Functional Degeneracy in Paracoccus denitrificans Pd1222 Is Coordinated via RamB, Which Links Expression of the Glyoxylate Cycle to Activity of the Ethylmalonyl-CoA Pathway

ABSTRACT Metabolic degeneracy describes the phenomenon that cells can use one substrate through different metabolic routes, while metabolic plasticity, refers to the ability of an organism to dynamically rewire its metabolism in response to changing physiological needs. A prime example for both phenomena is the dynamic switch between two alternative and seemingly degenerate acetyl-CoA assimilation routes in the alphaproteobacterium Paracoccus denitrificans Pd1222: the ethylmalonyl-CoA pathway (EMCP) and the glyoxylate cycle (GC). The EMCP and the GC each tightly control the balance between catabolism and anabolism by shifting flux away from the oxidation of acetyl-CoA in the tricarboxylic acid (TCA) cycle toward biomass formation. However, the simultaneous presence of both the EMCP and GC in P. denitrificans Pd1222 raises the question of how this apparent functional degeneracy is globally coordinated during growth. Here, we show that RamB, a transcription factor of the ScfR family, controls expression of the GC in P. denitrificans Pd1222. Combining genetic, molecular biological and biochemical approaches, we identify the binding motif of RamB and demonstrate that CoA-thioester intermediates of the EMCP directly bind to the protein. Overall, our study shows that the EMCP and the GC are metabolically and genetically linked with each other, demonstrating a thus far undescribed bacterial strategy to achieve metabolic plasticity, in which one seemingly degenerate metabolic pathway directly drives expression of the other. IMPORTANCE Carbon metabolism provides organisms with energy and building blocks for cellular functions and growth. The tight regulation between degradation and assimilation of carbon substrates is central for optimal growth. Understanding the underlying mechanisms of metabolic control in bacteria is of importance for applications in health (e.g., targeting of metabolic pathways with new antibiotics, development of resistances) and biotechnology (e.g., metabolic engineering, introduction of new-to-nature pathways). In this study, we use the alphaproteobacterium P. denitrificans as model organism to study functional degeneracy, a well-known phenomenon of bacteria to use the same carbon source through two different (competing) metabolic routes. We demonstrate that two seemingly degenerate central carbon metabolic pathways are metabolically and genetically linked with each other, which allows the organism to control the switch between them in a coordinated manner during growth. Our study elucidates the molecular basis of metabolic plasticity in central carbon metabolism, which improves our understanding of how bacterial metabolism is able to partition fluxes between anabolism and catabolism.

of metabolic plasticity in central carbon metabolism, which improves our understanding of how bacterial metabolism is able to partition fluxes between anabolism and catabolism. KEYWORDS Paracoccus denitrificans, anaplerosis, carbon metabolism, ethylmalonyl-CoA pathway, glyoxylate cycle, metabolic regulation, proteobacteria, transcriptional regulation A cetyl-CoA is an essential intermediate in central carbon metabolism, where it serves as energy and/or carbon source. While acetyl-CoA is oxidized through the tricarboxylic acid (TCA) cycle for NADH and ATP generation, bacteria have evolved different strategies for the assimilation of acetyl-CoA into biomass. Several acetyl-CoA assimilation routes have been described to date (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12) among them the glyoxylate cycle (2,3) and the ethylmalonyl-CoA pathway (4)(5)(6)(7)12). The glyoxylate cycle (GC) consists of only two enzymes, isocitrate lyase (Icl) and malate synthase (Ms), which branch off the TCA cycle and are often organized in one operon under joint transcriptional regulation (13). The ethylmalonyl-CoA pathway (EMCP), instead, is a rather complex pathway that involves at least 12 enzymes (4), among them crotonyl-CoA carboxylase/ reductase (Ccr). Ccr catalyzes a unique reaction, the reductive carboxylation of crotonyl-CoA into ethylmalonyl-CoA, and is considered the key enzyme of the EMCP (4,6). In contrast to the GC, genes encoding the EMCP are typically distributed across the host genome (14).
Despite their biochemical and genetic differences, the EMCP and the GC seem to serve a functionally degenerate purpose; both pathways use three or two molecules of acetyl-CoA, respectively, and bypass the decarboxylation steps of the TCA cycle to allow the cell to form the TCA cycle intermediates succinate and malate from acetyl-CoA (2-5, 7, 12; Fig. 1B; Table 1). This shifts the function of the TCA cycle from acetyl-CoA oxidation (i.e., energy generation) toward acetyl-CoA assimilation (i.e., biomass formation). For optimal growth, bacteria require a tight and fine-tuned regulation of flux between acetyl-CoA oxidation and assimilation (15).
Among various bacteria, different regulatory mechanisms have been described that regulate flux distribution between acetyl-CoA oxidation and assimilation by controlling the enzymes of the TCA cycle as well as the key enzymes of the GC or EMCP, respectively, at the gene and/or protein level (16)(17)(18)(19)(20). However, these studies primarily focused on organisms that possess only one of the two alternative acetyl-CoA assimilation pathways, mostly the GC (summarized in 13). Notably, several organisms exist that harbor both the GC and the EMCP (14,(21)(22)(23).
These organisms are not only faced with the challenge to distribute metabolic flux between acetyl-CoA oxidation and assimilation, but also with the question of how to coordinate two seemingly degenerate metabolic pathways without compromising fitness.
