(rPHAMAX, rFPMAX) were calculated by the slope of PHA or FP, respectively, as function of the cor-responding time period. the yield of PHA production per substrate consumed (YPHA/FP; COD basis) was calculated as follows: ∆PHA/ ∆FP and the yield of HHx produced by caproate con-sumed was calculated by the ratio between the HHx produced and the caproate concon-sumed on a COD basis: ∆HHx/∆Caproate.
Figure 2.1. Trends of the OLR applied to the reactor and of the concentrations of each FP in the outlet of the acidogenic reactor throughout the study.
In the first 2 weeks of operation, the gas phase of the reactor contained a relatively high quantity of oxygen, which was stripped out by the gas produced by the culture towards the end of this period (Figure 2.2). The end of the transition phase coincides with the end of a residual production of methane and a sudden increase in hydrogen concentration in the gas outlet stream. Methanogens were effectively outcompeted from the reactor around day 60 of operation, as indicated by the decrease in methane in the gas feed. This was achieved through a combination of unfavorable conditions to these microorganisms, namely low HRT and pH.
The inhibition of methanogenic activity freed up the precursors for methane production (e.g.
H2 and acetate) for caproate production by chain elongating bacteria. The gas outlet stabilized at a flow rate of 0.43 L min-1 with a high fraction of hydrogen (30%) and containing also carbon dioxide (59%), oxygen (3%) and nitrogen (8%).
Figure 2.2. Trends of fraction of each gas identified in the headspace of the UASB reactor during operation.
The microbial culture in the acidogenic reactor (suspended and granular biomass) was characterized through 16S rRNA gene sequencing analysis (Table 2.3). Despite the reactor con-figuration enabling the partition of the cells in both an aggregated (granules) and planktonic (suspended biomass) form, the microbial community profile did not change significantly be-tween the two. The genus Ruminiclostridium 5 was predominant throughout the whole oper-ation. Members of this genus (specifically strain CPB6) have previously been associated with a CE process using lactate as electron donor, giving caproate and caprylate [36]. In addition, genera Atopobium, Olsenella and Prevotella 7, which together constituted more than 40% of the biomass on day 129, may have also played an important role in the caproate-rich FP profile.
Several authors have reported that genera Atopobium and Olsenella are present in lactate-based CE processes and that they are responsible for the conversion of carbohydrates into lactate, thus becoming the driving force for CE [36–38]. In fact, the genera Atopobium, Olsen-ella and Ruminiclostridium 5 have been reported to be part of a MMC capable of converting lactate and xylose into caproate through chain elongation [36]. In that study, it was observed that caproate production decreased when the prevalence of these microorganisms decreased as well. Moreover, Ruminiclostridium 5 was also the most abundant of the three, suggesting
that this process requires the prevalence of a caproate producer rather than lactate producers.
In other studies, Prevotella 7 has been associated with SCFA and H2 production which are rel-evant in CE [39,40]. SCFA, such as acetate and butyrate, can be directly used as precursors for CE by caproate-producing microorganisms, such as Ruminiclostridium 5. However, for that to happen, sufficient H2 must be maintained in the reactor to prevent fatty acids (FA) oxidation [41,42]. The operation of the UASB in the current study at a pH of 5 resulted in the inhibition and inactivation of hydrogenotrophic methanogens, thus reducing the overall hydrogen con-sumption, leading to a high hydrogen partial pressure. A high partial pressure of hydrogen is known to provide the adequate conditions to the conversion of carbohydrates into lactate, which could then be converted into caproate. This was possible due to the enrichment of the adequate lactate- and caproate-producing populations in the reactor.
Table 2.3. Microbial community dynamics of 10 most abundant genera of the acidogenic reactor. Values indicate percentage of total number of reads. OUTs listed indicate phylum and the lowest assigned taxonomic classifica-tion that could be obtained. G samples correspond to granular biomass, S samples correspond to suspended
bio-mass (p - phylum; g - genus).
