The global economy relies substantially on fossil fuels as a source of carbon compounds with applications ranging from energy to medicine. The pressing need to reduce dependency on these nonrenewable sources has inspired interest in the development of sustainable energy and bioproduct feedstocks. Due to its availability and energy density, lignocellulosic biomass has long been a target for the bioproduction of fuels, bioplastics, and other commodity chemicals [1
]. However, while biological methods for upcycling the cellulosic portion are well developed, a significant challenge to the economic feasibility of using lignocellulose for bioenergy is the chemical complexity of lignin and its recalcitrance to breakdown.
Lignin can compose 15–40% of unprocessed plant matter; initial processing of lignin yields complex mixtures of aromatic compounds, which may vary between types of feedstock [4
]. There is no known single organism capable of catabolizing every compound in this complex mixture, and degradation can result in toxic intermediates that inhibit the growth of some of the very organisms involved. For instance, lignin-derived aromatic compounds are heavily substituted by methoxy (-OCH3
) groups, which are transformed to formaldehyde in the process of microbial degradation [6
]. Chemical pretreatment of lignin requires extreme chemicals or heat and may result in the destruction of important organisms or enzymes. Lignin is, therefore, considered an economically ineffective waste product of the food and other commercial industries. For lignocellulose to compete with nonrenewable carbon feedstocks, efficient, cost-effective processes of transformation must be developed that are robust over time and adaptable to diverse feedstocks.
In nature, lignocellulose is consumed by complex and dynamic communities of microbes, where distinct catabolic niches allow symbiosis by crossfeeding and detoxification of dangerous intermediates. While industries may mimic this “natural” strategy by using complex, undefined microbial communities (as is done in wastewater treatment [12
]), there are long-standing problems with an approach in which the chemical transformations are not fully understood, particularly when it comes to the efficient production of specific desired bioproducts.
To the same ends, with very different means, synthetic and systems biology research frequently attempts to build a single metabolic powerhouse: one well-understood, often genetically engineered species capable of carrying out the entire process (e.g., [13
]). This strategy has the advantage of not requiring the careful balance of growth conditions tailored to multiple species, and it is theoretically possible to maintain a consistent community in a fermentation system. However, when the specific enzymes responsible for particular transformations—as in the case of many lignocellulose components—are unknown, or function poorly in organisms with well-developed systems, complex and multistep processes can be a challenge.
Incorporating the best of both systems, the synthetic ecology approach describes a highly defined and specifically engineered community of organisms [14
]. By narrowing the number of organisms, compared to “natural” undefined communities, it is possible to create conditions where each can thrive, especially if organisms are engineered or evolved for their roles. It is also possible, in contrast to the “powerhouse” strain approach, to select organisms with native affinities and tolerances suited for their role that are complex to engineer, including resistances to heat, acid, toxic intermediates, or the ability to store carbon or biomass in a form that can be used economically in downstream processes.
In this work, we describe a synthetic microbial community designed for the degradation of lignocellulose, with a particular focus on addressing the problem of a toxic compound generated in lignin degradation: formaldehyde. Formaldehyde is a small aliphatic aldehyde that is often overlooked in bioprocessing. Yet it is inhibitory to ethanol-generating yeast at concentrations as low as 1.0 mM, lower than is found in many chemically pretreated lignocellulosic feedstocks, resulting in a reduction in product generation of up to tenfold [18
]; formaldehyde accumulation is often a challenge overcome via engineered resistance [19
]. It can be inhibitory even to the organisms that generate it: formaldehyde detoxification can prove a rate-limiting step in the microbial degradation of methoxylated aromatic compounds [8
]. In this study, we have taken advantage of the single-carbon (C1
) metabolism of the model methylotroph Methylorubrum extorquens
(formerly Methylobacterium extorquens
), which uses formaldehyde as a central metabolic intermediate [21
]. Our aim was to investigate the potential for M. extorquens
, as a member of a defined lignocellulose-degrading community, to increase the efficiency of lignocellulosic breakdown by consuming the inhibitory formaldehyde.