While "functional degeneracy" describes the phenomenon that cells can use and metabolize the same substrate through different pathways (1), "metabolic plasticity" refers to the ability of a cell to dynamically rewire metabolic routes in response to changing physiological needs (24). The latter is well known from cancer cells. There, metabolic plasticity mediates survival and metastatic outgrowth by facilitating the rapid adaptation to changing conditions, which enables cancer cells to outcompete their neighbors by ensuring optimal nutrient supply (25)(26)(27)(28)(29). Several examples of functional degeneracy and metabolic plasticity have been observed in microbial communities (30)(31)(32)(33) as well as individual bacterial populations. One example is the presence of two functionally degenerate routes for the oxidation of methylamine in Methylobacterium extorquens AM1 (34,35) or the simultaneous existence of the EMCP and GC in several alphaproteobacteria such as P. denitrificans Pd1222 (22) or Rhodobacter capsulatus (23). Yet, the molecular mechanisms coordinating functional degeneracy and metabolic plasticity in bacteria remain largely unknown.
Here, we study the regulation of functional degeneracy during acetyl-CoA metabolism in P. denitrificans Pd1222. We show that acetyl-CoA metabolism follows a dedicated pattern in which the EMCP and GC are sequentially upregulated during a switch to acetate as the sole carbon source for growth. Using a combination of genetic, molecular biological, and biochemical approaches, we identified the transcriptional regulator controlling the GC in P. denitrificans Pd1222, characterized its DNA-binding site and identified CoA esters as small-molecule ligands that induce a change in the oligomerization state of the protein. Notably, these CoA esters are direct intermediates of the EMCP, implying that the expression of the GC is directly linked to an active EMCP. Altogether, our study demonstrates an as of yet unknown regulatory mechanism in central carbon metabolism, in which one seemingly degenerate pathway drives the expression of another.

RESULTS
The ScfR family of transcription factors are potential regulators of acetate metabolism in P. denitrificans Pd1222. To study the regulation of the GC and the EMCP during a switch from succinate to acetate as the sole carbon source for growth, we used the reporter strain P. denitrificans Pd1222 ccr::ccr-mCherry icl::icl-cerulean, in which the key enzymes of the EMCP (Ccr) and the GC (Icl) were fused to distinct fluorescent proteins. During growth on succinate, neither Ccr-mCherry nor Icl-Cerulean fluorescence ccr::ccr-mCherry icl::icl-cerulean (TJE-KK12) on succinate and acetate. Growth is given as OD 600 on the left y axis of each graph. Fluorescence normalized to OD 600 is given on the right y axis of each graph. Replicates are shown as individual curves. (B) Assumed assimilation routes for succinate (left) and acetate (right) in Pd1222. Succinate is assimilated via the TCA cycle. The regeneration of acetyl-CoA from succinate could occur by decarboxylation of malate to pyruvate and subsequent oxidative decarboxylation by pyruvate dehydrogenase (Pdh). The assimilation of acetate requires the activity of the EMCP and GC. (C) Genetic neighborhood of scfR homologs as well as the genes encoding the key enzymes of the GC and EMCP. Ch1, chromosome 1; Ch2, chromosome 2. Genes of ScfR homologs (light orange), the MCC (dark orange), the GC (cyan), and the EMCP (red) are highlighted in color. Gene names are given as Pd1222 gene tags (Pden_). For a detailed description of the individual genes of interest and their products, see Table 1. was detected (Fig. 1A). Upon a switch to acetate, the Ccr-mCherry signal increased first followed by the Icl-Cerulean signal. This sequential activation of key enzymes of the EMCP (Ccr) and the GC (Icl) is in line with earlier reports (22) that additionally showed very basal levels of Ccr (and no Icl) activity in cell extracts of succinate-grown cells and subsequent induction of both enzymes on acetate ( Fig. 1A and B).
We next aimed to identify potential regulators of acetate metabolism in P. denitrificans Pd1222, henceforth termed Pd1222. Notably, several well-known homologs of transcriptional (e.g., IclR (18)) and posttranslational (e.g., AceK) (16,17) regulators of the GC are absent in Pd1222, indicating that the regulation of this pathway relies on other mechanisms. The family of short-chain fatty acid regulators (ScfR) drew our attention, as several of these proteins are involved in the regulation of central carbon metabolic pathways, in particular acetate and/or propionate assimilation (36). Three known ScfR family members are encoded in the genome of Pd1222 (36), namely, Pden_4472, Pden_1365, and Pden_1350. Based on sequence similarity and gene neighborhood analysis, Pden_1350 is suggested as regulator of the methylcitrate cycle (MccR) and Pden_1365 as regulator of acetate metabolism (RamB), while no specific function could be predicted for Pden_4472 due to its genomic isolation (Fig. 1C) (36).
