Waste Management 127 (2021) 80–89

Contents lists available at ScienceDirect

Waste Management
journal homepage: www.elsevier.com/locate/wasman

Selecting fermentation products for food waste valorisation with HRT

and OLR as the key operational parameters
Vicky De Groof a,b, Marta Coma c, Tom Arnot b,c,d, David J. Leak c,d,e, Ana B. Lanham b,d,⇑
a
EPSRC Centre for Doctoral Training in Sustainable Chemical Technologies, University of Bath, Claverton Down, Bath BA2 7AY, UK
b
Department of Chemical Engineering, University of Bath, Claverton Down, Bath BA2 7AY, UK
c
Centre for Sustainable and Circular Technologies (CSCT), University of Bath, Claverton Down, Bath BA2 7AY, UK
d
Water Innovation & Research Centre (WIRC), University of Bath, Claverton Down, Bath BA2 7AY, UK
e
Department of Biology & Biochemistry, University of Bath, Claverton Down, Bath BA2 7AY, UK

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

Article history: Acidogenic fermentation is attractive for food waste valorisation. A better understanding is required on
Received 18 December 2020 how operation affects product selectivity. This study demonstrated that the hydraulic retention time
Revised 29 March 2021 (HRT) and organic loading rate (OLR) selected fermentation pathways in a single-stage, semi-
Accepted 10 April 2021
continuous stirred tank reactor. Three combinations of HRT and OLR were tested to distinguish the effect
Available online 28 April 2021
of each parameter. Three fermentation profiles with distinct microbial communities were obtained.
Predominantly n-butyric acid (13 ± 2 gCOD L
1, 55 ± 14% of carboxylates) was produced at an HRT of
Keywords:
8.5 days and OLR around 12 gCOD L
1d
1. Operating at an HRT two days longer, yet with similar OLR,
Acidogenic fermentation
Butyric acid
stimulated chain elongation (up to 13.6 gCOD L
1 of n-caproic acid). This was reflected by a microbial
Caproic acid community twice as diverse at longer HRT as indicated by first and second order Hill number (1D = 24
Food waste valorisation ± 4, 2D = 12 ± 3) and by a higher relative abundance of genera related to secondary fermentation, such
Lactic acid as the VFA-elongating Caproiciproducens spp., and secondary lactic acid fermenter Secundilactobacillus
Resource recovery spp.. Operating at a higher OLR (20 gCOD L
1d
1) but HRT of 8.5 days, resulted in typical lactic acid fer-
mentation (34 ± 5 gCOD L
1) harbouring a less diverse community (1D = 8.0 ± 0.7, 2D = 5.7 ± 0.9) rich in
acid-resistant homofermentative Lactobacillus spp. These findings demonstrate that a flexible product
portfolio can be achieved by small adjustments in two key operating conditions. This improves the eco-
nomic potential of acidogenic fermentation for food waste valorisation.
Ó 2021 Published by Elsevier Ltd.

1. Introduction bio-economy and reach international sustainability targets

(European Commission, 2019; Gasper et al., 2019). However, land-
Effective food waste (FW) management systems should provide filling and incineration are still widely used for FW disposal, which
cost-efficient resource and energy recovery to enable a circular lead to greenhouse gas emissions, are energy intensive and result
in an irrevocable loss of resources (Edwards et al., 2017; Kibler
et al., 2018). To attain the full material and economic potential
Abbreviations: AD, anaerobic digestion; C1, formic acid, formate; C2, acetic acid,
acetate; C3, propionic acid, propionate; C4, n-butyric acid, n-butyrate; C5, n-valeric
of FW valorisation, recycling systems that allow for an assort-
acid, n-valerate; C6, n-caproic acid, n-caproate; C7, n-heptanoic acid, n-heptanoate; ment of products are key (Engelberth, 2020; Venkata Mohan
C8, n-caprylic acid, n-caprylate; CA, carboxylic acids and carboxylates; CSTR, et al., 2016).
continuous stirred tank reactor; F/M, feed-to-microbial ratio; FW, food waste; HH/ A range of promising technologies are emerging that use acido-
LO, reactor operating condition at high HRT and low OLR; HRT, hydraulic retention
genic fermentation with anaerobic microbial communities to gen-
time; LH/HO, reactor operating condition at low HRT and high OLR; LH/LO, reactor
operating condition at low HRT and low OLR; MCCA, medium chain carboxylic acids erate different value-added products from bio-wastes, often
(C5-C8); NA, not applicable / not available; OLR, organic loading rate; PHA, alternative or supplementary to more conventional anaerobic
polyhydroxyalkanoate; SBR, sequencing batch reactor; STR, stirred tank reactor; t/s digestion (AD) (Capson-Tojo et al., 2016; Moscoviz et al., 2018).
COD, total/soluble chemical oxygen demand; TS, total solids; VFA, volatile fatty For instance, carbohydrates in FW can be fermented to produce
acids (C1-C4); VS, volatile solids.
⇑ Corresponding author at: Department of Chemical Engineering, University of lactic acid, ethanol, volatile fatty acids (carboxylic acids with 1 to
Bath, Claverton Down, Bath BA2 7AY, UK. 4 carbon atoms, VFA) or hydrogen (Im et al., 2020; Tang et al.,
E-mail address: [email protected] (A.B. Lanham). 2016; Zhang et al., 2020). Consecutive fermentation steps can

https://doi.org/10.1016/j.wasman.2021.04.023
0956-053X/Ó 2021 Published by Elsevier Ltd.
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