clade encompasses a diverse range of species in a variety of metabolic niches, including commensal relationships with plants as well as independently, in soil and leaf litter [22
]. While the extent of the metabolic capabilities of the genus is not fully characterized—and recent work suggests some species may have the ability to utilize lignin-derived aromatic compounds [23
]—we chose to include in our consortium, M. extorquens
PA1, a model organism for which extensive metabolic and physiological data exist [21
]. The other bacterial members of this defined community included Pseudomonas putida
, a canonical lignin degrader that has been studied extensively for its aromatic catabolism [20
], and Cellulomonas fimi
, a cellulose degrader of interest for its ability to utilize diverse polysaccharides and to channel the products of their degradation to other organisms in co-culture [31
]. In some experiments, a fourth microbial strain, the oleaginous yeast Yarrowia lipolytica
, was included for its ability to grow on organic acids generated by other consortium members, and to produce neutral lipids as a potential end product [33
]. We envision, ultimately, developing a stirred aerobic bioreactor for the transformation of lignocellulose hydrolysate; for this reason, we chose not to work with mycelial fungi or filamentous bacteria.
In place of a complex and undefined plant biomass substrate, we opted for a simple and defined set of compounds to stand in for lignocellulose: cellobiose (a disaccharide of glucose, a product of cellulose hydrolysis), xylose (a 5-carbon sugar found in hemicellulose), and vanillic acid (a simple methoxylated aromatic compound, a derivative of guaiacyl lignin phenylpropanoids, for which formaldehyde generation is the first step in catabolism by P. putida
]). Of these compounds, P. putida
could consume only vanillic acid and C. fimi
only cellobiose and xylose (with cellobiose the preferred substrate [33
]); M. extorquens
and Y. lipolytica
could consume neither and, therefore, subsisted on metabolites generated by P. putida
and C. fimi
. The hypothesized interactions around which this community is built are shown in Figure 1
Our ultimate goal is the development of a metabolically efficient and ecologically robust model microbial lignocellulose-degrading community, in which we can take advantage of recent developments in metabolomic measurement and metabolic modeling to enable a flexible, predictive strategy to maximize community output [36
]. To achieve this, we sought to establish robust and reproducible methods for the culture of this novel community and for measurement of organism growth, activity, and interactions; to characterize the dynamics of formaldehyde during the consumption of lignin-derived aromatic compounds.
2. Materials and Methods
Specifics of the strains and culture conditions used in each experiment are given in Table S1
. Details are provided below; abbreviations used in Table S1
are denoted by square brackets.
: The strains used were Cellulomonas fimi
ATCC 484 [CF], Pseudomonas putida
KT2440 [PP], Yarrowia lipolytica
CLIB122 [YL], and three strains of Methylorubrum extorquens
PA1 [ME]. Strain CM2730 is the wild-type strain, M. extorquens
: the cellulose-synthesis locus was deleted to eliminate clumping and facilitate growth analysis by OD [26
]. Strain CM3745 is M. extorquens
, a mutant with increased tolerance to formaldehyde [37
]. CM3745 was used in early experiments, but as the formaldehyde concentrations in coculture never exceeded the maximum inhibitory concentration (MIC) of wild-type M. extorquens
and we observed no difference in performance between the two strains, later experiments were conducted with CM2730.
Strain CM4744 was a methionine-overproducing strain developed for this study, by taking advantage of the fact that methionine overproduction can confer resistance to the methionine-analog ethionine [38
]. M. extorquens
CM2730 was grown to stationary phase in MP + 15 mM methanol. A total of 200 µL of culture was spread-plated onto each of two plates with MP-methanol agar medium containing 1 mg/mL of ethionine. A total of 8 colonies were chosen and re-streaked onto fresh MP-methanol-ethionine plates. These isolates were then tested for their ability to promote the growth of C. fimi
in the absence of methionine, and all performed equally well. Results shown in this text (Figure 9) are from a single isolate, CM4744.
For all experiments, freezer stocks were streaked onto MPI agar plates with 15 mM succinate or 125 mM methanol (M. extorquens), or onto Nutrient Agar (Difco) (other strains), to obtain colonies. A single colony was used to inoculate 5 mL of liquid culture medium and grown overnight until stationary phase. Unless otherwise noted, pre-growth medium was MP with methionine/thiamine/biotin supplements in standard concentrations and 15 mM methanol (M. extorquens) or 10 mM glucose (other strains). When necessary, this inoculum was subcultured once into a different medium (e.g., Model Lignocellulose) for an additional 24 h of growth to acclimate it for the experiment. After 24 h, stationary-phase cultures of all species were diluted to normalize their ODs to match that of the least-dense culture, then equal volumes of all cultures were inoculated into the experimental medium, at a dilution of 1:64 (vol:vol) into the final medium, unless otherwise stated.