We used the enzyme similarity tools EFI-EST (37) and Cytoscape (38) to construct a sequence similarity network for ScfR family members in Pd1222 (Fig. S1). In line with previous predictions (36), Pden_1350 clustered together with PrpR, a recently characterized MccR homolog from Mycobacterium tuberculosis (39,40), while Pden_1365 clustered in a node of RamB-like proteins that belong to subclass 2 and are distinct from the archetypal Corynebacterium glutamicum RamB that defines subclass 1 (36,41). Pden_4472 was located in a node together with the regulator of propionyl-CoA carboxylase (PccR) that has recently been characterized as transcriptional activator of the methylmalonyl-CoA pathway (MMCP) in Rhodobacter sphaeroides (Fig. S1) (36). Notably, the MMCP is part of the canonical EMCP where it converts propionyl-CoA into succinyl-CoA. Two additional proteins of Pd1222 appeared in our ScfR network, namely, Pden_1785 and Pden_2985. The corresponding genes clustered together with genes for a two-component system and an isocitrate lyase/phosphoenolpyruvate mutase family protein, respectively. Based on the sequence similarity network analysis, we focused on Pden_1350, Pden_1365 and Pden_4472 as potential regulators of acetate metabolism in the following. Pden_1365 is a RamB homolog that regulates acetate metabolism in Pd1222. To test whether one or several of the ScfRs identified were involved in the regulation of the EMCP or GC, we studied the function of Pden_1350, Pden_1365, and Pden_4472 in Pd1222 in vivo. To this end, we deleted each of these genes individually in Pd1222 reporter strains with ccr-mCherry or icl-mCherry/icl-cerulean fusions. Neither the deletion of Pden_4472 nor that of Pden_1350 affected growth or pathway expression patterns in any of the reporter strains under any condition tested ( Fig. S2A and B), which excludes these genes as regulators of the GC or the upper part of the EMCP in Pd1222. In contrast, and in line with its proposed function, the deletion of Pden_1365 abolished the acetate-triggered production of Icl-mCherry in the icl::icl-mCherry reporter background and reduced the growth rate on acetate by 50% compared to the wild-type (WT) strain (m icl::icl-mCherry = 0.4 h 21 versus m icl::icl-mCherryDPden_1365 = 0.2 h 21 ) ( Fig. 2A, column  2). This phenotype could be complemented by expression of flag-Pden_1365 from an inducible promoter on plasmid pIND4 (42) ( Fig. 2A, column 3), while the empty pIND4 had no effect on growth or fluorescence of the Pden_1365 knockout reporter strain ( Fig. 2A, column 4). Note that plasmid pIND4 is known to show leaky expression in P. denitrificans Pd1222 (42). Accordingly, the DPden_1365 growth and expression phenotypes could also be complemented by pIND4_flag-Pden_1365 in the absence of inducer ( Fig. 2A, column 3, row 3).
Deletion of Pden_1365 in the WT background confirmed these results. Similar to the icl:: icl-mCherry DPden_1365 strain, the Pden_1365 deletion in the WT reduced the growth rate on acetate 2-fold (m WT = 0.4 h 21 versus m DPden_1365 = 0.2 h 21 ) (Fig. 2B). Notably, the growth phenotype of the DPden_1365 mutant resembled that of a Dicl mutant (22) (m Dicl = 0.2 h 21 ) (Fig.  2B), suggesting that the growth defect caused by deletion of Pden_1365 could be caused by a lack of Icl in the cell. Together, these results verified the function of Pden_1365 as regulator of acetate metabolism in Pd1222 and suggested that the protein serves as transcriptional activator for the GC. Therefore, we will refer to Pden_1365 as RamB Pd in the following.
RamB Pd binds to a conserved motif upstream of the GC operon. Next, we aimed at identifying the promoter region of GC genes that is targeted by the transcription factor RamB Pd . The gene encoding RamB Pd (Pden_1365) is located in reverse orientation on chromosome 1, directly upstream of the genes for malate synthase (Pden_1364; aceB, in the following referred to as ms) and icl (Pden_1363) (Fig. 1C, row 2). We assessed the upstream region of ms as well as the 120 bp intergenic region between ms and icl in more detail. To that end, we transformed Pd1222 with fluorescence reporter plasmids carrying either the upstream region of ms or the ms-icl intergenic region fused to a downstream mCherry reporter gene (Fig. S3A). In these experiments during growth on acetate, we detected fluorescent expression for the construct containing the ms upstream region, whereas no signal was observed for the ms-icl intergenic region or a promoterless mCherry gene. The fluorescence pattern obtained with the ms upstream region (Fig. S3D) was comparable to the signal observed for the Pd1222 icl::icl-mCherry strain (see above). However, when we analyzed the reporter construct containing the ms upstream region in the DramB background (Fig. S3D), we did not observe any fluorescence, supporting the hypothesis that RamB Pd functions as a transcriptional activator. In summary, these results confirmed the role of RamB Pd as the central activator that acts by binding upstream of the ms gene.
RamB Pd is a class 2 RamB homolog (36) (Fig. S1). While specific binding motifs are known (or predicted) for class 1 RamB homologs and other classes of ScfR-type transcription factors, no consensus motif has been described for class 2 RamB homologs to date (36). To identify such a binding motif, we searched for genomes that contain class 2 RamB homologs clustering with a putative GC operon. We identified 67 bacterial genomes that fulfilled this requirement and analyzed the sequences upstream of the GC operon with MEME (Multiple Em for Motive election) (43) to determine a putative class 2 RamB binding motif. As is common for ScfR binding sites, this binding motif consists of two individual binding boxes that are nearly identical and separated from each other by an 18 bp spacer (Fig. 3A). In the Pd1222 genome, this motif is located 82 bp upstream of the ms gene. We mutated either one or both binding boxes in a reporter plasmid containing the ramB-ms intergenic region fused to an mCherry reporter gene (Fig. 3A). When introduced into the WT strain, only the reporter construct in which both binding boxes were intact yielded an mCherry signal during growth on acetate, while mCherry expression was abolished for constructs where one or both binding boxes were mutated (Fig. 3B). In the Pd1222 DramB control strain, none of the constructs showed mCherry expression, which is consistent with the notion that RamB Pd functions Replicates are shown as individual curves. The corresponding growth rates are given on the right. Asterisks indicate the level of significance as determined by t test. ****, P , 0.0005; n.s., not significant.