Day 65 Day 129 Day 171
G S G S G S
p_Firmicutes; g_Ruminiclostridium 5 81.5 93.3 87.8 56.8 92.9 85.8 p_Actinobacteria; g_Atopobium 1.6 0.6 5.6 15.7 6.5 12.8 p_Actinobacteria; g_Olsenella 5.6 4.1 4.9 16.7 0 0 p_Bacteroidetes; g_Prevotella 7 6.2 1.4 0.9 9.9 0.6 0.8
p_Firmicutes; g_Megasphera 1.7 0.1 0 0 0 0
p_Firmicutes; g_Selemonas 0.8 0.1 0 0 0 0
p_Firmicutes; g_Solobacterium 0.4 0.1 0.1 0 0 0
p_Firmicutes; g_Lactococcus 0 0.1 0 0.2 0 0.3
p_Firmicutes; g_Ruminococcaceae UCG-014 0.1 0 0.1 0.2 0 0 p_Firmicutes; g_Clostridium sensu stricto 11 0 0 0.2 0.2 0 0
The selection of these two types of organisms was specific for the operation conditions used in this study. Indeed, the operation of the same UASB in very similar conditions, including inoculum and feedstock [29], where the only difference was that the reactor was sparged with pure nitrogen at start-up and every time the reactor was opened for cleaning or probes cali-bration. This difference in operation resulted in a different microbial profile, which did not
produce any detectable amounts of caproate. In the current study, nitrogen sparging was not used and oxygen was allowed inside the reactor in the start-up of the UASB. Oxygen was then either slowly consumed during the two first weeks of operation by aerobic heterotrophs or simply stripped out eventually to a value below 5% as a result of gas production from fermen-tation reactions. We hypothesize that the microaerophilic conditions during start-up of the study may have been the trigger for the selection of a caproate-producing MMC. It is possible that the presence of oxygen enabled oxygen-tolerating bacteria to thrive over strict anaerobes.
In fact, Ruminiclostridium 5 has been reported to grow under microaerophilic conditions, Olsenella has been reported as a microaerophilic anareobic genus and Atopobium, as faculta-tive anaerobes [43–45]. A study conducted by Lambrecht and colleagues achieved a similar conclusion with a different consortium, where it was reported that a low oxygen partial pres-sure indirectly supported caproate formation by favoring lactate production over other more oxygen-sensitive fermentation pathways [37]. While this hypothesis still has to be validated, it is possible that the presence of microaerophilic conditions in the UASB operation might have shaped the microbial ecology of the MMC, enabling the production of caproate.
Sequencing batch reactor
The caproate-rich fermentate produced during the first stage of the process was used as carbon source to select a culture in the SBR capable of producing a HHx-rich polymer. The fermented FW used as SBR substrate had the following composition (gCODFP L-1): lactate (0.05
± 0.06), acetate (0.95 ± 0.16), propionate (0.07 ± 0.06), ethanol (0.48 ± 0.11), butyrate (2.25 ± 0.35), valerate (0.17 ± 0.26) and caproate (9.97 ± 1.72). The culture was selected by undergoing a series of cycles of feast and famine. During the feast phase, the SBR was fed with the carbon-rich fermented feedstock, while during the famine phase a nutrient solution containing nitro-gen and phosphorus was fed. The uncoupled supply of carbon and nutrients is an effective approach to enhance the selection of the PHA-accumulating organisms since only these are able to grow, using the stored carbon, when nutrients are provided in the famine phase (ab-sence of external carbon) [30]. Table 2.4 summarizes the SBR operation conditions.