occur, for instance, lactic acid can be further metabolised to VFA, or particular feedstock and vice versa, which is often overlooked. In
VFA can be elongated with lactic acid or ethanol as electron donor addition, limited studies provide a systematic investigation of the
to generate medium chain carboxylic acids (5–8 carbon atoms, impact of these operating parameters on the microbial community.
MCCA) (Asunis et al., 2019; Contreras-Davila et al., 2020). These Improved understanding is required on how the OLR and HRT each
products can be extracted to serve as renewable commodity chem- direct fermentation pathways so that an operating strategy can be
icals or liquid fuels (Bhatia & Yang, 2017; Jang et al., 2012; designed to target each specific product in acidogenic FW fermen-
Kannengiesser et al., 2016). Alternatively, the enriched fermenta- tation in CSTRs. This would allow developing manageable operat-
tion effluent could serve as feedstock for other biological processes, ing strategies similar as currently applied in AD that provide a
e.g., polyhydroxyalkanoate (PHA) production, a biodegradable sub- broad and flexible product portfolio for FW valorisation. Therefore,
stitute to fossil fuel plastics (Girotto et al., 2017). this study aimed at untangling the effect of OLR and HRT on acido-
However, microbial communities in acidogenic fermentation genic fermentation in a semi-continuous STR system. To observe
have a wide and complex metabolic capacity, which makes it chal- the impact of HRT and OLR on product outcome and on the respec-
lenging to steer acidogenic fermentation selectively towards a tar- tive microbial community, duplicate semi-continuous STRs were
get product (Arslan et al., 2016). Generally, specific fermentation operated at three different conditions: high HRT with low OLR
products are targeted in studies by manipulating operating condi- (HH/LO), low HRT with low OLR (LH/LO), and low HRT with high
tions, such as the operating temperature, pH, applied organic load- OLR (LH/HO). These were selected to ensure an OLR and HRT suit-
ing rate (OLR) and retention times (Zhou et al., 2018). For instance, able for acidogenic fermentation while minimising carbon loss to
various parameters influence chain elongation of VFA into MCCA methanogenesis. Previous reports demonstrated that an OLR above
when fermenting a complex organic feedstock with microbial com- 8.5 and up to 21 gCOD L
1 d
1 accumulated a range of VFA and
munities (De Groof et al., 2019). However, when it comes to MCCA in a similar setup of semi-continuous FW fermentation
industrial-scale operation as currently applied in AD, control is (De Groof et al., 2020b). These values were thus chosen as limits
most commonly exerted by manipulating the feeding rate, which for low and high OLR (LO at 12 gCOD L1d
1 and HO at 20 gCOD L
1-
determines retention time and OLR (Nguyen et al., 2015). Continu- d
1). An HRT above 8 days was selected to ensure a sufficient level
ous stirred tank reactors (CSTRs) are the most common reactor of hydrolysis (low HRT, LH at 8.5 days)(Lim et al., 2008). The upper
configuration for AD on commercial scale due to their ease of oper- limit (high HRT, HH at 10.5 days) was set to still allow wash-out of
ation and low investment and operating costs (Bensmann et al., methanogenic species (Fdez.-Güelfo et al., 2013). The results pro-
2013). Being able to select for target acidogenic fermentation prod- vide a concrete strategy to direct acidogenic fermentation of FW
ucts within a CSTR configuration by adjusting the HRT or OLR, thus into different and specific products of interest, and thus, demon-
similar as current AD operation, would be very attractive as sup- strate the opportunity of repurposing existing single stage AD
plementary FW valorisation technology. assets to diversify the product range from FW valorisation.
Retention time can be defined as i) in relation to the liquid, i.e.,
the hydraulic retention time (HRT); or ii) the solid feedstock frac-
tion and microorganisms, i.e., the sludge retention time (SRT). 2. Materials and methods
The SRT drives the composition of the microbial community via
differential growth rates between species. For instance, due to 2.1. Feedstock and inoculum
the slow growth rate of methanogens, AD must operate at long
SRT values. For the same reason, acidogenic fermentation can be The feedstock comprised unpackaged Category 3 mixed FW
selected instead of methanogenesis by reducing the SRT to<5– from a full-scale industrial digestion plant (GENeco, Bristol, UK).
8 days (Kim et al., 2016; Vanwonterghem et al., 2015). The HRT The FW is a relatively homogeneous slurry-like mixture obtained
dictates the time available for the feedstock to be hydrolysed and in the plant by grounding municipal FW from kerbside domestic
fermented, and hence overall throughput. For complex organic collection, FW beyond the sell-by date from supermarkets, and
feedstocks, acidification generally improves by operating at reten- FW collections from restaurants and other catering services, with
tion times of 15 days or more (Bolaji & Dionisi, 2017; Jankowska liquid streams from the food-processing industry and/or the liquid
et al., 2015). In addition, increased HRT generally results in a shift fraction of AD effluent.
in the product spectrum towards more reduced compounds, i.e., FW composition from this full-sized AD plant was determined
longer carboxylic acids (Arslan et al., 2016). based on a rigorous simulation exercise carried out on thousands
A combination of the HRT and the organic content of the feed- of data points across a two-year period from 2013 to 2014 (Table 1).
stock will determine the organic loading rate (OLR) and the feed- Forgacs et al. estimated the readily bio-degradable fraction, non-
to-microbial ratio (F/M) exerted (Arslan et al., 2016). Raising the biodegradable fraction etc. using BioWin 4.1 (EnviroSim Associates
OLR to increase F/M is a strategy applied to target carboxylic acid Ltd., Canada) to run a steady-state simulation of the full-sized AD
production by inhibiting methanogenesis via organic overload site. BioWin implements the IWA ADM1 model of AD (Batstone
(Coma et al., 2017). Extensive literature reviews show that an et al., 2002). Some of the key substrate characteristics were anal-
increase in OLR generally increases carboxylic acid production, ysed at the GENeco Avonmouth laboratories, including solids,
up to a peak between 50 and 100 gCOD L
1d
1, which then pla- COD, VFA, nitrogen and ammonia content (Forgacs et al., 2015;
teaus or declines (Khan et al., 2016). In addition, operating at Forgacs et al., 2017).
increased OLR can be used as strategy to shift AD of FW towards Two batches of FW were collected, further characterised
VFA accumulation and further to stimulate chain elongation for (Table 2) and stored in aliquots for feeding purposes at
18 °C.
MCCA production (De Groof et al., 2020b). However, other studies The measured TS and total COD were within the minima and max-
have found that increasing the OLR in acidogenic FW fermentation ima range of the values obtained during the two-year simulation