Coordination of Functional Degeneracy in Pd1222
Applied and Environmental Microbiology as a transcriptional activator. We further verified these findings in vitro by analyzing the interaction of purified RamB Pd (Fig. S4) with target DNA using biolayer interferometry. To this end, an 80 bp fragment of the gc promoter region comprising the two binding boxes separated by an 18 bp spacer and additional 17 bp on each side was immobilized on a biosensor and probed with increasing concentrations of protein, yielding an equilibrium dissociation constant K D of 1.9 mM ( Fig. 3C and D). Mutation of both binding boxes resulted in a complete loss of RamB Pd binding (Fig. 3C). Altogether, these experiments identified and confirmed the putative binding site of RamB Pd . RamB Pd is a transcriptional activator and repressor. Next, we analyzed the changes in the global transcriptome of the P. denitrificans Pd1222 WT and the DramB mutant induced upon a switch from succinate to acetate as the sole carbon source during mid-exponential growth close to the point at which we expected Icl production in the WT near its maximum (OD 600 at the time of sampling: 0.8, see Fig. 1A for comparison). In the WT, 43 genes were significantly upregulated under this condition ( Fig. S5A; Table S1). These included genes for acetate-CoA ligases (Pden_4231, Pden_4550, Pden_4966), several transporters as well as several enzymes of core metabolism, such as a-ketoglutarate dehydrogenase (Pden_4984 and Pden_4985) or glucose-1-phosphate dehydrogenase (Pden_4982); (Table S1). Notably, genes of the prp-operon also showed a strong (300-fold) upregulation on acetate ( Fig. 4A; Fig. S5A; Table S1). The prp-operon encodes the enzymes of the methylcitrate cycle (MCC), which converts propionyl-CoA via methylcitrate and methylisocitrate to succinate and pyruvate (8)(9)(10)(11). As propionyl-CoA is an intermediate in the EMCP, upregulation of the MCC indicated that this pathway might be active in parallel to or instead of the MMCP-branch of the EMCP on acetate (Fig. 4B, see Discussion). Interestingly, transcripts of ramB showed an average of 10 reads, which is below the limit of statistical significance, indicating a very low, basal expression of ramB under all conditions in the WT in line with its proposed function as a transcriptional regulator.
In addition, and as expected in the WT, expression of EMCP and GC genes was also increased upon growth switch to acetate. GC genes were upregulated more than 600fold on acetate in mid-exponential phase, while EMCP genes were only slightly upregulated (Fig. 4A, row 1; Table S1; Fig. S5A) (see Fig. 1A), which is consistent with the observed decrease of Ccr production in this growth phase. Unlike in the WT, the expression of GC genes did not change significantly in the DramB mutant upon a switch to acetate (  Table S4), indicating that RamB Pd might serve a dual function as a transcriptional activator on acetate and as a repressor on succinate. This hypothesis is in agreement with the observations that (i) RamB Pd was able to bind its target DNA even in the absence of a potential ligand ( Fig. 3C and D); (ii) addition of succinate to an acetate-grown culture in mid-log phase resulted in the downregulation of the GC (Fig. S7A); and (iii) deletion of ramB in the icl::icl-mCherry background resulted in higher basal mCherry fluorescence on succinate (Fig. S6). Interestingly, EMCP genes were also strongly upregulated in the DramB mutant compared to the WT strain, with 8-to 15-fold higher transcript levels of ccr, ecm, mcd, and mcl-1 compared to the WT on acetate (Fig.  4A, row 3; Table S3; Fig. S2C). Collectively, these results confirmed the central role of RamB Pd as regulator of acetate metabolism in Pd1222 and suggested that the protein can act both as a transcriptional activator and repressor.
Intermediates of the EMCP bind to RamB Pd . Previous studies showed that PrpR (from the MccR group of the ScfR family) from M. tuberculosis binds CoA-esters via a (lower panel; mutations correspond to those present in pTE714_1365/64_ig D1 1 2) was immobilized on Sartorius Octet SAX2 biosensors. The biosensor tips with immobilized DNA were probed with increasing concentrations of RamB Pd in binding buffer. Wavelength shifts resulting from an increase in the biolayer due to protein association were monitored over the course of time using a BLItz (FortéBio, USA). (D) Shown are the wavelength shifts reached at equilibrium at the end of the association phase plotted versus the concentration of protein present in the reaction. The K D was determined as the concentration at which the half-maximal wavelength shift was reached.

Coordination of Functional Degeneracy in Pd1222
Applied and Environmental Microbiology   (Fig. S9), which we confirmed by UV/Vis spectroscopy (see Materials and Methods).
Together, this suggested that RamB Pd also binds CoA-esters. Notably, the EMCP produces several CoA-ester intermediates and is always induced before the GC in the pre-exponential phase of Pd1222 cultures on acetate (OD600 , 0.1) regardless of the presence of additional carbon sources in the medium ( Fig. 5D; Fig. S8B) or the growth state of the preculture (Fig. S7C), suggesting that an active EMCP and/or its CoA-ester intermediates are required for GC expression. This hypothesis was further supported by the fact that a Dccr mutant showed a prolonged lag phase (between 30 h to 90 h) upon

Coordination of Functional Degeneracy in Pd1222
Applied and Environmental Microbiology a switch to acetate (22) (Fig. 5A and B), while the lag phase was unchanged in a Dicl strain under the same conditions (22) (Fig. 2B) (Fig. 4B). Growth on downstream metabolites, such as glyoxylate, propionate, or a mixture of both carbon sources, had no inducing effect on the expression of ccr or icl, and thus the EMCP or the GC (Fig.  5C). In contrast, growth on upstream substrates such as acetate (Fig. 1A), acetate plus glyoxylate (Fig. 5D), ethanol, or crotonate induced icl expression (22), independently of the presence of additional (i.e., 5% vol/vol) CO 2 , which is an important co-substrate of the EMCP and Ccr as well as propionyl-CoA carboxylase (PccAB) in particular (Fig. S7B). Altogether, these results indicated that the signal inducing expression of the GC was produced in the upper part of the EMCP. This notion was further supported by the fact that functional Ccr (and presumably an active EMCP) was required to induce icl. While icl expression was induced during growth on acetate plus glyoxylate (Fig. 5D, first  panel), it was lost under the same conditions in a Dccr strain (Fig. 5D, last panel).