Table 2.4. SBR performance parameters after pseudo-steady state
Parameter Average ± standard deviation
Ratio between feast phase length
and famine phase length (h h-1) 0.05 ± 0.05 PHA content at the end of feast
phase (gPHA gTS-1) 0.30 ± 0.02
Monomer composition at the end
of feast (HB/HV/HHx, % COD basis) 55 ± 0.6/9 ± 0.2/36 ± 0.8 rFPMAX (gCODFP L-1 h-1) -2.76 ± 0.79 rCapMAX (gCOD L-1 h-1) -1.83 ± 0.55 rValMAX (gCOD L-1 h-1) -0.24 ± 0.06 rButMAX (gCOD L-1 h-1) -0.78 ± 0.15 rProMAX (gCOD L-1 h-1) -0.08 ± 0.01 rAceMAX (gCOD L-1 h-1) -0.27 ± 0.11 rPHAMAX (gCODPHA L-1 h-1) 2.24 ± 0.19 rHHxMAX (gCOD L-1 h-1) 1.17 ± 0.26
rHVMAX (gCOD L-1 h-1) 0.22 ± 0.03
rHBMAX (gCOD L-1 h-1) 0.86 ± 0.14
YPHA/FP (gCODPHA gCODFP-1) 0.81 ± 0.24
YHHx/Cap (gCOD gCOD-1) 0.64 ± 0.24
YHB/(Ace+But) (gCOD gCOD-1) 0.82 ± 0.13
A short feast to famine ratio was observed throughout the reactor operation, averaging 0.05 ± 0.005 h h−1 in the pseudo-state phase. This condition is a well-known to foster the se-lection of a PHA-accumulating culture. The highest PHA content in the biomass, observed at the end of the feast phase, averaged 0.30 ± 0.02 gPHA gTS-1. The product consisted in a ter-polymer composed of HB, HV and HHx. Among the FP, caproate had the fastest volumetric consumption rate of -1.83 ± 0.55 gCOD L-1 h-1. The metabolic conversion of caproate into PHA in MMC is not fully understood yet. However, in pure cultures caproate is a known precursor of HHx monomers. The average conversion yield (YHHx/Cap) during pseudo-steady state opera-tion of the SBR was 0.64 ± 0.24 gCOD gCOD-1. The high yield of polymer on substrate (YPHA/FP
= 0.81 gCODPHA gCODFP-1) reveals that the conditions used in the SBR were effective for the selection of a culture enriched in PHA-storing microorganisms.
The community was dominated by well-established PHA-storing microorganisms, such the genera, Amaricoccus, Zoogloea, Paracoccus and Plasticicumulans, which have often been reported as the most predominant genera in PHA-producing MMC (Table 2.5) [46–49]. This study indicates that these organisms are also capable of taking up caproate and producing HHx. However, acclimatization to caproate is necessary, as discussed below.
Table 2.5. Semi-quantitative FISH characterization of the MMC enriched in the SBR during pseudo-steady state (c – class; g – genus).