systems stimulated lactic acid production rather than carboxylic exercise (5.49 to 18.5% for TS and 69.2 to 264.0 gCOD L
1 for total
acids (Bo & Pin-Jing, 2014; Wu et al., 2015). COD) (Forgacs et al., 2015; Forgacs et al., 2017). Thus, the profile of
Thus, depending on the study, either VFAs, lactic acid or MCCAs FW received by the plant during the separate simulation exercise is
were targeted by manipulating OLR or retention times. However, in the same as that when feedstock samples were taken for the work
CSTRs, the SRT equals the HRT, and the OLR is inversely related to presented in this paper. In addition, the FW composition data are
the HRT. Thus, the selection of an optimal HRT or SRT to target a consistent for data on FW published by other authors (Zhang
specific product will affect which OLR can be applied with a et al., 2013; Zhang et al., 2007).
81
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

Table 1 Table 3
Food waste characteristics from the Avonmouth (Bristol) site operated by GENeco, UK Operating conditions for semi-continuous fermentation of food waste (FW) –
as determined by Forgacs et al. (2015, 2017). duplicates were used.

Food Waste Fractional Unit Fractional Condition HRT (d) OLR (gCOD L
1 d
1) FW batch
Characteristics composition HH/LO 10.4 ± 1.4 12.7 ± 1.6 1
Readily biodegradable gCOD g
1 tCOD 0.280 LH/LO 8.5 ± 0.5 11.9 ± 0.6 2 (2/3 dilution)
Acetic acid gCOD g
1 readily 0.250 LH/HO 8.3 ± 0.4 19.6 ± 1.0 2
biodegradable COD
Non-colloidal slowly gCOD g
1 slowly 0.900
biodegradable degradable COD
Non-biodegradable soluble gCOD g
1 tCOD 0.030
5.9. The FW used for LH/LO systems was diluted with tap water
Non-biodegradable particulate gCOD g
1 tCOD 0.130
Ammonia g NH3-N g
1 TKN 0.250 to obtain the required organic content as determined by the oper-
Particulate organic nitrogen g N g
1 Organic N 0.400 ating OLR (Table 3). Gas production was monitored via water dis-
Soluble non-biodegradable TKN gN g
1 TKN 0.040 placement using 2 L graduated glass columns containing acidified
N:COD ratio for non- g N g
1 COD 0.070 water (pH < 3.5, HCl) to avoid carbon dioxide absorption in the
biodegradable components
COD
water, and calibrated at Standard Temperature and Pressure (STP,
Phosphate g PO4-P g
1 TP 0.750 273.15 K and 100 kPa). The reactors were operated airtight, but
P:COD ratio for non- g P g
1COD 0.011 during feeding and pH correction, these were briefly opened to
biodegradable components the atmosphere. The columns were connected to the reactor head-
COD
spaces via a buffer bottle to prevent liquid from the columns from
entering the reactors.
At start-up, reactors were inoculated and diluted with tap water
Each set of duplicate semi-continuous reactors were operated to obtain a biomass content of approx. 14 gVS L
1. They were then
with the same feed throughout the experiment. left to acclimatise overnight to operating temperature (35 °C). Dur-
The fermentation runs HH/LO and LH/LO were inoculated with ing start-up for low OLR fermentations (LH/LO and HH/LO), the
an acidogenic fermentation culture from an in-house reactor that first feed was diluted 50% with tap water. For the higher OLR fer-
operated under similar conditions, which was stored at 4 °C (pH mentation (LH/HO), the first feed cycle comprised a full feed. Day
5.41 ± 0.01, VS 187 ± 4 g L
1) (De Groof et al., 2020b). The LH/HO 0 of operation corresponds to the first day of feeding at full organic
systems were inoculated with effluent from HH/LO reactors (pH strength.
5.65 ± 0.06, VS 52.8 ± 0.1 g L
1).
2.3. Evaluation of fermentation performance
2.2. Experimental set-up and reactor operation
To monitor fermentation performance over the full length of
For each combination of OLR and HRT, a pair of 2 L glass biore- reactor operation, liquid fermentation products and pH were anal-
actors were operated in duplicate, each with a 0.6 L working vol- ysed in each effluent sample, solids and COD were measured
ume, vertical mechanical stirrers (Bioprocess Control, Lund, weekly and gas composition in the water displacement columns
Sweden), and heated at 35 °C by a warm water bath. Reactors were was evaluated before feed addition (for details on analysis see Sec-
operated semi-continuously. The semi-continuous operation tion 2.4). In addition, after operating for approx. three HRT to allow
involved sampling and feeding every 3.5 days, i.e., a fixed volume the system to stabilise, cycle studies between feeding events were
of reactor effluent determined by the operating HRT was replaced performed in situ, where periodic samples for analysis of pH and
manually with an equal volume of thawed FW (Table 3). After product composition were taken from the reactors (without open-
addition of fresh feed, the pH was recorded and corrected manually ing the reactor headspace), to elucidate the underlying fermenta-
to 5.9 ± 0.1 by addition of sodium hydroxide (1 or 2 M) to minimize tion metabolism. Samples were immediately processed for
acid inhibition. The pH was not further corrected in between feed- product analysis. Only one of the two HH/LO reactors was sub-
ing events. Overall, pH ranged from minima of 4.8 to maxima of jected to a cycle study as the second reactor had ingested water
from the water displacement gas measurement column in the final
feeding cycle due to under pressure.
Table 2
The rate of acidification was calculated as the amount of OH–
Characteristics of the two batches of food waste (FW) collected and the reactor
system in which they were used. HH/LO was operated at 10.4 ± 1.4 days HRT and required to correct the pH in the reactor at the point of feeding,
12.7 ± 1.6 gCOD L
1d
1 OLR, LH/LO at 8.5 ± 0.5 days HRT and 11.9 ± 0.6 gCOD L
1d
1 normalized over the time in between feeding events (Eq. (1)).
OLR, and LH/HO at 8.3 ± 0.4 days HRT and 19.6 ± 1.0 gCOD L
1d
1 OLR. Samples were
C NaOHðiÞ  V NaOHðiÞ
analysed in duplicates. AcidificationðiÞ ¼ ½mMd
1 ð1Þ
V reactor  ðt ðiÞ
t ði
1Þ Þ
Parameter Unit FW1 FW2
Reactor run HH/LO LH/LO, LH/HO where i represents the feeding event, C NaOH is the molar concentra-
pH 4.05 ± 0.10 4.07 ± 0.02 tion of the sodium hydroxide solution applied, V represents volume
Conductivity mS cm
1 6.9 ± 0.2 8.3 ± 0.4 in L, and t represents the day of reactor operation, with tðiÞ
t ði
1Þ
Total Solids (TS) % w/w 10.7 ± 0.7 11.3 ± 0.2
being the time between feeding events. The OLR, HRT and net pro-
Volatile Solids (VS) % w/w 8.8 ± 0.3 10.3 ± 0.2
Total COD gCOD L
1 130 ± 9 163 ± 2 duct yields (Y p ) were calculated for each feeding cycle as a proxy to
Soluble COD gCOD L
1 63.7 ± 0.8 76.4 ± 1.5 continuous operation according Eqs. (2), (3) and (4) (De Groof et al.,
Ethanol g L
1 12.7 ± 0.5 4.9 ± 0.4 2020b).
Lactic acid g L
1 19.7 ± 1.3 21.7 ± 0.2
Acetic acid g L
1 4.5 ± 0.3 2.6 ± 0.2 C feedðiÞ  V feedðiÞ C feedðiÞ
g L
1 OLRðiÞ ¼ ¼ ½gCODL
1d
1; ð2Þ
n-Propionic acid 0.7 ± 1.0 0.7 ± 0.6 V reactorðiÞ  ðt ðiþ1Þ
t ðiÞ Þ HRT ðiÞ
Other VFA and MCCA g L
1 0.0 ± 0.0 0.0 ± 0.0
Glucose g L
1 1.4 ± 0.3 0.1 ± 0.1
Sugar compounds* g L
1 0.1 ± 0.3 0.3 ± 0.1 V reactorðiÞðtðiþ1Þ
tðiÞ Þ
HRT ðiÞ ¼ ½d; ð3Þ
* fructose, overlap with sucrose and xylose.
V feedðiÞ