We next aimed to identify the potential ligand of RamB Pd . The upper part of the EMCP, including the reactions between the reactions of Ccr and Mcl-1 produces four intermediates, namely (2S and 2R)-ethylmalonyl-CoA, (2S)-methylsuccinyl-CoA, mesaconyl-C1-CoA, and (2R/3S)-b-methylmalyl-CoA (4, 12) (Fig. 4B). To study the effect of the different-CoA esters on RamB Pd , we assessed the oligomeric state of RamB Pd in vitro in the absence and presence of the different CoA-esters using mass photometry (45,46). Besides a background peak at 80 kDa, RamB Pd was detected at a mass corresponding to a molecular weight of approximately 210 kDa (Fig. 6A), indicating that RamB Pd forms a tetramer in solution.
Incubation of the protein in the presence of free CoA resulted in a second peak that increased in size with higher concentrations of CoA, until reaching saturation at 62.5 mM CoA with a peak ratio of 1:1 dimer to tetramer. The protein species in this second peak has a molecular weight of ;105 kDa, corresponding to a dimer. These results suggest that the tetrameric RamB Pd complex dissociates into a dimer upon the addition of CoA. We next tested the influence of the different CoA ester intermediates of the EMCP on dimer formation (Fig. 6C to E). Ethylmalonyl-CoA did not induce dimer formation, while methylsuccinyl-CoA (which contained 54% free CoA) showed similar effects as free CoA. Incubation of RamB Pd with mesaconyl-CoA or crotonyl-CoA resulted in a 3:1 ratio of dimers to tetramers. In contrast, incubation with b-methylmalyl-CoA increased the ratio of dimers to tetramers to 9:1, with an apparent K D of the interaction of 700 mM. The strong shift toward dimer formation at physiologically relevant concentrations suggested that b-methylmalyl-CoA is the preferred ligand of RamB Pd .
Acetate assimilation is different in Rhodobacter capsulatus SB1003. We finally turned our attention to Rhodobacter capsulatus SB1003, an alphaproteobacterium that is closely related to P. denitrificans Pd1222 and possesses the genes of the EMCP, the GC and the MCC as well (47). Like Pd1222, this organism harbors several ScfR family members. Based on their locations in the sequence similarity network (Fig. S1) and the neighborhoods of their respective genes, we identified the proteins as putative RamB (regulator of the GC genes), putative PccR (regulator of propionyl-CoA carboxylase), and putative MccR (regulator of the MCC genes) (36), indicating a similar strategy of R. capsulatus SB1003 to assimilate acetate as Pd1222. To test this hypothesis, we deleted the genes of the key enzymes and (putative) regulators of acetate metabolism in R. capsulatus SB1003 (ccr, icl, pccR, and mccR genes were independently deleted) and analyzed the growth of the mutant strains on different carbon sources. Deletion of these genes did not affect the growth of R. capsulatus SB1003 on malate, the negative control (Fig. 7, line 1). As observed for Pd1222, deletion of ccr impaired the growth of R. capsulatus SB1003 on acetate (Fig. 7, column 3, line 2), emphasizing the importance of the EMCP for its ability to assimilate acetyl-CoA. However, unlike in Pd1222, deletion of icl did not affect the growth of R. capsulatus SB1003 on acetate (Fig. 7, column 2, line 2), indicating that Icl (and likely the GC) plays a minor role in acetyl-CoA assimilation in this species. Deletion of the regulatory pccR or mccR genes did not affect growth on acetate, but the deletion of pccR uniquely impaired growth on propionate. Deletion of mccR left growth of R. capsulatus SB1003 on propionate and CO 2 unaffected, indicating that the MCC was not able to substitute for the loss of the expression of MMCP genes, which requires PccR (Fig. 7, column 4, line 3) (36). In summary, these findings suggested that despite the presence of a common set of pathways and regulators, acetate metabolism strongly differs between these two organisms.

DISCUSSION
Organisms must balance metabolic flux between catabolic and anabolic routes during growth. One of the key intermediates in central carbon metabolism is acetyl-CoA, which can be either oxidized in the citric acid cycle (catabolism) or assimilated into biomass (anabolism). Notably, the alphaproteobacterium P. denitrificans Pd1222 employs two functionally degenerate acetyl-CoA assimilation pathways, the EMCP and the GC. It has been suggested that the EMCP represents the ancestral pathway in P. denitrificans Pd1222, while the GC was likely acquired through lateral gene transfer (22). The existence of two apparently redundant central metabolic traits in one organism is surprising and raises the question why P. denitrificans Pd1222 uses two parallel acetyl-CoA assimilation strategies and how the expression and activity of the two pathways are coordinated during growth of P. denitrificans Pd1222.
Here, we propose a novel mechanism of transcriptional control of GC genes that involves metabolic cross talk via RamB, a regulator of the ScfR family that binds upstream of the gc operon (Fig. 8). Induction of the GC is driven by flux through the EMCP, an alternative pathway for assimilating the same initial substrate.