Basal medium and buffer
: The medium used as a basis for all growth experiments was a modified PIPES-buffered medium previously described [26
] [MP]: 30 mM PIPES, 1.45 mM K2
, 1.88 mM NaH2
, 0.5 mM MgCl2
, 5.0 mM (NH4
, 0.02 mM CaCl2
, 45.3 µM Na3
, 1.2 µM ZnSO4
, 1.02 µM MnCl2
, 17.8 µM FeSO4
, 2 µM (NH4
, 1 µM CuSO4
, 2 µM CoCl2
, 0.338 µM Na2
, pH 6.7. As described in the Results and in Table 1
, we conducted some experiments with a variant of MP with lower PIPES or higher phosphate [P] concentrations. “P” refers to the combination of K2
in the same ratio as in the final medium. Most experiments used either 1× (30 mM) or 0.1× (3 mM) PIPES, and either 1× P (~3 mM phosphate) or 5× P (~15 mM phosphate).
Carbon substrates and supplements: Carbon substrates consisted of glucose [G], methanol [MeOH], succinate [S], vanillic acid [VA], protocatechuic acid [PCA], cellobiose [CB], and xylose [XY]. All were stored as sterile aqueous stock solutions and added to cultures by pipet. Because of their poor solubility in water, VA and PCA were stored as 50 mM stock in MP medium so that their addition would not dilute the final culture medium. Whereas most carbon substrates were sterilized by autoclaving, VA and PCA were sterilized by filtration out of an abundance of caution because we found that autoclaving changed their color (brown and purple, respectively).
For the experiments identifying the nutritional needs of C. fimi
), all amino acid stocks were made in MP medium at 10 g/L (10 mL vol), except asparagine (6.67 g/L), aspartic acid (2 g/L, NaOH added to bring pH to 6.5), and tyrosine (50 g/L in DMSO), due to solubility. For all stocks, regardless of concentration, 100 µL of stock added to 10 mL culture medium, with the exception of the tyrosine DMSO stock, of which only 10 µL was added. Wolfe’s vitamins made according to [40
] and provided at 1×. Yeast extract (VWR) was provided at a final concentration of 0.3 g/L and tryptone (Peptone from Casein, Sigma Aldrich) at 0.16 g/L.
In Model Lignocelluose medium, methionine [M], thiamine [T], and biotin [B] aqueous stocks were provided at concentrations of 2 mg/L, 5 µg/L, and 40 µg/L, respectively (standard concentrations), unless otherwise noted. Stocks were made in water, filter-sterilized, stored at 4 °C, and added to medium prior to each experiment. For experiments testing vanillic acid toxicity, glucose was provided for C. fimi and methanol for M. extorquens as additional carbon substrates as they cannot grow on vanillic acid, but no additional carbon substrate was provided for P. putida. For experiments testing the effect of iron concentration, 17 mM (1000×) aqueous, autoclave-sterile FeSO4 stock was added.
For use in enumerating colony-forming units of the different species from cocultures, standard MP medium was prepared with 1.8 g/L glucose and 4.05 g/L sodium succinate dibasic hexahydrate, methionine/thiamine/biotin in standard concentrations, and 15 g/L agar.
Formaldehyde was produced fresh weekly as 1 M stock by combining 0.3 g paraformaldehyde powder (Sigma Aldrich, St. Louis, MO, USA), 9.95 mL ultrapure water, and 50 µL 10 N NaOH solution in a sealed tube and immersing in a boiling water bath for 20 min to depolymerize. The stock was stored at room temperature and removed from the sealed tube using a syringe when needed.
