 0.23 and 0.32 g/g, respectively, outputs more typical of Clostridium sp. and closer to the theoretical maximum yield of butanol on glucose of 0.41 g/g. Nevertheless, as the present objective was to demonstrate that EBA could be used to enhance butanol production by preventing its accumulation to inhibitory levels—a threshold that the maximum achievable butanol titers were here able to approach—the present conditions were deemed suitable for this purpose. Further optimization of the fermentation and overall process conditions will surely be key to achieving higher and more representative yields and maximizing the performance.
Figure 3. C. acetobutylicum ATCC 824 control fermentation (i.e., without EBA).
Upper: DCW (open squares) and glucose (solid squares) concentrations. Middle: Concentrations of butanol (open diamonds), acetone (open circles), and ethanol (open inverse triangles). Lower: Concentrations of acetic acid (open stars) and butyric acid (left triangle) as well as pH (solid line). Error bars represent standard error from triplicate measurements.
Download figure to PowerPoint
Table 3. Comparing the Performance of C. acetobutylicum Fermentations Operated in Batch Mode and Fed-Batch with In Situ Product Recovery by Expanded Bed Adsorption
|Operating Mode||Aqueous/Net Accumulation* (g/L)||Productivity (g/L h)||Glucose Yield (g/g)|
|Butanol||Acetone||Ethanol||Total Solvents||Butanol||Total Solvents||Butanol||Total Solvents|
|Batch||11.5 ± 0.2||3.9 ± 0.4||2.7 ± 0.1||17.7 ± 0.7||0.41 ± 0.02||0.63 ± 0.03||0.11 ± 0.01||0.17 ± 0.01|
|Fed Batch + EBA||12.1 ± 0.1/27.1 ± 0.2||8.2 ± 0.1/10.7 ± 0.1||2.5 ± 0.1/3.1 ± 0.1||22.8 ± 0.3/40.7 ± 0.4||0.48 ± 0.02||0.72 ± 0.03||0.19 ± 0.02||0.28 ± 0.02|
EBA for in situ product recovery from a C. acetobutylicum ATCC 824 fermentation
The EBA column was at last integrated with a C. acetobutylicum ATCC 824 fermentation and evaluated for its ability to enhance butanol production. The collective results are depicted in Figure 4 and summarized in Table 3. Following a lag phase, cell growth and acidogenesis began after approximately 15 h. Once again, pH adjustment was used as a strategy to encourage the shift to solventogenesis, although this time it was accordingly performed after approximately 19 h. Acetic and butyric acids accumulated to similar maximum levels relative to the control fermentation (2.4 ± 0.2 and 4.5 ± 0.1 g/L, respectively). Following the appearance of butanol in the medium (first detected at 33.5 h) circulation through the EBA column was initiated. Note that circulation was not initiated earlier so as to minimize possible stress to the culture. This promptly resulted in a decrease in aqueous butanol concentration, from 10.1 ± 0.1 to 5.1 ± 0.1 g/L by 37 h. With an increased capacity for butanol accumulation in the system, the supplementation of additional glucose in a fed-batch manner was explored in an effort to increase its net production. At both 43.4 and 56.1 h, 30 g of additional glucose was added to the culture and quickly assimilated. This resulted in additional biomass production and, most importantly, the increased accumulation of butanol (together with acetone and ethanol). At the end of 72 h, butanol titers, which reached up to 12.1 ± 0.1 g/L, again ultimately approached their inhibitory threshold. Acetone and ethanol titers, meanwhile, peaked at 8.2 ± 0.1 and 2.5 ± 0.1 g/L, respectively. However, in addition to its accumulation in the aqueous media, butanol also accumulated significantly upon L-493, which, by use of the equilibrium isotherm (Table 1), was predicted to reach a final specific loading of 300 g/kg (or 3,850 mmol/kg). Thus, by accounting for all butanol accumulated in the system (i.e., aqueous and adsorbed), an effective titer of 27.1 ± 0.2 g/L was estimated by simple mass balance. Likewise, the adsorption of acetone and ethanol on L-493 were predicted as 48.5 and 10.0 g/kg (or 835 and 217 mmol/kg), respectively, boosting their effective titers to 10.7 ± 0.1 and 3.1 ± 0.1 g/L, and that of total solvent production to 40.7 ± 0.4 g/L. In this case, 90% of the total glucose added was consumed by the culture and the average (as determined between 22.8 and 72 h) butanol productivity was elevated to 0.48 ± 0.02 g/L h (17% greater). Meanwhile, butanol and total solvent yields on glucose were improved to 0.19 ± 0.02 and 0.28 ± 0.02 g/g, respectively; both of which were approximately 70% greater than achieved in conventional batch operation. DCW, however, reached a maximum of only 4.36 ± 0.07 g/L, or 13% less than the control. The growth reduction suggests that the EBA process may have negatively impacted culture fitness, perhaps as a result of its continuous pumping and recirculation through the column. Further optimization of the re-circulation rate may help to minimize such detrimental effects.