82
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

C pðiÞ;effluent
C pði
1Þ;feed 3. Results and discussion

Y p ð iÞ ¼ ½%; ð4Þ
C feed
3.1. Stimulating chain elongation at longer retention times
where C p is the product concentration in the effluent in gCOD L
1
and C feed is the total COD in the feed in gCOD L
1. The average
The effects of HRT and OLR were evaluated separately by com-
OLR and HRT were then obtained by averaging over the entire oper- paring the performance of semi-continuous fermentation of FW in
ation. The average product concentration, Y p and rate of acidifica- duplicate STRs. Reactors operated as HH/LO were fed with the first
tion were calculated by averaging the values obtained at each batch of substrate, FW1, at a HRT of 10.4 ± 1.4 days and OLR of
feeding event. This was done for each operating condition tested 12.7 ± 1.6 gCOD L
1 d
1. To determine the impact of HRT, a second
and started after one full HRT cycle to exclude start-up conditions. batch of substrate was obtained, FW2, and diluted to allow opera-
All data has been made available as a dataset (De Groof et al., tion of the LH/LO reactors at an HRT that was 2 days lower than
2020a). HH/LO but with a similar OLR (Table 3). The total carboxylic acid
(CA) yield was similar for HH/LO and LH/HO, i.e. Y CA of 26 ± 3%
2.4. Chemical analysis and 22 ± 4% respectively. However, the product distribution dif-
fered (Fig. 1).
C1-C4 carboxylic acids, ethanol, lactic acid, glucose were mea- The LH/LO system mainly produced n-butyric acid (C4, Y C4 of
sured by high pressure liquid chromatography with a refractor 13 ± 2%), resulting in an average concentration of C4 in the effluent
index detector as previously described (Coma et al., 2017) with of 13 ± 2 gCOD L
1, i.e., 55 ± 14% of the total CAs (Fig. 1-B). In other
the oven temperature adjusted to 65 °C. The more hydrophobic studies focusing on butyrate-type FW fermentation selectivity was
MCCAs (C5-C8) were analysed by gas chromatography (GC) improved to 80% by, for example, control of pH at 6 (Jin et al., 2019;
equipped with flame ionization detector (FID), and the gas compo- Wang et al., 2014). In the LH/LO systems in this work the pH
sition (N2, O2, H2, CH4 and CO2) for the HH/LO and LH/HO fermen- decreased to a minimum of 5.5 between feeding events as pH
tations were determined by GCs equipped with FID and thermal was only corrected after feed addition (Fig. 2-B). Thus, a more con-
conductivity detector as previously described (De Groof et al., stant pH control might have improved selecting for C4 during
2020b). Due to a technical problem, composition from the biogas fermentation.
produced in the LH/LO system could not be analysed. Air intrusion In comparison, at longer HRT (HH/LO systems) less C4 (Y C4 of
was corrected by assuming produced biogas contains 0% O2 and N2. 10 ± 3%) and more acetic acid (C2, Y C2 of 3 ± 1%) and n-caproic
Total solids (TS) and volatile solids (VS) were determined acid (C6, Y C6 of 7 ± 3%) were produced (Fig. 1-A). Thus, a more
according to Standard Methods 2540G using fresh samples mixed-acid metabolism was obtained in this system, where C2,
(APHA, 2017). Chemical oxidation demand (COD) was measured C4 and C6 respectively comprised on average 21 ± 2%, 33 ± 10%
using kits (LCK014, LCI400, Hach, Düsseldorf, Germany) before and 21 ± 11% of the total CAs. Ethanol accumulated under all
and after filtration (0.45 mm) for total and soluble COD, operating conditions, as it was present in the influent (Table 2).
respectively. The ethanol concentration averaged 9.7 ± 0.5 gCOD L
1 and
18 ± 5 gCOD L
1 in LH/LO and HH/LO systems, respectively. In
2.5. Microbial community analysis the LH/LO reactors, ethanol was net produced (Y ethanol of
3.4 ± 0.5%). In contrast, in the HH/LO systems ethanol was net
Biomass samples for microbial community analysis were taken consumed (Y ethanol of
6 ± 4%) despite its high concentration. This
and stored at
18 °C. The inoculum was sampled in duplicate at was caused by the higher concentration of ethanol in the influent
the start-up of each duplicate reactor (i.e., n = 4 for each opera- (FW1), i.e., ethanol was net consumed compared to its loading
tional condition), before the first feeding event. To evaluate the rate. For both LH/LO and HH/LO operating conditions, lactic acid
effect of operating conditions on the microbial communities of from the influent (Table 2) was detected in the effluent at the
the reactors, the final two effluent collections were sampled in start but was nearly fully consumed once semi-continuous oper-
each duplicate reactor and sequencing results were pooled ation was established.
together for bioinformatics processing (i.e., n = 4 for each opera- The maximum C6 concentration reached in the effluent of the
tional condition). Microbial samples were send to DNAsense (Aal- HH/LO reactors (13.6 gCOD L
1) was over three times higher than
borg Øst, Denmark) for DNA extraction, 16S rRNA gene amplicon those in the LH/LO reactors (3.4 gCOD L
1). Trace concentrations of
sequencing and bioinformatics processing where reads were clus- heptanoic (C7) and caprylic acid (C8) were detected near the end of
tered by operational taxonomic units (OTUs), taxonomy assigned operation (<0.5 gCOD L
1). The higher concentrations of these
and relative abundance determined, as previously described (De MCCAs in the HH/LO systems is likely from chain elongation of
Groof et al., 2020b). Sequences were deposited with the ENA data- VFAs to MCCAs with ethanol and lactic acid being utilised as elec-
base (accession number PRJEB40478). The results were analysed tron donors (Spirito et al., 2014). Under both operating conditions,
using the DNAsense app (DNAsense, Aalborg Øst, Denmark), which ethanol accumulated to concentrations that promote chain elonga-
is based on the ampvis package (v.2.5.8) in R (v. 3.5.1) (Albertsen tion (Lonkar et al., 2016). Yet, only in HH/LO was ethanol net
et al., 2015; R Core Team, 2017). The number of sequencing reads consumed.
per sample was between 7705 and 27323. Rarefraction curves, To clarify the difference in ethanol consumption and carboxylic
sequences, relative abundances and taxonomic classifications for acid production, the fermentation mechanisms were evaluated in
all samples have been made available as part of a dataset (De cycle studies: the reactor contents were sampled between feeding
Groof et al., 2020a). The OTUs with <0.1% relative abundance in points after operating for 3 HRT (Fig. 2). In the LH/LO systems lac-
any sample were removed and a Non-Metric Multidimensional tate fell below detection limits and ethanol and CAs, mostly C4,
Scaling (NMDS) plot on the Bray-Curtis dissimilarity matrix was increased in the first 13 h after feeding (Fig. 2-B). After, limited
constructed to compare similarity of the microbial community further fermentation was observed. In contrast, when operating
structures in each biomass sample (Bray & Curtis, 1957). Alpha- at a longer HRT in the HH/LO reactors, primary acidogenic
diversity indices were calculated as indicator for biodiversity of fermentation was followed by a consecutive fermentation stage
the microbial communities and converted to effective diversity of chain elongation (Fig. 2-A). Lactic acid and VFAs showed no
based on Hill numbers 1D and 2D (Hurlbert, 1971; Jost, 2006; net change for the first 10 h after feed addition, however, the pH
Lucas et al., 2017). dropped to 5.68 indicating acidogenic fermentation. After 24 h of
83
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