Vessels and incubation conditions: All cultures were grown at 30 °C, either in culture flasks, culture tubes, multiwell culture plates, or culture plates with solid medium [agar]. For multiwell culture plates [multiwell], we used 48-well tissue culture plates (Corning Costar, Tewksbury, MA, USA) with a total volume of 640 µL per well, and incubated in a LPX44 Plate Hotel (LiCONiC, Mauren, Liechtenstein) with shaking at 650 RPM. For glass tubes, we originally used Balch tubes with serum stoppers [Balch] (Chemglass, Vineland, NJ, USA) in order to ensure that volatile compounds such as formaldehyde were not lost in the gas phase; however, experiments demonstrated no difference in any measured compounds between Balch tubes and simple 16 × 150 mm glass culture tubes with loose-fitting lids [tube], so we ultimately used aerobic tubes for convenience. All culture tubes contained 5 mL of liquid medium unless otherwise stated and were incubated with shaking at 250 rpm. For culture flasks [flask], we used 50 mL-capacity glass Erlenmeyer flasks, containing 10 mL of liquid culture, also shaken at 250 rpm.
: In growth experiments conducted in tubes or flasks with multiple timepoints, a typical sampling procedure was as follows: 100 µL of culture was removed from the vessel, by syringe and needle through the stopper for Balch tubes or by pipet otherwise, transferred into a microcentrifuge tube, and centrifuged for 1 min at 14,000× g
to pellet the cells. 60 µL of supernatant was used for formaldehyde measurement and 20 µL for GC-MS or HPLC analysis. The cell pellet was reserved for species-specific analysis of cell abundance as colony-forming units [CFU]. Not all analyses were carried out for all timepoints, but the sampling procedure nonetheless remained the same. Abbreviations in Table S1
are as follows: optical density at 600 nm [OD]; formaldehyde [F]; cellobiose by HPLC [HPLC]; vanillic acid or protocatechuic acid by GC-MS [GC-MS].
For CFU measurement, cell pellets were resuspended in 980 µL of MP medium without carbon substrate (1:10 dilution) and then subjected to serial 1:10 dilutions down to 10−6
(7 dilutions total). These dilutions were either spread-plated or spot-plated onto solid culture medium. For spot-plates, three replicates spots of 10 µL of each dilution were pipetted onto plates and spots were dried under a laminar flow hood, then incubated at 30 °C for 4 days or until colonies were visible. Species were identified by colony morphology (Figure S1
). The colonies in each series of 7 dilution spots was counted; the number of colonies in the two spots of the highest dilution levels that had countable colonies were summed, then multiplied by 1.1 times the lower of the two dilution factors to calculate the number of colony-forming units (CFU) per mL in the original undiluted sample. The mean and standard deviation were calculated for the three replicate spot series representing each sample. For spread-plates, between 1 and 3 dilutions were chosen for plating based on predicted cell abundance; from each dilution, 100 µL of the dilution was spread onto a culture plate. Plates were dried and incubated as for spot-plates. CFU/mL was calculated by multiplying by the dilution factor and accounting for the volume plated.
Measurement of optical density at 600 nm (OD600) for cultures in glass tubes was carried out nondestructively by reading the whole tube with a Spectronic 200 spectrophotometer (Thermo Fisher, Waltham, MA, USA). For cultures in flasks, a 100 µL sample was transferred to a trUVue low-volume cuvette (Bio-Rad, Hercules, CA, USA) and read in a SmartSpec Plus spectrophotometer (Bio-Rad). For experiments in multiwell plates, optical density was assessed using a Wallac 1420 Victor2 Microplate Reader (Perkin Elmer, Waltham, MA, USA), reading OD600 for 0.4 s. In experiments involving different carbon substrates, blank wells were included for each medium composition for blanking purposes, as solutions containing PCA and VA were purple and brown, respectively.
Formaldehyde was measured using the method of Nash [41
]. Reagent B was prepared as described (2 M ammonium acetate, 50 mM glacial acetic acid, 20 mM acetylacetone); for each assay, equal volumes of sample (or standard) and Reagent B were combined in a microcentrifuge tube and incubated for 6 min at 60 °C. Absorbance was read on a spectrophotometer at 412 nm and formaldehyde concentration calculated using a standard curve made from freshly prepared formaldehyde stock. To assay large numbers of samples, a 96-well polystyrene flat-bottom culture plate (Olympus Plastics, San Diego, CA, USA) was used, with a total volume of 200 µL per well and incubation time of 10 min before absorbance at 432 nm was read using a Wallac 1420 Victor2 plate reader. A clean plate was used for each assay; each plate contained each sample in triplicate as well as a standard curve run in triplicate. The absorbance of vanillic acid was not found to interfere with formaldehyde measurements.