Figure 4. C. acetobutylicum ATCC 824 fermentation with EBA for in situ product recovery.
Upper: DCW (open squares) and glucose (solid squares) concentrations. Glucose additions indicated with gray arrows (note: samples taken after addition). Middle: Concentrations of butanol (open diamonds), acetone (open circles), and ethanol (open inverse triangles). Lower: Concentrations of acetic acid (open stars) and butyric acid (left triangle), as well as pH (solid line). Dashed vertical line indicates time at which EBA was initiated; continuous circulation through the EBA column resumed for the duration of the experiment. Error bars represent standard error from triplicate measurements.
Download figure to PowerPoint
A comparison of actual solvent titers achieved in the control fermentation with effective titers achieved using EBA clearly demonstrates the impact of this strategy. As seen in Figure 5, total solvent production was increased by 2.3-fold compared to the control fermentation, whereas net butanol production was improved 2.2-fold. In contrast, in previous works we found that net butanol production could be improved by only approximately 1.8-fold when L-493 was added directly to a batch culture. The net impact of EBA on butanol production was similar to that of previous reports for succinic acid, where 2.9- and 2-fold improvements were made in net production relative to batch and fed-batch operation, respectively. Alternative strategies for the in situ recovery of butanol have also been explored including through solvent extraction. For example, Roffler et al. reported that butanol productivity could be increased 1.7-fold (up to 1.0 g/L-h) in fed batch cultures by circulating the culture through a Karr reciprocating plate extraction column that contained oleyl alcohol as solvent. Meanwhile, to date, gas stripping has been found to provide perhaps the greatest enhancement to butanol and solvent production. For example, improvements in total solvent production and productivity of about 2.3- and 1.7-fold, respectively, have been reported in batch configurations with further and more significant improvements (up to 14- and 2.6-fold, respectively) possible through the adoption of continuous operating protocols. It is likewise anticipated the further evolution of the proposed EBA design—for example, through incorporation of simulated moving-bed protocols to support semicontinuous or continuous operation—will lead to additional and sizable gains in butanol production. Of course, whereas the focus of this study was predominantly directed toward in situ butanol recovery, the efficient recovery of ethanol and acetone will ultimately be of importance to the net viability and sustainability of the process. As the accumulation of these species to inhibitory levels is not the foremost concern (see above), this can be achieved through downstream processing also with the aid of adsorption (following butanol removal) or other suitable strategies (e.g., gas stripping, solvent extraction, distillation).
Figure 5. Comparing effective titers of acetone (circles), butanol (diamonds), ethanol (inverted triangle), and total solvent products (upright triangle) between control (solid shapes) and EBA (open shapes) fermentations. Error bars represent standard error from triplicate measurements.
Download figure to PowerPoint
Assessing the prospects of biofouling
Biofouling is an important concern whenever biological and separation processes are interfaced. Here, for instance, fouling could reduce the surface area available for adsorption and thus the apparent capacity of the adsorbent. SEM imaging was performed on L-493 samples taken from the EBA column to provide a preliminary appraisal of biofouling. As seen from Figure 6, surface fouling by whole cells and cell debris was minimal on L-493. The low apparent density (∼2 ×·106 cells/cm2) and dispersed nature of the fouling suggests that the mechanism is predominantly random and not the result of active biofilm formation. Since it is known that Clostridium sp. can form biofilms on porous surfaces (e.g., brick and bone char[59, 60]) and that hydrophobic surfaces can be particularly susceptible to increased biofilm formation, the limited degree of fouling observed here may have been due to the short duration of the experiment (72 h total with only 38.5 h of media contact with resin) and should be reinvestigated in subsequent, longer-term studies. The preliminary findings are, however, consistent with those of a prior report wherein the anion exchange adsorbent used to recover succinic acid from Actinobacillus succinogenes cultures via EBA was subjected to whole cell biofouling at a comparable level of approximately 1.85·106 cells/cm2 over the course of 126 h. In that case, fouling was easily reversed through simple water washing, a strategy that should likewise be tested for its efficacy with L-493. Meanwhile, from Figure 6 it also appears that biofouling by noncellular debris—perhaps extracellular polymeric substances (EPS)—may have occurred. EPS, which C. acetobutylicum ATCC 824 has been reported to produce as a prelude cell surface attachment and biofilm maturation, may be indicating that significant whole cell fouling in longer fermentations is a distinct possibility. Further characterization, however, would be required to confirm the nature of the observed noncellular debris, as well as its influence on fouling, biofilm formation and, most importantly, its effect on butanol adsorption and EBA performance.
Figure 6. SEM images of fresh L-493 at ×500 magnification (A), and L-493 recovered from the EBA column following fermentation magnified at ×500 (B), ×1,000 (C), and ×2,500 (D).
Asterisk and “+” symbols included as points of reference. Inset rectangles show approximate regions imaged at next magnification level, as appropriate.
Download figure to PowerPoint