Fig. 1. Average concentration profiles (left) and net yields of liquid fermentation products (right) for three sets of operating conditions: (A) HH/LO at 10.4 ± 1.4 days HRT and
12.7 ± 1.6 gCOD L
1d
1 OLR, (B) LH/LO at 8.5 ± 0.5 days HRT and 11.9 ± 0.6 gCOD L
1d
1 OLR, (C) LH/HO at 8.3 ± 0.4 days HRT and 19.6 ± 1.0 gCOD L
1d
1 OLR. Values are
averaged over duplicate reactors. Overall net yields (right) are determined by averaging net yields at each feeding after in the period from one HRT (indicated by dotted line)
to end of operation.

fermentation, the pH recovered to 5.83 due to lactic acid being To date chain elongation in FW fermentation, without supple-
fully consumed, ethanol decreased slightly, and MCCAs increased mentation of electron donors, has predominantly been attributed
to 18.7 gCOD L
1. This is in line with reports for repeated batch fer- to lactate as the electron donor (Contreras-Davila et al., 2020;
mentation of FW, where initial acidogenic fermentation is followed Nzeteu et al., 2018). However, several studies have reported that
by lactate-based chain elongation where the pH increases addition of ethanol to pre-fermented FW also resulted in chain
(Contreras-Davila et al., 2020). When lactate was depleted, no elongation (Grootscholten et al., 2014; Roghair et al., 2018). Wu
major changes were seen in the concentrations of fermentation et al. found that combining ethanol and lactic acid resulted in a
products over the following two days. No more chain elongation syntrophic interaction improving MCCA production in the fermen-
took place even though ethanol and VFAs were still present. Inhi- tation of Chinese liquor wastewater (Wu et al., 2018). An initial
bition of further fermentation was not likely as the concentration ethanol content around 4 gCOD L
1 initiated chain elongation in
of protonated C6 (maximum HC6 of 0.72 g L
1), the most antimi- batch fermentation of liquid FW (nearly 2 gCOD L
1 C6 was
crobial form, and ethanol remained below reported inhibitory formed) (Coma et al., 2017). Due to the net consumption of ethanol
levels (0.87 g L
1 HC6 and 40 g L
1 for ethanol) (Angenent et al., in the HH/LO reactors, we hypothesize that either chain elongation
2016; Lonkar et al., 2016). It is more likely that the limited avail- occurred with both the electron donors, and/or that ethanol due to
ability of lactate hindered further fermentation. Production of its high concentration in the HH/LO systems could be oxidized to
MCCA could probably be optimised by operating at a higher C2, resulting in a higher Y C2 , with co-production of H2 (Roghair
organic load to stimulate in situ lactic acid production, yet without et al., 2018). The HH/LO reactors produced 0.6 ± 0.1 L L
1d
1 biogas
compromising on retention times, thus by using a feedstock with a containing 22 ± 9% H2 (and 78 ± 9% CO2, no CH4). The increased
higher COD content (De Groof et al., 2020b). level of C2 and H2 production stimulate chain elongation, the

84
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

Fig. 2. Cycle study performed on the semi-continuous reactors after three HRT: the concentration of liquid fermentation products and pH was evaluated at structured time
intervals in between feeding events. Time represents hours after feed addition. Three sets of operating conditions were compared: (A) HH/LO at 10.4 ± 1.4 days HRT and
12.7 ± 1.6 gCOD L
1d
1 OLR, (B) LH/LO at 8.5 ± 0.5 days HRT and 11.9 ± 0.6 gCOD L
1d
1 OLR, (C) LH/HO at 8.3 ± 0.4 days HRT and 19.6 ± 1.0 gCOD L
1d
1 OLR. Values are
averaged over duplicate reactors and error bars present standard deviations (with the exception of the HH/LO system which show only 1 replicate, as the second replicate
became contaminated).

former as a substrate and the latter to ensure a reductive environ- HH/LO systems where no lactic acid was detected in the effluent.
ment preventing MCCA degradation (Angenent et al., 2016). Ethanol accumulated up to 14 ± 5 gCOD L
1, but with Y ethanol near
zero as it was present in the FW2 feed. Longer chain carboxylic
3.2. Lactic acid production at elevated organic loading rate acids, i.e., C5 and C6, were present at start-up of the LH/HO sys-
tems due to their presence in the inoculum, but they were fully
The effect of varying OLR was evaluated by comparing the per- washed-out after two HRT cycles. Biogas was produced at an aver-
formance of the LH/LO systems to operation at higher OLR with the age rate of 0.13 ± 0.09 L L
1d
1, and CO2 comprised over 99%. Fer-
same HRT (the LH/HO reactors). The same substrate, FW2, was mentation studies focusing on synthesis of lactic acid from FW
used and an increased OLR was obtained by not diluting the sub- reached similar outcomes, i.e., 30 to 40 gCOD L
1 lactic acid, but
strate as was done for the LH/LO systems (Table 3). In the LH/HO at lower HRT values of 3 to 5 days (Tang et al., 2016; Wu et al.,
reactors, lactic acid was the dominant product (34 ± 5 gCOD L
1, 2015). Thus, yields for lactic acid might increase by shortening
Y lactate of 7 ± 2%) (Fig. 1-C). This is in contrast with the LH/HO and the HRT below 8.5 days.