The hypothesis that RamB links the expression of the GC to the activity of the EMPC is supported by several findings. First, the EMCP is (always) active at very basal levels in P. denitrificans Pd1222 and becomes upregulated in the pre-exponential growth phase after a switch to acetate and prior to induction of the GC. Second, deletion of the

Coordination of Functional Degeneracy in Pd1222
Applied and Environmental Microbiology EMCP-essential ccr gene results in a significantly prolonged lag in GC expression. Third, in vitro experiments with purified RamB Pd show that several intermediates of the EMCP, and in particular b-methylmalyl-CoA, cause a shift from the tetrameric to the dimeric state of RamB Pd at physiologically relevant ligand concentrations. Our findings show that P. denitrificans Pd1222 achieves metabolic plasticity through a complex entanglement and coordination of functionally degenerate central carbon metabolic pathways. Functional degeneracy is not uncommon in prokaryotes, and alphaproteobacteria in particular. Methylobacterium extorquens AM1, a related methylotroph, for instance, features two different methylamine assimilation routes that serve different

Coordination of Functional Degeneracy in Pd1222
Applied and Environmental Microbiology purposes and confer distinct advantages under different growth conditions (34,35). In the case of acetyl-CoA assimilation in P. denitrificans Pd1222, the EMCP might serve as a basic assimilation route, which is replaced by the more specialized and more efficient GC, when high fluxes require a strong commitment to acetyl-CoA assimilation (22). This work clarifies the molecular basis of GC regulation in P. denitrificans Pd1222, but several questions remain unanswered. While upregulation of the GC requires an active EMCP, as shown in this work, the mechanism upregulating the EMCP itself remains enigmatic. In addition, the regulation of acetate metabolism in P. denitrificans Pd1222 is more complex and seems to involve catabolite repression, because succinate suppresses acetate metabolism (Fig. S8) but apparently not EMCP expression. Moreover, externally added succinate directly represses the expression of GC genes. This indicates a (post-) transcriptional/translational feedback loop and points to a more integrated regulation of acetate metabolism in the context of central carbon metabolism in P. denitrificans Pd1222. This hypothesis is further supported by the observation that the deletion of ramB or icl increases EMCP expression, possibly to compensate for the loss of GC activity.
An interesting finding of our study is that the MCC is upregulated in addition to the EMCP and GC during growth on acetate (and even further increased in the absence of a functional GC in the DramB mutant). This indicates that propionyl-CoA might be additionally assimilated through the MCC under these conditions, possibly as a strategy to prevent accumulation of potentially toxic concentrations of propionyl-CoA in the cell (48). In any way, the additional upregulation of the MCC represents an interesting variation of the EMCP that might add to the functional degeneracy of P. dentrificans Pd1222 and, potentially, other organisms expressing the EMCP and the MCC.
We would like to note that the functional degeneracy of acetyl-CoA assimilation routes is accompanied by regulatory degeneracy: although R. capsulatus SB1003 possesses in principle the same metabolic capabilities (i.e., codes for the EMCP, GC, and MCC) and homologous regulatory proteins (i.e., RamB, PccR, MccR), the regulation of acetate metabolism (i.e., the induction of the different pathways), and the relevance of functional degeneracy of acetyl-CoA assimilation differs from P. denitrificans Pd1222. Future studies focusing on acetyl-CoA assimilation in P. dentrificans Pd1222 and/or R. capsulatus SB1003 will shed more light on and thus eventually complete our picture of functional degeneracy and metabolic plasticity in alphaproteobacteria. These findings are also of importance for efforts to reprogram and rewire bacterial central carbon metabolism in synthetic biology and biotechnology, which aim at introducing new metabolic modules into alphaproteobacteria and/or transplanting alphapreoteobacterial pathways into other model systems (49,50).

MATERIALS AND METHODS
Chemicals. All chemicals used in this study were obtained from Sigma-Aldrich (Steinheim, Germany) and Carl Roth (Karlsruhe, Germany) unless specified otherwise and were of the highest purity available.
Strains, medium, and cultivation conditions. All plasmids and strains used in this study are listed in Table 2. Escherichia coli was grown in Luria Bertani (LB) medium in the presence of appropriate antibiotics (ampicillin [100 mg/mL]; gentamycin [30 mg/mL]; kanamycin [50 mg/mL]; streptomycin [20 mg/mL]; tetracycline [10 mg/mL]) at 37°C. For the cultivation of E. coli ST18, media were supplemented with 5-aminolevulinic acid (50 mg/mL). Paracoccus denitrificans was grown in the presence of appropriate antibiotics (kanamycin [50 mg/mL]; spectinomycin [50 mg/mL]; rifampicin [30 mg/mL], tetracycline [10 mg/mL]) in LB medium or mineral salt medium supplemented with trace elements TE3-Zn (51) and defined carbon compounds at a total carbon concentration of 120 mM at 30°C. For strains carrying derivatives of the suicide plasmid pK18mobsacB, LB medium was replaced with super optimal broth (SOB) to lower the risk of point mutations in the sacB gene. R. capsulatus was grown at 30°C with peptone yeast extract (PYE) medium, as well as a modified version of minimal media (7) that, undiluted, contained 20Â salt solution (20 g/L NH 4   Escherichia coli F -mcrA D(mrr-hsdRMS-mcrBC) w 80lacZDM15 DlacX74 recA1 araD139 D(ara-leu)7697 galU galK l -rpsL(Str R ) endA1 nupG New England Biolabs S17-1 Escherichia coli pro, res 2 hsdR17 (rK 2 mK 1 ) recA 2 with an integrated RP4-2- measuring the OD 600 at regular intervals. The fluorescence of mCherry was measured at an emission wavelength of 610 nm after excitation at 575 nm with a constant gain of 120. The fluorescence of Cerulean was measured at an emission wavelength of 475 nm after excitation at 433 nm with a constant gain of 90. The growth of R. capsulatus SB1003 and its derivatives was followed in a SpectraMax i3 plate reader (Molecular Devices, USA) by monitoring the OD 600 . Genetic manipulation of P. denitrificans. The transfer of plasmids to P. denitrificans was achieved by biparental mating using E. coli ST18 as a donor strain (52). The deletion of genes from the P. denitrificans genome was carried out as described (22) with LB medium being replaced with SOB medium. The transfer of replicative plasmids to P. denitrificans was done according (53). The selection of transformants occurred on mineral salt medium supplemented with 120 mM methanol and appropriate antibiotics. The successful transfer of plasmids and the deletion of genes were verified by colony PCR. All strains and plasmids used in this study are listed in Table 2.