(-), not present; (+), almost non-existent (1-5%); (++), present (5-20%); (+++), abundant (20-50%); (++++), dominant (> 50%)
Accumulation reactor
The successfully selected culture, capable of producing PHA and, specifically, the mono-mer HHx, was used as inoculum of the fed-batch accumulation reactor, which was fed with caproate-rich fermentate. Figure 2.3 shows the profiles of FP (2.3a) and PHA content in the biomass (2.3b) during a typical accumulation assay, where 8 pulses of caproate-rich fermentate were added. While caproate and butyrate were readily consumed, acetate was essentially con-sumed only after the former FP were depleted. Caproate was concon-sumed at a rate approx. 8-fold higher than butyrate. The fact that butyrate was preferred to acetate as substrate for PHA storage is in line with the work of Wang et al. [50]. The reason for this is the fact that the energetic cost of SCFA uptake is 1 mol of ATP per molecule, which means that less energy is used per C-mol when butyrate is taken up. Assuming that the general MMC metabolism is similar to that of PHA-accumulating isolates such as Pseudomonas putida, one mol of caproate
Relative abundance c_Alphaproteobacteria; g_Rhodobacter+Roseobacter (G Rb) +++
c_Alphaproteobacteria; g_Amaricoccus (AMAR839) +++
c_Alphaproteobacteria; g_Paracoccus (Par651) ++
c_Betaproteobacteria; g_Zoogloea (Zra23a) +++
c_Betaproteobacteria; g_Lampropedia (Lamp444) +
c_Betaproteobacteria; g_Azoarcus (Azo644) +
c_Betaproteobacteria; g_Thaurea (THAU832) +
c_Gamaproteobacteria; g_Plasticicumulans (UCB-823) ++
would be activated by an acyl-CoA dehydrogenase at the cost of 1 mol of ATP [51]. This means that the energetic requirement per C-mol for the cell to metabolize caproate is even lower than butyrate, which explains the preference and higher consumption rates observed. Since nitro-gen was not present in the feed, cell growth is negligible and approximately all FP were used for PHA production, resulting in conversion yields around 100% and a PHA content of 0.71 gPHA
gTS−1. The final product consisted of a terpolymer of HB/HV/HHx in the following proportion:
33/1/66% wt. (Table 2.6). The final HHx content in the biomass was higher than the HB content, alike to the caproate and HB precursors proportions in the feed, which suggests that caproate is the direct precursor for HHx production. HV monomers were barely produced since odd-chain FP such as valerate and propionate, HV precursors, were not present in the fermentate using in the accumulation assays.
Figure 2.3. Concentration of FP (a) and PHA content and monomer content in the biomass (b) in a typical accumu-lation assay.
The HHx content reported in this study is the highest obtained to date with a MMC. The maximum HHx content in PHA previously reported for an MMC was 6.3% (mol basis), where a synthetic feed containing sodium laureate was supplemented [27]. The productivity obtained in the current pilot scale study is roughly 5 times higher than Shen et al., and the PHA content achieved a maximum of 0.71 gPHA gTS-1 which was 3 times higher than Shen and colleagues [27].
The conversion yield obtained by Ntaikou et al. [28] was also considerably lower (0.12-0.15), than the one here reported. Consequently, this study has significantly improved on each pa-rameter when compared to the two most relevant studies previously reported in the literature.
Table 2.6. Summary of the parameters obtained in the accumulation reactor at pilot scale; Feed concentration (gCODFP L-1): [Lac] = 0.09, [Ace] = 1.03, [But] = 1.61, [Val] = 0.03, [Cap] = 11.5
Kinetic assays
In order to investigate the metabolic transformations of caproate, as well as the remain-ing FP, in the PHA-accumulatremain-ing process, a series of 5 batch tests were carried out. The same PHA-accumulating biomass, selected using a caproate-rich fermentate, was used as the inoc-ulum in all the tests. In Test A, the reactor was fed with 3 pulses of fermented fruit waste. In Test B, a synthetic medium mimicking the real fermentate used in Test A, was used to confirm that switching from fermented fruit waste to synthetic medium would not affect the results.
Test C was carried out with a synthetic medium without caproate and Test D was carried out with a synthetic medium comprised only of caproate, aiming at understanding the behavior of the selected culture when fed with caproate as the only substrate. A final test (Test E) was performed with a similar feed as used in Test A, but using as inoculum a PHA-accumulating culture that had not been selected with a caproate-rich fermented fruit waste. The goal was to
Parameter Value
rPHA (gCODPHA L-1 h-1) 3.17 ± 1.27
rHHx (gCOD L-1 h-1) 2.09 ± 0.99
rHB (gCOD L-1 h-1) 1.06 ± 0.24
rFP (gCODFP L-1 h-1) -3.39 ± 0.77
rCap (gCOD L-1 h-1) -6.61 ± 0.94
rBut(gCOD L-1 h-1) -0.80 ± 0.11
rAce (gCOD L-1 h-1) -0.48 ± 0.21
Final PHA content (gPHA gTS-1) 0.71 ± 0.01
YPHA/FP (gCODPHA gCODFP-1) 1.00 ± 0.33
YHHx/Cap (gCOD gCOD-1) 0.44 ± 0.14
PPHA (gCODPHA L-1 h-1) 3.29
PHHx (gCOD L-1 h-1) 2.31
Final monomer compostion (HB/HV/HHx, % wt.) 33/1/66
understand the impact of acclimatization of the biomass to caproate on HHx production. Fig-ure 2.4 represents FP consumption over time for each of the batch tests, and FigFig-ure 2.5 depicts the profile of PHA and monomer content in the biomass in each test. Similarly to what was observed in the accumulation test at pilot scale, each pulse fed to the reactor was readily con-sumed, eventually resulting in complete consumption of all FP. As the FP were concon-sumed, PHA was produced. A summary of the results is shown in Table 2.7.