85
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

The lactic acid production in the LH/HO reactors was accompa-

nied by a high rate of acidification. NaOH was dosed at an average
of 33 ± 5 mM d
1 to maintain pH, whereas only 11 ± 2 and
15 ± 4 mM d
1 of NaOH were required for the LH/LO and HH/LO
systems, respectively. In LH/HO the pH decreased rapidly in the
first 12 h after feed addition and plateaued at 5.0 (Fig. 2- C). This
was accompanied by minimal change in the concentrations of liq-
uid products, with only C2 and C3 increasing slightly by 2 ± 1 and
1.6 ± 0.7 gCOD L
1 respectively. Rapid acidification of the environ-
ment is a known outcome of the biotic activity of lactic acid bacte-
ria (LAB), which gives them a competitive advantage over
microorganisms with lower acid tolerance (Bachmann et al.,
2017). Tang et al. (2017) found FW to be fermented mostly into lac-
tic acid at pH 5, while VFAs were obtained at pH 6. Because pH was
not corrected during the cycle, and only at the feeding points, the
acidification of the media by LAB likely allowed them to outcom-
pete other fermentation bacteria. This effect did not take place in
the LH/LO or HH/LO reactors as the pH remained above 5.5 during
the entire cycle in each case (Fig. 2).
Gu et al. (2018) noted in their leach bed reactor that a load of
100 g per day of FW gave C4 as the dominant VFA product. How-
ever, when operating at 200 g per day of FW the product spectrum
contained mainly lactic acid. Thus, similarly, at an HRT around
8.5 days and OLR around 12 gCOD L
1 d
1 in the LH/LO system
C4 was produced while at the same HRT but nearly double OLR
(20 gCOD L
1 d
1), acidogenic lactic acid-type fermentation
occurred. However, the OLR was still within a suitable range for
carboxylic acid fermentation (5–50 gCOD L
1 d
1) according to lit-
erature (Arslan et al., 2016). Another study, seemingly in contrast,
reported MCCA production to be stimulated in FW fermentation by
an OLR around 20 gCOD L
1 d
1 because it allows to accumulate
electron donors, i.e., precursors of chain elongation (De Groof
et al., 2020b). However, in that study HRT was almost double
(14 days) than the HRT applied in the LH/HO system (8.5 days). Fig. 3. Microbial community analysis for each set of operating conditions (n = 4, 2
Hence, suggesting that at insufficiently long HRT, an increase in reactor duplicates and two sample duplicates). (A) Non-Metric Multidimensional
Scaling Analysis (NMDS) based on the Bray-Curtis distance measure of 24 samples
OLR will overload the system and result in acidogenic lactic acid
and 177 OTUs. Each point represents the community in a specific sample. The closer
production. To determine if the effects observed are simply due the sample points on the plot, the more similar the communities in the samples are.
to the operational parameter or if these have in fact selected for Arrows indicate the connection from inoculum to corresponding reactor culture for
different microbial communities, a microbial community analysis each set of operating condition. (B) Microbial community composition on the genus
was undertaken. level with relative abundance > 1% and coloured per phylum: red = Proteobacteria;
green = Firmicutes; orange = Bacteroidetes; grey = Actinobacteria. HH/LO operation
at 10.4 ± 1.4 days HRT and 12.7 ± 1.6 gCOD L
1d
1 OLR, LH/LO at 8.5 ± 0.5 days HRT
3.3. Reactor operation selected different lactic and carboxylic acid and 11.9 ± 0.6 gCOD L
1d
1 OLR, and LH/HO at 8.3 ± 0.4 days HRT and 19.6 ± 1.0
bacteria gCOD L
1d
1 OLR. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
Duplicate samples were taken from each replicate reactor
before the first feeding to characterise the inoculum, and at the
end of operation to characterise the microbial communities at of the LH/LO reactors. The change of microbial community compo-
quasi-steady state. The communities from sample duplicates, and sition during storage is in agreement with previous reports on
reactor duplicates were highly similar (Fig. 3-A). In contrast, the preservation of microbial communities performing chain elonga-
microbial communities from each reactor at quasi-steady state tion (Candry et al., 2020). Nevertheless, eventually similar commu-
were very distinct from those of the respective inoculums, suggest- nities evolved in the LH/LO and LH/HO systems and alpha-diversity
ing that each set of operational conditions influenced a shift in the increased again over operation. Thus, while storage of microbial
microbiome. The cultures of the inoculum for the LH/LO reactors communities for research purposes of environmental biotechno-
and of the LH/HO reactors in quasi-steady state were in proximity logical applications remains a challenge, reactor operation eventu-
on the NMDS plot, which is expected since the first was inoculated ally dictated community structure.
from the second. However, despite the HH/LO and LH/LO reactors Generally, less OTUs were detected in the reactor communities
being given the same inoculum, i.e., effluent from an enriched compared to the inocula (Table 4). The alpha-diversity in the reac-
chain elongation reactor fermenting FW, their microbial cultures tors operated at an HRT that was two days longer (HH/LO) was
differed. This could be explained since before inoculation of the double based on the first and second order Hill number (1D = 24
LH/LO system the inoculum had been stored around ten months ± 4, 2D = 12 ± 3) compared to the other systems, i.e., LH/LO and
longer at 4 °C compared to the inoculum used in HH/LO. This LH/HO. Thus, a more diverse community was obtained at longer
resulted in an alpha-diversity for the LH/LO inoculum that was less HRT likely due to slower growing fermenters being able to remain
than half of that for HH/LO based on comparison of Hill numbers in the system.
(Table 4). Interestingly, Rummeliibacillus stabekisii, a strictly aerobic The major dominant phyla were Firmicutes, Actinobacteria and
Firmicute isolated from Antarctic soil (Faira da Mota et al., 2016), Bacteroidetes for all the three operational strategies (Fig. 3-B). The
was found at high relative abundance (45 ± 3%) in the inoculum genera with high relative abundances were mostly obligate or

86
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

Table 4
Biodiversity parameters of the microbial community in the inoculum and in the reactors operated according three sets of conditions (HH/LO at 10.4 ± 1.4 days HRT and 12.7 ± 1.6
gCOD L
1d
1 OLR; LH/LO at 8.5 ± 0.5 days HRT and 11.9 ± 0.6 gCOD L
1d
1 OLR; LH/HO at 8.3 ± 0.4 days HRT and 19.6 ± 1.0 gCOD L
1d
1 OLR). Four biomass samples were
averaged per reactor operation, i.e., two samples from each of duplicate reactors. Samples were rarefied to a depth of 10,000 reads.
1 2
Sample Observed OTUs D D Pielou evenness
Inoculum
HH/LO 268 ± 5 26 ± 3 8.7 ± 0.7 0.53 ± 0.02
LH/LO 151 ± 5 8.5 ± 0.6 4.0 ± 0.4 0.35 ± 0.01
LH/HO 210 ± 12 20.6 ± 0.3 10.1 ± 0.2 0.50 ± 0.00
Operation
HH/LO 201 ± 10 24 ± 4 12 ± 3 0.52 ± 0.01
LH/LO 49 ± 5 11 ± 1 6.5 ± 0.4 0.38 ± 0.02
LH/HO 44 ± 3 8.0 ± 0.7 5.1 ± 0.9 0.34 ± 0.02

facultative anaerobic bacteria, tolerant to mildly acidic environ- instance, in the LH/HO reactors the relative abundance of
ments, with wide-ranging hydrolytic capabilities, and the ability homofermentative Lactobacillus (e.g., OTU_6 at 14 ± 7%) was higher
to produce organic acids such as lactic acid and VFAs from fer- compared to the systems operated at lower OLR (e.g., OTU_6 at
mentable sugars (Kraatz et al., 2011; Krieg et al., 2010; Mattarelli 3 ± 3% in LH/LO and <1% in HH/LO). This aligns with the high lac-
& Biavati, 2012; Parte et al., 2009; Zheng et al., 2020). Various tate yield in the LH/HO system. Another example is the almost
VFA-producing genera were detected, especially in the LH/LO reac- exclusive detection of Secundilactobacillus spp. in the HH/LO reac-
tors, which is in line with their predominant functionality, i.e., C4 tors. Secundilactobacillus spp. generally appear in systems after pri-
production. For example, Prevotella 7 was present in all reactors mary fermenters have depleted hexoses and disaccharides, where
but was most abundant in the LH/LO systems (Fig. 3-B). This spe- they perform secondary fermentation of metabolizing pentoses
cies converts lactic acid to VFA in FW fermentation (Feng et al., to pyruvate. Thus, by operating at an HRT of two days longer, sec-
2018). The VFA-producing Megasphaera and Acidaminococcus were ondary fermentation was stimulated which was reflected in the
almost exclusive to the LH/LO reactors. microbial community.
LAB consistently showed the highest relative abundance values. Caproiciproducens spp. was detected in the HH/LO (5 ± 2%) and
The LAB were assigned predominantly to the genus of Lactobacillus LH/LO (3 ± 2%) systems but not in the LH/HO reactor. The sequence
in the reactors operated at shorter HRT (41 ± 4% in LH/LO and of the OTU assigned to the genus Caproiciproducens with the high-
68.8 ± 0.6% in LH/HO). Lactobacillus spp. are known to be dominant est relative abundance was closely similar to the bacteria
in lactate-type fermentation of FW at low pH (<4.5) (Bonk et al., Ruminococcaceae CPB6, found in several lactic acid-based chain
2017; Feng et al., 2018). The high relative abundance of Lactobacil- elongation systems (98.63% identity match in BLAST) (Liu et al.,
lus spp. in LH/LO, but the absence of lactic acid in the effluent, indi- 2020; Zhu et al., 2017). Caproiciproducens spp. belongs to a distinct
cates C4 likely resulted from lactate being metabolised (Asunis lineage of Clostridium group IV in the family of Ruminococcaceae,
et al., 2019). and can produce ethanol, lactic acid, C2, C4 and C6 via chain elon-
In the microbial communities of the reactors, up to 31 different gation (Flaiz et al., 2020; Kim et al., 2015). It was enriched during
OTUs were classified as Lactobacillus spp., and these had varying cheese whey fermentation for VFA production by increasing SRT
relative abundances according to reactor operating conditions. from 10 to 15 days (Lagoa-Costa et al., 2020). Thus, operating at
The OTUs classified in the genus of Lactobacillus with a relative longer HRT in the HH/LO selected for a higher relative abundance
abundance > 1% were fed into a BLAST search to distinguish them of Caproiciproducens spp. However, this genus was also detected
according their novel genera, as recently proposed by Zheng et al. at lower HRT in LH/LO, thus at the lower HRT it did not fully wash
(2020). For instance, some of the most relatively abundant Lacto- out. Its absence (undetected) in the LH/HO system suggests that at
bacillus OTUs were reclassified into the novel genus of heterofer- higher OLR it was unable to compete in the microbial community
mentative Limosilactobacillus spp, common for food fermentation of acidogenic lactate-fermentation.
where they metabolise carbohydrates into lactate, ethanol and/or Caproiciproducens was not the only genus found that relates to
C2 and CO2 (Table 5). C6 production. Some OTUs were assigned to Clostridium sensu
The reclassification of the Lactobacillus genus by Zheng et al. stricto 12 with a relative abundance >1% only found in the HH/LO
(2020) allowed a better comparison of microbial community and systems. Members of this genus have been found to either produce
function across different operating conditions (Table 5). For C6 via ethanol-based chain elongation or compete with it by