Genetic manipulation of R. capsulatus. R. capsulatus and E. coli cultures were grown overnight for at least 24 h. R. capsulatus and E. coli S17-1 were allowed to conjugate for at least 24 h to transfer the desired plasmids into R. capsulatus. R. capsulatus strains containing the plasmids were selected using PYE with 20 mg/mL kanamycin, followed by colony PCR. Plasmid-containing cells were grown in liquid PYE with no antibiotics for at least 24 h. Cultures were then centrifuged, diluted, and plated on PYE agar 1 10% sucrose or nutrient agar (NA) 1 10% sucrose. Colonies were patch-plated in parallel onto PYE agar and PYE agar with antibiotics. Colonies that grew on PYE agar but not on PYE agar with antibiotics were identified as potential deletion or fusion strains. PCR was performed to confirm the deletion of the desired gene.
Generation of plasmids. Plasmids pTE1630-pTE1632 were generated by amplification of the up-and downstream genomic regions of targeted genes from Pd1222 genomic DNA with oligonucleotides 1 to 12 and subsequent Gibson assembly (54) of matching PCR products as indicated in Table 3 with EcoRI-linearized, dephosphorylated pK18mobsacB. Plasmids pTE1633 and pTE1634 were generated by amplification of inserts from Pd1222 gDNA with oligonucleotides 13 to 16 and subsequent ligation of adequately restricted PCR products into equally restricted and dephosphorylated pTE714. Plasmids pTE1637-1639 were generated by inverse amplification of pTE1634 with oligonucleotides 17 to 21, followed by removal of the template strand with DpnI, phosphorylation and ligation (KLD reaction). Plasmid pTE5000 was generated by inverse amplification of pIND4 with oligonucleotides 22 to 23, amplification of Pden_1365 with oligonucleotides 24 to 25 from Pd1222 genomic DNA and subsequent Gibson assembly of the resulting fragments. Plasmid pTE5007 was generated in two steps. First, the insert fragment was amplified from Pd1222 gDNA with oligonucleotides 26 to 27 and inserted into SapI-restricted pTB146 by Gibson assembly. A missing codon for a glycine-residue in the future Ulp1 recognition site was then inserted into the resulting construct by inverse amplification with oligonucleotides 28 to 29 and followed by a KLD reaction. Plasmids pMJ009, pYN011, pYN019, and pPK005 were generated by amplification of up-and downstream genomic regions of targeted genes from SB1003 genomic DNA and subsequent Gibson assembly of matching products according to Table 3 with BamHI-and XbaI-linearized, dephosphorylated pK18mobsacB. All oligonucleotides used in this study and their purpose are listed in Table 3.
Global transcriptome analysis. Samples for transcriptome analysis were taken in triplicates from succinate-and acetate-growing cultures of P. denitrificans Pd1222 and the Pd1222 DramB deletion at OD 600 0.8. For sample collection, 2 mL of culture were transferred into sterile 2-mL Eppendorf tubes and pelleted by centrifugation at 10,000 Â g and 4°C for 10 min. Supernatant was discarded and samples were snap-frozen at 280°C. Storage of samples occurred at 280°C until further usage. The Qiagen miRNeasy Kit was used for total RNA isolation starting from a pellet of 2-mL cultured cells. The pellet was resuspended in 1 mL QiAzol lysis reagent and homogenized in a FastPrep sample preparation system using Lysing Matrix B containing 0.1 mm silica beads (MP biomedicals) and the following settings: 4 Â 6,500 rpm for 20 s, 15-s break. The supernatant was transferred into a new tube and RNA isolation was performed according to the manufacturer's instruction, including the optional on column DNase I digestion. Depletion of rRNA was performed and cDNA libraries were prepared with the Illumina Stranded Total RNA Prep Kit (Ligation with Ribo-Zero Plus; Illumina, USA). Library quality was assessed with the fragment analyzer employing the Agilent HS NGS Fragment kit (Agilent, USA). Quantification of libraries was done by qPCR and the KAPA library quantification Kit (Roche, Switzerland) in a qTOWER3 G Thermal Cycler (Analytic Jena, Germany). Libraries were paired-end sequenced on an Illumina MiSeq using the MiSeq reagent kit V3 featuring 150-bp read length. Sequencing data are available from ArrayExpress (reference number: E-MTAB-12482). The differential gene expression analysis was done as described elsewhere (55) using the RNA-seq pipeline Curare 0.3.1 (https://github.com/pblumenkamp/ Curare). Briefly, reads were preprocessed with Trim Galore 0.6.7 (56) and Cutadapt 3.5 (57) with a quality   (63) unless specified otherwise. b-Methylmalyl-CoA was synthesized from propionyl-CoA and a 12-fold molar excess of glyoxylate (in buffer: 100 mM MOPS/KOH pH 7.5, 5 mM MgCl 2 ) using heterologously expressed and purified Rhodobacter sphaeroides Mcl-1 (accession number ACI22682) as catalyst. The resulting esters were purified via preparative HPLC-MS with a 1260 Infinity II LC System (Agilent, USA) in combination with a 6130 Single Quadrupole LC/MS (Agilent, USA). Lyophilized CoA esters were stored at 280°C and dissolved in ddH 2 O before use. The concentrations of the respective CoA esters were determined photometrically using a Carry 60 UV-Vis spectrophotometer (Agilent, USA) and calculated from the absorbance at 260 nm using defined extinction coefficients (22, Heterologous production and purification of proteins. 