Table 2.7. Summary of the parameters obtained in the kinetic batches at lab scale
Parameter
Test A Fruit waste
(FW)
Test B Synthetic
FW
Test C Synthetic no
caproate
Test D Synthetic
only ca-proate
Test E Non accli-matized
bio-mass and FW rPHAMAX (gCODPHA L-1 h-1) 4.19 ± 1.14 6.24 ± 0.05 2.24 ± 0.15 4.22 ± 1.09 3.19 ± 0.30
rHHxMAX (gCOD L-1 h-1) 8.17 ± 0.09 8.91 ± 0.85 -0.04 ± 0.23 3.44 ± 0.97 1.04 ± 0.15 rHVMAX (gCOD L-1 h-1) n.a. n.a. n.a. n.a. 0.42 ± 0.36 rHBMAX (gCOD L-1 h-1) 2.60 ± 0.27 2.50 ± 0.15 1.47 ± 0.25 0.98 ± 0.12 2.21 ± 0.42 rFPMAX (gCODFP L-1 h-1) -4.24 ± 1.25 -7.21 ± 0.88 -2.32 ± 0.11 -4.56 ± 0.37 -3.57 ± 0.40
rCapMAX (gCOD L-1 h-1) -11.0 ± 1.0 -9.86 ± 0.94 n.a. -4.85 ± 0.41 -3.13 ± 0.45 rButMAX (gCOD L-1 h-1) -0.97 ± 0.14 -1.30 ± 0.18 -2.82 ± 0.69 0.11± 0.02 -0.83 ± 0.13 rAceMAX (gCOD L-1 h-1) 0.76 ± 0.18 0.88 ± 0.65 -1.54 ± 0.35 0.42 ± 0.03 -1.24 ± 0.01
YPHA/FP (gCODPHA
gCODFP-1) 0.99 ± 0.40 0.87 ± 0.10 0.97 ± 0.11 0.93 ± 0.25 0.90 ± 0.19 YHHx/Cap (gCOD gCOD-1) 0.74 ± 0.07 0.91 ± 0.17 n.a. 0.71 ± 0.37 0.34 ± 0.10
Initial monomer com-postion (HB/HV/HHx, %
wt.)
18/5/78 16/0/84 13/0/87 11/0/89 75/25/0
Final monomer compo-sition (HB/HV/HHx, %
wt.)
53/2/46 30/0/70 68/1/31 22/0/78 73/6/22
n.a., not applicable;
Figure 2.4. Trends of FP concentration over time in the kinetic batches at lab scale.
2.3.4.1 The influence of non-FP organic matter in the tests (Test A and B)
In the case of Test A and Test B, the culture responded slightly differently to the two substrates. Similar trends of FP consumption were obtained, where caproate was the most rapidly consumed FP, whereas acetate was only consumed when the other FP were depleted.
In fact, acetate concentration tended to increase in these tests while caproate and butyrate were being consumed (Figure 2.5A and Figure 2.5B). Despite the difference in the final mono-mer composition, similar maximum HB and HHx production rates were obtained in both tests.