Table 5
Overview of the different OTUs assigned to the genus Lactobacillus with a minimal relative abundance of 1% in at least one of the reactors (HH/LO at 10.4 ± 1.4 days HRT and
12.7 ± 1.6 gCOD L
1d
1 OLR; LH/LO at 8.5 ± 0.5 days HRT and 11.9 ± 0.6 gCOD L
1d
1 OLR; LH/HO at 8.3 ± 0.4 days HRT and 19.6 ± 1.0 gCOD L
1d
1 OLR). The OTU sequences were
run in BLAST, and the hits with highest identity match (%) were used to indicate potential reclassification of these Lactobacillus OTUs into the new classification of 25 genera
proposed by Zheng et al. (2020).

OTU_ID Relative abundance (%) BLAST result Reclassification

HH/LO LH/LO LH/HO
OTU_1 11 ± 6 32 ± 2 36 ± 8 98.63% with various L. mucosae and L. reuteri strains Limosilactobacillus
OTU_6 0.9 ± 0.7 3±3 14 ± 7 98.63% with various L. amylovorus and L. acidophilus strains Lactobacillus
OTU_27 0.1 ± 0.1 0.0 ± 0.0 6±1 98.97% with unclassified homofermentative L. (Acharya et al., 2011) Lactobacillus
OTU_24 0.1 ± 0.2 0.0 ± 0.0 5±2 98.29% with various L. oris, L. reuteri and L. vaginalis strains Limosilactobacillus
OTU_171 0.5 ± 0.2 5±2 1.8 ± 0.6 98.63% with various L. mucosae and L. fermentum strains Limosilactobacillus
OTU_13 4±2 0.0 ± 0.0 0.1 ± 0.2 97.26% with various L. johnsonii strains Lactobacillus
OTU_9 3±2 0.0 ± 0.0 0.0 ± 0.0 98.63% with various strains: L. kimchicus, L. pentosiphilus, L. odoratitofui, L. silage Secundilactobacillus
OTU_35 1.7 ± 0.3 0.3 ± 0.1 0.3 ± 0.2 98.63% with various L. ginsenosidimutans and L. versmoldensis strains Companilactobacillus

87
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

excessive production of C4 (Candry et al., 2020; Liu et al., 2020; Wu Declaration of Competing Interest
et al., 2018). The presence of both genera that contain
chain-elongating species using both lactic acid and ethanol The authors declare that they have no known competing finan-
strengthens the hypothesis that both types of chain elongation cial interests or personal relationships that could have appeared
could take place. to influence the work reported in this paper.
Other prominent LAB included Olsenella, which had the highest
relative abundance (38 ± 4%) outscoring Lactobacillus spp. (22 ± 8%) Acknowledgements
in the HH/LO reactors. Olsenella spp. were also detected in the LH/
LO systems but to a lesser extent (6 ± 1%), whilst in the LH/HO The authors would like to thank Tom Phelps and Wesley Wong
reactors the relative abundance was below 1%. This genus ferments from GENeco for reliable and enthusiastic support of the research
carbohydrates to predominantly lactic acid and is linked with performed in the supply of samples and information. We would
hydrolysis and primary fermentation in chain elongating systems, also like to acknowledge the DNAsense staff for their insightful
usually co-occurring with the chain-elongating Pseudoramibacter guidance and assistance regarding the 16s rRNA gene amplicon
spp. (Kraatz et al., 2011; Lambrecht et al., 2019; Liu et al., 2020). sequencing. We would also like to extend our thanks to our col-
However, the latter were not detected here. leagues at the Environmental Engineering Cluster at the Singapore
The other LAB detected were part of the family of Bifidobacteri- Centre for Life Science and Engineering for some helpful discus-
aceae where the dominant genus depended on which OLR was sions regarding microbial community analysis.
employed. Bifidobacterium spp. had a relative abundance of
7 ± 4% in the LH/LO and HH/LO systems but were <1% when oper- References
ating at higher OLR (LH/HO reactors). In contrast, Aeriscardovia spp.
were the second most abundant in the LH/HO reactors (15.3 ± 0.9%) Acharya, B.K., Pathak, H., Mohana, S., Shouche, Y., Singh, V., Madamwar, D., 2011.
but showed < 1.1% in the reactors operated at lower OLR, i.e., LH/LO Kinetic modelling and microbial community assessment of anaerobic biphasic
fixed film bioreactor treating distillery spent wash. Water Res. 45 (14), 4248–
and HH/LO. They are oxygen and acid tolerant bacteria that pro- 4259.
duce C2 and lactate, and were found before in lactic acid fermenta- Albertsen, M., Karst, S.M., Ziegler, A.S., Kirkegaard, R.H., Nielsen, P.H., 2015. Back to
tion from FW and lactate-based chain elongation reactors Basics-The Influence of DNA Extraction and Primer Choice on Phylogenetic
Analysis of Activated Sludge Communities. PLoS ONE 10 (7), e0132783.
(Lambrecht et al., 2019; Liu et al., 2020; Wu et al., 2015). Thus, Angenent, L.T., Richter, H., Buckel, W., Spirito, C.M., Steinbusch, K.J., Plugge, C.M.,
operating at elevated OLR or HRT selected for different types of Strik, D.P., Grootscholten, T.I., Buisman, C.J., Hamelers, H.V., 2016. Chain
LAB, each linked to different fermentation pathways to lactic acid, Elongation with Reactor Microbiomes: Open-Culture Biotechnology To
Produce Biochemicals. Environ. Sci. Technol. 50 (6), 2796–2810.
C4 or C6 and in accordance with the kinetic profiles and pheno-
APHA. 2017. Standard methods for the examination of water and wastewater, vol.
types observed for each set of reactors. 23.
Arslan, D., Steinbusch, K.J.J., Diels, L., Hamelers, H.V.M., Strik, D.P.B.T.B., Buisman, C.J.
N., De Wever, H., 2016. Selective short-chain carboxylates production: A review
of control mechanisms to direct mixed culture fermentations. Critical Rev.
4. Conclusions Environ. Sci. Technol. 46 (6), 592–634.
Asunis, F., De Gioannis, G., Isipato, M., Muntoni, A., Polettini, A., Pomi, R., Rossi, A.,
This study has demonstrated how separately manipulating only Spiga, D., 2019. Control of fermentation duration and pH to orient biochemicals
and biofuels production from cheese whey. Bioresour. Technol. 289, 121722.
two of the key operational parameters in a semi-continuous STR, Bachmann, Herwig, Molenaar, Douwe, Branco dos Santos, Filipe, Bas, Teusink, 2017.
i.e., the HRT and OLR, is sufficient to stimulate three different path- Experimental evolution and the adjustment of metabolic strategies in lactic acid
ways in FW fermentation. To target n-butyric acid in FW fermen- bacteria. FEMS Microbiol. Rev. 41, S201–S219. https://doi.org/10.1093/
femsre/fux024.
tation: operating at lower OLR (12 gCOD L
1d
1) can avert Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A.,
excessive lactic acid production; and a lower HRT (8.5 d) can limit Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002. The IWA Anaerobic Digestion
secondary fermentation. Targeting lactic acid requires a higher Model No 1 (ADM1). Water Sci. Technol. 45 (10), 65–73.
Bensmann, A., Hanke-Rauschenbach, R., Sundmacher, K., 2013. Reactor
OLR (20 gCOD L
1 d
1) whereby rapid acidification allows configurations for biogas plants – a model based analysis. Chem. Eng. Sci.
homofermentative Lactobacillus spp. to outcompete other fermen- 104, 413–426.
tative bacteria. When the goal is MCCA production, a longer HRT Bhatia, S.K., Yang, Y.H., 2017. Microbial production of volatile fatty acids: current
status and future perspectives. Rev. Environ. Sci. Bio/Technol. 16 (2), 327–345.
(10.5 d) is required for consecutive secondary fermentation by
Bo, Z., Pin-Jing, H., 2014. Performance assessment of two-stage anaerobic digestion
species related to chain elongation. These shifts in product and of kitchen wastes. Environ. Technol. 35 (9–12), 1277–1285.
microbial community were reached even after only operating for Bolaji, I.O., Dionisi, D., 2017. Acidogenic fermentation of vegetable and salad waste
three HRT, highlighting how quickly acidogenic fermentation can for chemicals production: Effect of pH buffer and retention time. J. Environ.
Chem. Eng. 5 (6), 5933–5943.
be steered. It would be valuable to evaluate further how modifying Bonk, F., Bastidas-Oyanedel, J.R., Yousef, A.F., Schmidt, J.E., 2017. Exploring the
HRT or OLR separately affect systems in steady-state operation and selective lactic acid production from food waste in uncontrolled pH mixed
how these effects are maintained. The optimal OLR and HRT will culture fermentations using different reactor configurations. Bioresour.
Technol. 238, 416–424.
need to be adjusted according to reactor type and feedstock. How- Bray, R.J., Curtis, J.T., 1957. An ordination of the upland forest community of
ever, these findings provide an essential contribution to how OLR southern Wisconsin. Ecol. Monogr. 27 (4), 325–349.
and HRT can serve as key parameters in the design of a flexible Candry, P., Huang, S., Carvajal-Arroyo, J.M., Rabaey, K., Ganigue, R., 2020.
Enrichment and characterisation of ethanol chain elongating communities
acidogenic FW fermentation platform. The possibility of having from natural and engineered environments. Sci. Rep. 10 (1), 3682.
an adaptable product range according to market demand by repur- Capson-Tojo, G., Rouez, M., Crest, M., Steyer, J.-P., Delgenès, J.-P., Escudié, R., 2016.
posing existing single stage AD assets will allow integrating this Food waste valorization via anaerobic processes: a review. Rev. Environ. Sci.
Bio/Technol. 15 (3), 499–547.
technology within waste management systems, and hence acceler- Coma, M., Martinez-Hernandez, E., Abeln, F., Raikova, S., Donnelly, J., Arnot, T.C.,
ate a broader range of circular economy outcomes from FW Allen, M.J., Hong, D.D., Chuck, C.J., 2017. Organic waste as a sustainable
valorisation. feedstock for platform chemicals. Faraday Discuss. 202, 175–195.
Contreras-Davila, C.A., Carrion, V.J., Vonk, V.R., Buisman, C.N.J., Strik, D., 2020.
Funding
Consecutive lactate formation and chain elongation to reduce exogenous
This study was supported by the European Union’s Horizon chemicals input in repeated-batch food waste fermentation. Water Res. 169,
2020 research and innovation programme under the Marie 115215.
Skłodowska-Curie grant agreement [H2020-MSCA-CO-FUND De Groof, V., Coma Bech, M., Arnot, T., Leak, D., lanham, A. 2020a. Dataset for
‘‘Selecting fermentation products for food waste valorisation with HRT and OLR
665992]; by the EPSRC [EP/L016354/1]; and Wessex Water and as the key operational parameters”, University of Bath Research Data Archive.
GENeco (Bristol, UK). Bath.