Vanillic acid was measured by gas chromatography-mass spectrometry (GC-MS) using an extraction and derivatization procedure modified from [42
]. 20 µL of culture supernatant was combined with 1.2 µL of 1 M HCl to acidify to pH ~ 2. The sample was combined with 100 µL of a 1:100 mixture of 2-chlorobenzoic acid:ethyl acetate, and vortexed to extract the vanillic acid into the organic phase. The sample was centrifuged at 14,000× g
for 1 min to separate the phases, and 80 µL of the organic phase was transferred to a clean GCMS sample vial (1 mL capacity). Samples were dried in a fume hood, then 400 µL of derivatization reagent was added. The derivatization reagent consisted of a 99:1:1000 mixture of N,O-Bistrifluoroacetamide:Trimethylsylil chloride:acetonitrile (that is, BSTFA-TMCS (TCI America, Portland, OR, USA) diluted 1:10 in acetonitrile). The sample was incubated, sealed, at 70 °C for 30 min, then cooled. Samples were analyzed on a Shimadzu GCMS-QP2010 Plus with a 30 m × 0.25 mm dimethyl polysiloxane column (Rxi-1ms, Restek, Bellefonte, PA, USA) in splitless mode with a 1-min injection at 280 °C. The GC run program was as follows: hold at 80 °C for 1 min; ramp to 110 °C at 20 degrees/min; ramp to 240 °C at 10 degrees/min; ramp to 280 °C at 40 degrees/min; hold at 280 °C for 5 min. The MS was run on SIM mode. Vanillic acid was detected as as 3-methoxy-4-[(trimethylsilyl)oxy]-benzoic acid trimethylsilyl ester, with retention time 10.4 min. Characteristic fragments used for quantitation were at m/z
267 and 297, with other fragments at 126, 193, 253, 312. Standard curves were generated from vanillic acid stocks made in lab and extracted alongside the samples.
Cellobiose and xylose were measured using a Shimadzu LC-20 high-performance liquid chromatograph (HPLC). Supernatant samples were filtered through a 0.2 µm syringe filter to remove any particles, and diluted in water if necessary to maintain signal within quantifiable range. They were run on an Aminex HPX-87h column (Bio-Rad) at a flow rate of 0.6 mL/min, column temperature 30 °C, with 5 mM H2SO4 as eluent. Peaks were detected using a RID-20A refractive index detector: cellobiose with a retention time of 7.1 min and xylose at 9.3 min.
Formaldehyde tolerance distributions
: The distribution of formaldehyde tolerance phenotypes within a population was assessed by counting colony-forming units on agar medium containing formaldehyde, as described previously [43
]. MP medium was prepared with the necessary carbon substrates and supplements for each species; after autoclaving, the medium was cooled to 50 °C and formaldehyde was rapidly mixed in. Agar was poured into 100 mm culture plates, dish lids were replaced, and plates were allowed to cool on the benchtop. Plates were stored at 4 °C for no longer than 1 week. Cultures were grown to stationary phase on MP medium with a preferred carbon source (methanol for M. extorquens
; glucose for other species), then CFU were spot-plated as described above onto a series of plates containing a range of formaldehyde concentrations, and the number of cells capable of forming colonies at each concentration of formaldehyde was calculated. Note that an abundance of 34 CFU/mL is necessary to observe 1 cell per 30 µL plated, so for a population of ~2 × 108
CFU/mL (as was typical for M. extorquens
samples), this method has a limit of detection of 1.65 × 10−7
Spent medium experiment
: To generate P. putida
spent medium, P. putida
was grown on Model Lignocellulose medium (Table 1
) to stationary phase. The culture was then centrifuged and the supernatant filtered through a 0.2 µm filter to remove cells. Vanillic acid was added again to a final concentration of 4 mM, to replenish the vanillic acid consumed by P. putida
. This was then used as the growth medium for C. fimi
Data analysis and visualization
: Original data are available as spreadsheets in Supplemental Data File 1
. All data were analyzed using R v. 4.0.2 in Rstudio v.1.3.959. Growth rates (r
) were calculated by fitting the exponential portion of the growth curve to the model N
) = N0ert
. Lag time was calculated as the intersection of the fitted growth curve with OD = 0.0126 (the threshold of detection in multiwell plates). The relationships between lag time and substrate concentration, or between growth rate and concentration, were calculated by fitting a linear relationship using the lm package in R. Inkscape v. 1.0 was used to generate the conceptual figures (Figure 1
and Figure 10) and for customizing layout and annotations on other figures.