This difference in final monomeric composition could potentially be attributed to the fraction of caproate metabolized into HHx, which was lower in Test A, resulting in a lower content of
HHx in the polymer. Nonetheless, in these tests only HB and HHx were produced, and the kinetic performance was reasonably similar in both tests, the small difference observed prob-ably being justifiable by the fact that the fermented FW had different non-FP composition from the chemically defined solution used in Test B.
Figure 2.5. PHA content in the biomass during the kinetic batch tests at lab scale.
2.3.4.2 The fate of caproate in PHA-production with caproate-acclimatized biomass (Test B, C and D)
As shown in Table 2.7, the total FP consumption rates were higher when caproate was present in the FP feed (tests A, B and D), as a result of the higher caproate consumption rate.
However, when caproate was not present, butyrate consumption rate was higher, suggesting
that butyrate was the second preferred substrate in the absence of a longer FP. These results are in agreement with the energetic requirements for uptake of individual FP, as discussed above. In these tests, it was also found that caproate was partially converted into short-chain fatty acids such as acetate and butyrate (Figure 2.4D). In Test B the conversion of caproate into HHx was the highest, which means that the loss of caproate into SCFA was smaller than in the other two tests.
2.3.4.3 The fate of each FP on the production of HB, HV and HHx (Tests B, C, D and E) The composition of the media fed in Tests B, C and D was defined to mimic the compo-sition of the real FW fermentate used in Test A, which happened to be exclusively comprised of even-chain compounds. Therefore, only even-chain monomers such as HB and HHx were produced, as expected. Regarding the production of HHx, it was clear that it resulted from the metabolism of caproate. Firstly, in Test D, HHx was produced when the biomass was exposed to caproate only, thus suggesting that caproate was converted into HHx. But since conversion caproate conversion into butyric acid and acetic acid has been previously shown, this conclu-sion could not necessarily be true. However, if Test C is also taken into account, when caproate was absent from the feed, HHx content in the biomass was essentially constant over time, confirming that caproate was the source of the HHx produced in the other tests. Since precur-sors for HB, such as acetate and butyrate, were present, HB was produced in this test. Finally, Tests A and B confirm that at least some caproate must have been used for HHx production, as known HB precursors such as acetate and butyrate could not account for all the HHx pro-duced, which would mean that at least some HHx would have been produced by caproate.
However, considering that YHHx/Cap was never close to 1, that all caproate was consumed and that growth was not possible as no nitrogen was available, it means that the culture used some caproate to produce HB. The production of acetate from caproate is also consistent with this result. Regarding acetate and butyrate, in this study, the culture used them to produce HB, which is in line with the mechanisms described in literature [50]. It is also interesting to notice that the highest maximum PHA production rate was obtained in Test B, where all FP were present. In Test E, due to the presence of some valerate which is a known HV precursor, a co-polymer containing three monomers was produced. As far as the authors know, this is the first report of a ter-polymer of PHA produced using mixed microbial cultures from a real waste fermentate.
2.3.4.4 The importance of the acclimatization to caproate on the production of HHx-contain-ing PHA (Tests A and E)
Tests A and E, performed to investigate the need for acclimatization to caproate, revealed some important differences in performance. Firstly, in Test E, neither butyrate nor acetate con-centration increased during the test, unlike Test A. In fact, Test E was the only test where acetate was consumed as soon as it was fed, even when caproate was present. Secondly, the rates of PHA production as well as FP consumption were lower in Test E. Furthermore, the yield of conversion of caproate into HHx (YHHx/Cap) was the lowest of all tests in Test E, which means that the biomass not acclimatized to the caproate could not convert as much of the caproate into HHx as a biomass acclimatized to caproate. Considering that all the caproate was consumed in Test E, caproate was likely converted into HB, resulting in a polymer with lower HHx content as compared to the other tests. This test points out the importance of acclimatizing the culture to the same substrate to be used in the accumulation stage, otherwise a different metabolism may occur in the last step of the PHA production process.