6xhis-sumo-ramB was heterologously expressed from plasmid pTE5007, respectively, in E. coli BL21 AI (Novagen, Germany) grown in terrific broth (TB; 24 g/L yeast extract, 20 g/L tryptone, 0.017 M KH 2 PO 4 , 0.072 M K 2 HPO 4 , 10% [wt/vol] glycerol) in the presence of ampicillin. Gene expression was induced by addition of 0.25% arabinose and 0.5 mM IPTG when the cultures had reached an OD 600 of 1.0 to 2.0. 6xhis-ulp1 was expressed from plasmid pTB145 (64) in E. coli Rosetta (DE3)pLysS (Novagen, Germany) grown in TB supplemented with ampicillin and gentamicin. Gene expression was induced by addition of 0.5 mM IPTG when cultures had reached an OD 600 of 1.0 to 2.0. Overproduction of all proteins occurred at 25°C overnight and was verified by SDS-PAGE analysis of culture samples taken before and after induction. The cultures were harvested by centrifugation at 8,000 Â g and 10°C for 15 min. For lysis, the cells were resuspended in buffer A (50 mM Tris, 500 mM NaCl, pH 7.8) supplemented with 0.5 mM DTT, DNase I, and 5 mM MgCl 2 and treated by four cycles of sonication (50%, 1s pulse, 60 s) using a Sonopuls GM200 sonicator (BANDELIN, Germany). Cell lysates were cleared by ultracentrifugation at 100,000 Â g and 4°C for 45 min followed by filtration through a Filtropur S 0.45 mm filter (Sarstedt, Germany). His-tagged proteins in cell lysate were loaded on pre-equilibrated Ni-NTA agarose beads in Protino drop down columns (Macherey-Nagel, Germany), washed with buffer A and eluted with increasing concentrations of buffer B (50 mM Tris, 500 mM NaCl, 500 mM imidazole, pH 7.8). Buffer exchange was performed with Cytiva PD-10 Desalting columns (Cytiva, USA) and elution in desalting buffer (50 mM Tris, 350 mM NaCl, 10% glycerol, pH 7.8). His-SUMO tag was cleaved off by treatment of the fusion proteins with His-Ulp1 protease (1 U per 2 mg of target protein) in desalting buffer, followed by removal of His-SUMO protein with pre-equilibrated Ni-NTA agarose beads. Successful purification of proteins, as well as proteolytic cleavage of tags was verified by SDS-PAGE. Protein concentrations were determined using a NanoDrop 2000 (Thermo Fisher Scientific, USA). After successful purification, RamB Pd was routinely analyzed via UV-vis spectroscopy to verify the presence of the previously described [4Fe4S] cluster as displayed by a characteristic peak at 410 nm in UV-vis spectroscopy (39).
Mass photometry. The molecular weights of protein-protein complexes were determined in phosphate-buffered saline (PBS; pH 7.4) on 1.5 H, 24 Â 60 mm microscope coverslips (Carl Roth, Germany) and Culture Well Reusable Gaskets (GRACE BIO-LABS, USA) using a Refeyn One mass photometer (Refeyn Ltd., UK) with the AcquireMP software (Refeyn Ltd., UK). A previously measured standard was used at a concentration of 20 nM as a reference. The processing and analysis of mass photometry images was performed using DiscoverMP (Refeyn Ltd., UK).
Biolayer interferometry. The analysis of protein-DNA interactions by biolayer interferometry (BLI) was performed using Sartorius Octet SAX2 biosensors in binding buffer (desalting buffer supplemented with 0.01% Tween 20 and 10 mM bovine serum albumin (BSA)) on a BLItz platform (FortéBio, USA). Biotinylated dsDNA at 40 mM concentration was generated by annealing oligonucleotides 46 and 47 for wt DNA with intact binding sites and oligonucleotides 48 and 49 (Table 3) for DNA with mutated binding sites in Phusion high GC buffer (Thermo Fisher Scientific, USA) dsDNA stocks were diluted 20-fold in binding buffer to reach a working concentration of 2 mM. To immobilize biotinylated dsDNA fragments on the biosensor tips, 4 mL of the 2 mM stock were applied to the sample drop holder and the biosensor tip was placed in it. DNA loading was followed for 120 s until equilibrium was reached. Afterwards, the biosensor tip was transferred back into binding buffer to wash off excess DNA fragments for 30 s. The association of protein with the DNA-loaded biosensor tip was followed as described for DNA loading. After each sample loading step, the sample drop holder was cleaned with 0.5 M NaOH, followed by two washes with ddH 2 O. After the association step, the biosensor tip was transferred into binding buffer again and protein dissociation was followed for 120 s. A new biosensor tip was used for each measurement.

SUPPLEMENTAL MATERIALS
Supplemental material is available online only. SUPPLEMENTAL FILE 1, DOCX file, 7.3 MB.

ACKNOWLEDGMENTS
We acknowledge technical assistance by the Bioinformatics Core Facility at the professorship of Systems Biology at JLU Giessen and provision of compute resources and general support by the BiGi service center (BMBF grant 031A533) within the de.NBI