88
V. De Groof, M. Coma, T. Arnot et al. Waste Management 127 (2021) 80–89

De Groof, V., Coma, M., Arnot, T., Leak, D.J., Lanham, A.B., 2019. Medium Chain Lagoa-Costa, B., Kennes, C., Veiga, M.C., 2020. Cheese whey fermentation into
Carboxylic Acids from Complex Organic Feedstocks by Mixed Culture volatile fatty acids in an anaerobic sequencing batch reactor. Bioresour.
Fermentation. Molecules 24 (3). Technol. 308, 123226.
De Groof, V., Coma, M., Arnot, T.C., Leak, D.J., Lanham, A.B., 2020b. Adjusting Organic Lambrecht, J., Cichocki, N., Schattenberg, F., Kleinsteuber, S., Harms, H., Muller, S.,
Load as a Strategy to Direct Single-Stage Food Waste Fermentation from Strauber, H., 2019. Key sub-community dynamics of medium-chain carboxylate
Anaerobic Digestion to Chain Elongation. Processes 8 (11). production. Microb. Cell Fact. 18 (1), 92.
Edwards, J., Othman, M., Crossin, E., Burn, S., 2017. Life cycle inventory and mass- Lim, S.J., Kim, B.J., Jeong, C.M., Choi, J.D., Ahn, Y.H., Chang, H.N., 2008. Anaerobic
balance of municipal food waste management systems: Decision support organic acid production of food waste in once-a-day feeding and drawing-off
methods beyond the waste hierarchy. Waste Manag 69, 577–591. bioreactor. Bioresour. Technol. 99 (16), 7866–7874.
Engelberth, A.S., 2020. Evaluating economic potential of food waste valorization: Liu, B., Kleinsteuber, S., Centler, F., Harms, H., Strauber, H., 2020. Competition
Onward to a diverse feedstock biorefinery. Current Opinion in Green and Between Butyrate Fermenters and Chain-Elongating Bacteria Limits the
Sustainable. Chemistry 26. Efficiency of Medium-Chain Carboxylate Production. Front. Microbiol. 11, 336.
European Commission, 2019. COM/2019/190 final, on the implementation of the Lonkar, S., Fu, Z., Holtzapple, M., 2016. Optimum alcohol concentration for chain
circular economy action plan. Brussels. elongation in mixed-culture fermentation of cellulosic substrate. Biotechnol.
Faira da Mota, F., Vollu, R.E., Jurelevicius, D., Seldin, L., 2016. Whole-Genome Bioeng. 113 (12), 2597–2604.
Sequence of Rummeliibacillus stabekisii Strain PP9 Isolated from Antarctic Soil. Lucas, R., Groeneveld, J., Harms, H., Johst, K., Frank, K., Kleinsteuber, S., 2017. A
Genome Announc 4 (3). critical evaluation of ecological indices for the comparative analysis of
Fdez-Güelfo, L.A., Álvarez-Gallego, C.J., Sales Márquez, D., Romero García, L.I., 2013. microbial communities based on molecular datasets. FEMS Microbiol. Ecol. 93
Destabilization of an anaerobic reactor by wash-out episode: Effect on the (1).
biomethanization performance. Chem. Eng. J. 214, 247–252. Mattarelli, P., Biavati, B. 2012. Genus II. Aeriscardovia Simpson, Ross, Fitzgerald and
Feng, K., Li, H., Zheng, C., 2018. Shifting product spectrum by pH adjustment during Stanton 2004, 405. 2 ed. In: Goodfellow, M., Kampfer, P., Busse, H.-J., Trujillo, M.
long-term continuous anaerobic fermentation of food waste. Bioresour. E., Suzuki, K., Ludwig, W., Whitman, W.B., Bergey’s manualÒ of Systematic
Technol. 270, 180–188. Biology: The Actinobacteria, Part A and B, vol. 5, Springer, pp. 206–207.
Flaiz, M., Baur, T., Brahner, S., Poehlein, A., Daniel, R., Bengelsdorf, F.R., 2020. Moscoviz, R., Trably, E., Bernet, N., Carrère, H., 2018. The environmental biorefinery:
Caproicibacter fermentans gen. nov., sp. nov., a new caproate-producing state-of-the-art on the production of hydrogen and value-added biomolecules
bacterium and emended description of the genus Caproiciproducens. Int. J. in mixed-culture fermentation. Green Chem. 20 (14), 3159–3179.
Syst. Evol. Microbiol. Nguyen, D., Gadhamshetty, V., Nitayavardhana, S., Khanal, S.K., 2015. Automatic
Forgacs, G., law, I., Arnot, T. 2015. Development of a food waste anaerobic digestion process control in anaerobic digestion technology: A critical review. Bioresour.
model using BioWin. In: 14th World Congress on Anaerobic Digestion Viña del Technol. 193, 513–522.
Mar, Chile. Nzeteu, Corine Orline, Trego, Anna Christine, Abram, Florence, O’Flaherty, Vincent,
Forgacs, G., Smyth, M., Law, I., Arnot, T. 2017. Co-digestion and food waste AD: 2018. Reproducible, high-yielding, biological caproate production from food
BioWin modelling to optimise OLR and HRT on the Avonmouth food waste waste using a single-phase anaerobic reactor system. Biotechnol. Biofuels 11
digesters. In: European Biosolids and Organic Resources Conference, Royal (108). https://doi.org/10.1186/s13068-018-1101-4.
Armouries, Leeds, UK. Parte, A.C., de Vos, P., Garrity, G.M., Jones, D., Krieg, N.R., Ludwig, W., Rainey, F.A.,
Gasper, D., Shah, A., Tankha, S., 2019. The Framing of Sustainable Consumption and Schleifer, K.-H., Whitman, W.B., 2009. Bergey’s manualÒ of Systematic
Production in SDG 12. Global Policy 10 (S1), 83–95. Bacteriology: The Firmicutes. Springer.
Girotto, F., Lavagnolo, M.C., Pivato, A., Cossu, R., 2017. Acidogenic fermentation of R Core Team, 2017. R: A Language and Envionment for Statistical Computing. R
the organic fraction of municipal solid waste and cheese whey for bio-plastic Foundation for Statistical Computing, Vienna, Austria.
precursors recovery - Effects of process conditions during batch tests. Waste Roghair, M., Hoogstad, T., Strik, D., Plugge, C.M., Timmers, P.H.A., Weusthuis, R.A.,
Manag 70, 71–80. Bruins, M.E., Buisman, C.J.N., 2018. Controlling Ethanol Use in Chain Elongation
Grootscholten, T.I.M., Strik, D.P.B.T.B., Steinbusch, K.J.J., Buisman, C.J.N., Hamelers, by CO2 Loading Rate. Environ. Sci. Technol. 52 (3), 1496–1505.
H.V.M., 2014. Two-stage medium chain fatty acid (MCFA) production from Spirito, C.M., Richter, H., Rabaey, K., Stams, A.J., Angenent, L.T., 2014. Chain
municipal solid waste and ethanol. Appl. Energy 116, 223–229. https://doi.org/ elongation in anaerobic reactor microbiomes to recover resources from waste.
10.1016/j.apenergy.2013.11.061. Curr. Opin. Biotechnol. 27, 115–122.
Gu, X.Y., Liu, J.Z., Wong, J.W.C., 2018. Control of lactic acid production during Tang, J., Wang, X., Hu, Y., Zhang, Y., Li, Y., 2016. Lactic acid fermentation from food
hydrolysis and acidogenesis of food waste. Bioresour. Technol. 247, 711–715. waste with indigenous microbiota: Effects of pH, temperature and high OLR.
Hurlbert, S.H., 1971. The nonconcept of species diversity: a critique and alternative Waste Manag 52, 278–285.
parameters. Ecology 52, 577–586. Tang, J., Wang, X.C., Hu, Y., Zhang, Y., Li, Y., 2017. Effect of pH on lactic acid
Im, S., Lee, M.-K., Yun, Y.-M., Cho, S.-K., Kim, D.-H., 2020. Effect of storage time and production from acidogenic fermentation of food waste with different types of
temperature on hydrogen fermentation of food waste. Int. J. Hydrogen Energy inocula. Bioresour. Technol. 224, 544–552.
45 (6), 3769–3775. Vanwonterghem, I., Jensen, P.D., Rabaey, K., Tyson, G.W., 2015. Temperature and
Jang, Y.S., Kim, B., Shin, J.H., Choi, Y.J., Choi, S., Song, C.W., Lee, J., Park, H.G., Lee, S.Y., solids retention time control microbial population dynamics and volatile fatty
2012. Bio-based production of C2–C6 platform chemicals. Biotechnol. Bioeng. acid production in replicated anaerobic digesters. Sci. Rep. 5, 8496.
109 (10), 2437–2459. Venkata Mohan, S., Nikhil, G.N., Chiranjeevi, P., Nagendranatha Reddy, C., Rohit, M.
Jankowska, E., Chwialkowska, J., Stodolny, M., Oleskowicz-Popiel, P., 2015. Effect of V., Kumar, A.N., Sarkar, O., 2016. Waste biorefinery models towards sustainable
pH and retention time on volatile fatty acids production during mixed culture circular bioeconomy: Critical review and future perspectives. Bioresour.
fermentation. Bioresour. Technol. 190, 274–280. Technol. 215, 2–12.
Jin,Y.,Lin,Y.,Wang,P.,Jin,R.,Gao,M.,Wang,Q.,Chang,T.C.,Ma,H.,2019.Volatilefattyacids Wang, K., Yin, J., Shen, D., Li, N., 2014. Anaerobic digestion of food waste for volatile
production from saccharification residue from food waste ethanol fermentation: fatty acids (VFAs) production with different types of inoculum: effect of pH.
Effect ofpH andmicrobial community. Bioresour.Technol. 292, 121957. Bioresour. Technol. 161, 395–401.
Jost, L., 2006. Entropy and diversity. Oikos 113 (2), 363–375. Wu, Q., Guo, W., Bao, X., Meng, X., Yin, R., Du, J., Zheng, H., Feng, X., Luo, H., Ren, N.,
Kannengiesser, J., Sakaguchi-Soder, K., Mrukwia, T., Jager, J., Schebek, L., 2016. 2018. Upgrading liquor-making wastewater into medium chain fatty acid:
Extraction of medium chain fatty acids from organic municipal waste and Insights into co-electron donors, key microflora, and energy harvest. Water Res.
subsequent production of bio-based fuels. Waste Manag. 47 (Pt A), 78–83. 145, 650–659.
Khan, M.A., Ngo, H.H., Guo, W.S., Liu, Y., Nghiem, L.D., Hai, F.I., Deng, L.J., Wang, J., Wu, Y., Ma, H., Zheng, M., Wang, K., 2015. Lactic acid production from acidogenic
Wu, Y., 2016. Optimization of process parameters for production of volatile fermentation of fruit and vegetable wastes. Bioresour. Technol. 191, 53–58.
fatty acid, biohydrogen and methane from anaerobic digestion. Bioresour. Zhang, C., Xiao, G., Peng, L., Su, H., Tan, T., 2013. The anaerobic co-digestion of food
Technol. 219, 738–748. waste and cattle manure. Bioresour. Technol. 129, 170–176.
Kibler, K.M., Reinhart, D., Hawkins, C., Motlagh, A.M., Wright, J., 2018. Food waste Zhang, L., Loh, K.C., Dai, Y., Tong, Y.W., 2020. Acidogenic fermentation of food waste
and the food-energy-water nexus: A review of food waste management for production of volatile fatty acids: Bacterial community analysis and semi-
alternatives. Waste Manag 74, 52–62. continuous operation. Waste Manag 109, 75–84.
Kim, B.C., Seung Jeon, B., Kim, S., Kim, H., Um, Y., Sang, B.I., 2015. Caproiciproducens Zhang, R., El-Mashad, H.M., Hartman, K., Wang, F., Liu, G., Choate, C., Gamble, P.,
galactitolivorans gen. nov., sp. nov., a bacterium capable of producing caproic 2007. Characterization of food waste as feedstock for anaerobic digestion.
acid from galactitol, isolated from a wastewater treatment plant. Int. J. Syst. Bioresour. Technol. 98 (4), 929–935.
Evol. Microbiol. 65 (12), 4902–4908. Zheng, J., Wittouck, S., Salvetti, E., Franz, C., Harris, H.M.B., Mattarelli, P., O’Toole, P.
Kim, H., Kim, J., Shin, S.G., Hwang, S., Lee, C., 2016. Continuous fermentation of food W., Pot, B., Vandamme, P., Walter, J., Watanabe, K., Wuyts, S., Felis, G.E., Ganzle,
waste leachate for the production of volatile fatty acids and potential as a M.G., Lebeer, S., 2020. A taxonomic note on the genus Lactobacillus: Description
denitrification carbon source. Bioresour. Technol. 207, 440–445. of 23 novel genera, emended description of the genus Lactobacillus Beijerinck
Kraatz, M., Wallace, R.J., Svensson, L., 2011. Olsenella umbonata sp. nov., a 1901, and union of Lactobacillaceae and Leuconostocaceae. Int. J. Syst. Evol.
microaerotolerant anaerobic lactic acid bacterium from the sheep rumen and Microbiol.
pig jejunum, and emended descriptions of Olsenella, Olsenella uli and Olsenella Zhou, M., Yan, B., Wong, J.W.C., Zhang, Y., 2018. Enhanced volatile fatty acids
profusa. Int. J. Syst. Evol. Microbiol. 61 (Pt 4), 795–803. production from anaerobic fermentation of food waste: A mini-review focusing
Krieg, N.R., Ludwig, W., Euzéby, J., Whitman, W.B., 2010. Phylum XIV. Bacteroidetes on acidogenic metabolic pathways. Bioresour. Technol. 248 (Pt A), 68–78.
phyl. nov. In: Krieg, N.R., Staley, J.T., Brown, D.R., Hedlund, B.P., Paster, B.J., Zhu, X., Zhou, Y., Wang, Y., Wu, T., Li, X., Li, D., Tao, Y., 2017. Production of high-
Ward, N.L., Ludwig, W., Whitman, W.B. (Eds.), Bergey’s ManualÒ of Systematic concentration n-caproic acid from lactate through fermentation using a newly
Bacteriology, Springer. New York. isolated Ruminococcaceae bacterium CPB6. Biotechnol. Biofuels 10, 102.

89