The fate of caproate
According to the most recent literature from MMC studies, SCFA can be metabolized into PHA according to the metabolic reactions shown in Figure 2.6 [24,52]. In summary, each SFCA is firstly activated into its corresponding acyl-CoA. While acetyl-CoA/propionyl-CoA are con-densed and reduced before being added into a PHA chain, and butyryl-CoA/valeryl-CoA un-dergo the first 2 steps of β-oxidation before being added into a PHA chain. The metabolism of caproate is not well established for MMC, but based on the results obtained in this study, it can be hypothesized that caproate undergoes a similar pathway as butyrate and valerate, being converted into (R)-3-hydroxyhexanoate by β-oxidation. At this point, (R)-3-hydroxyhexanoate can be either directly incorporated into a PHA chain, hence resulting in a polymer containing HHx, or cleaved into Acetyl-CoA and Butyryl-CoA by a β-ketothiolase. These products can then undergo known metabolic reactions as described in Figure 2.6. According to the results of this paper, the selection of a MMC in a SBR with caproate has a direct impact on the polymer production process in the accumulation reactor. As the results in the batch tests A to D show, the selection of a MMC with a caproate-rich fermented feedstock resulted in a polymer with a higher content of HHx when compared to Test E, using a MMC that had not been selected with caproate-rich feedstock. One possible reason for this behavior is that microorganisms selected using a fermentate absent of caproate simply cannot convert caproate into HHx, as it requires
a different type of PHA synthase (class 2 and 4, as opposed to class 1 and 3 for shorter-chain length PHA) [53]. As a consequence, in Test E, as caproate was uptaken by the cell, it underwent β-oxidation and it was converted into acetyl-CoA and butyryl-CoA which were used to elongate a PHA chain using a class 1 or class 3 PHA synthase into PHA. Alternatively, it could also be hypothesized that a minimum caproate concentration is necessary to activate the enzymes required to metabolize caproate.
Implications of the study
PHA production using MMC often results in the production of a polymer containing up to two monomers: HB and HV. Other monomers such as HHx, confer the polymer with elasto-meric properties, allowing the polymer to be used in different applications than the typical PHBV. Although a few MMC studies have previously reported the presence of HHx in their PHA products [25–28], the yield and productivity values achieved in the past were too low to be interesting for full scale implementation. Moreover, in literature studies using the 3-stage pro-cess, the precursor for HHx production was added as a supplement to the accumulation reac-tor. The current study is the first to fine-tune a complete 3-stage process, at pilot scale, to produce HHx-containing PHA from a waste feedstock. Unprecedented productivities were ob-tained for both the caproate and the HHx-containing polymer, corresponding to 1.3 and 5 fold the highest values previously reported for caproate [35] and HHx [27], respectively. These pilot-scale results are a significant advancement towards the full-pilot-scale implementation of HHx pro-duction from wastes by mixed cultures.
This work additionally unveiled key factors supporting the successful operation of the 3-stage process towards HHx production. Microaerofilic conditions in the start-up of the acido-genic reactor may help enhance the selection of caproate-producing microorganisms, whereas the presence of caproate in the selection stage, and not just the accumulation stage as in some previous studies [28], seemed to be detrimental for obtaining HHx-rich PHA polymers. The uncoupled C and N regime further contributed to the high productivities achieved, as previ-ously reported by Matos et al. [29]. It was also demonstrated that caproate was the precursor for HHx production, which had been shown in pure cultures [e.g. [10]] but not in MMC. Indeed, an additional contribution of this work was the proposal of two metabolic mechanisms for conversion by MMC of a caproate-containing feed into HHx-containing PHA. This more fun-damental aspect of the work is important to incorporate these pathways in mathematical mod-els in the future, towards further optimization and process control.