We have established a system for studying microbial lignocellulose degradation that uses tractable, well-characterized microbial species in ecological roles that take advantage of their evolved metabolic capabilities, and have developed a simple defined Model Lignocellulose medium with which to investigate interspecies interactions. This groundwork will enable future investigations involving metabolomic analysis and modeling for a more complete understanding of metabolic exchange within the community, and further work using engineering and laboratory evolution to optimize the efficiency and yield of the biomass transformation. While much remains to be learned about the mechanisms of the dynamics observed in this microbial community, the methods and initial findings described here comprise the first step toward using this promising model consortium for studying the microbial transformation of lignocellulosic biomass.
A valuable lesson learned from this study relates to the prevalence of unexpected ecological interactions that may be discovered even in a simple three-species community (Figure 10
). Dynamics we discovered that were not originally predicted included the dramatic interspecies differences in nutritional needs, individual growth rates, pH sensitivity, and tolerance to toxins such as formaldehyde and vanillic acid. Some of these differences may provide clues about the organisms’ ecological niches. For example, formaldehyde tolerance shows a link to species’ metabolic capabilities: in M. extorquens
, a methylotroph, formaldehyde is detoxified by the dephospho-tetrahydromethanopterin pathway [51
], and vanillic-acid-consuming P. putida
carries a set of redundant glutathione-dependent formaldehyde dehydrogenases [30
]. However, in C. fimi
, to our knowledge, formaldehyde detoxification has not been characterized. Another complication in our understanding of the community’s dynamics was the consumption by P. putida
of compounds (iron and methionine) seemingly in excess of what was limiting for its growth. Critically, very little about these findings would be captured by genome-scale metabolic modeling, a method often based on the assumption that each member achieves optimal metabolism on the available substrates [53
]. Our results point to the importance of taking into account traits outside of metabolism for their roles in interspecies interactions. As many model experimental systems in microbial ecology focus primarily on metabolite exchange and chemical warfare (e.g., [38
]), our microbial consortium could prove a useful model for the study of alternative modes of ecological interactions.
The discovery of C. fimi
’s dependence on methionine for growth provided a fortuitous opportunity to enrich the existing interspecies interactions though the development of a methionine-excreting strain of M. extorquens
). This is especially promising, as we have observed C. fimi
already has the capacity to promote M. extorquens
growth (Figure 7
), likely through the generation of carbon substrates. Prior work using exometabolomic analysis of C. fimi
growing on cellulose and galactomannan has shown it to produce several organic acids (e.g., alpha-ketoglutaric acid, 3-hydroxypropionic acid, D-malic acid, citric acid, and amino acids) that are among those known to support the growth of M. extorquens
]. The exchange of methionine for carbon substrates by cross-feeding is, in fact, the basis of another well-characterized model microbial consortium, which was developed through laboratory evolution and has been studied extensively as an example of a stable bidirectional costly mutualism [38
]. It is likely that the growth promotion observed here could in the future be similarly developed through laboratory evolution into a true codependence.
A central goal of our work was to address the issue of formaldehyde production during lignin degradation and to test the hypothesis that a methylotrophic organism could improve the efficiency of lignocellulose degradation by removing a toxic burden. We made significant contributions in that area, by documenting formaldehyde accumulation due to P. putida
growth on vanillic acid over time and the effect of formaldehyde consumption by M. extorquens
. However, much remains still to be investigated. Because formaldehyde concentrations in the consortium are dynamic, their effect may be dramatically different in different growth conditions. Formaldehyde-induced cell damage occurs over time [43
], and as formaldehyde remains in the medium as long as P. putida
is actively consuming vanillic acid, a continuous-culture growth environment might induce a more substantial effect of formaldehyde on C. fimi
viability. Yet it was clear in the present study that other interspecies interactions had a stronger effect on the community than formaldehyde, and those interactions must be resolved first before the role of formaldehyde can be accurately explored and quantified.