Co-catabolism of arginine and succinate drives symbiotic nitrogen fixation

Biological nitrogen fixation emerging from the symbiosis between bacteria and crop plants holds a significant promise to increase the sustainability of agriculture. One of the biggest hurdles for the engineering of nitrogen-fixing organisms is to identify the metabolic blueprint for symbiotic nitrogen fixation. Here, we report on the CATCH-N cycle, a novel metabolic network based on co-catabolism of plant-provided arginine and succinate to drive the energy-demanding process of symbiotic nitrogen fixation in endosymbiotic rhizobia. Using systems biology, isotope labeling studies and transposon sequencing in conjunction with biochemical characterization, we uncovered highly redundant network components of the CATCH-N cycle including transaminases that interlink the co-catabolism of arginine and succinate. The CATCH-N cycle shares aspects with plant mitochondrial arginine degradation path-way. However, it uses N2 as an additional sink for reductant and therefore delivers up to 25% higher yields of nitrogen than classical arginine catabolism — two alanines and three ammonium ions are secreted for each input of arginine and succinate. We argue that the CATCH-N cycle has evolved as part of a specific mechanism to sustain bacterial metabolism in the microoxic and acid environment of symbiosomes. In sum, our systems-level findings provide the theoretical framework and enzymatic blueprint for the rational design of plants and plant-associated organisms with new properties for improved nitrogen fixation. Significance Statement Symbiotic bacteria assimilate nitrogen from the air and fix it into a form that can be used by plants in a process known as biological nitrogen fixation. In agricultural systems, this process is restricted mainly to legumes, yet there is considerable interest in exploring whether similar symbioses can be developed in non-legumes including cereals and other important crop plants. Here we present systems-level findings on the minimal metabolic function set for biological nitrogen fixation that provides the theoretical framework for rational engineering of novel organisms with improved nitrogen-fixing capabilities.

designer nitrogenase gene clusters in bacteria (9)(10)(11)(12). Despite 23 these promising results, engineered organisms based on het-24 erologous expression of nitrogenase genes have not yet come 25 close to the efficiency of natural rhizobia-legume symbiosis sys-26 tems (4). While the molecular mechanism of the nitrogenase 27 reaction has been resolved with atomistic detail (13)(14)(15)(16), the 28 precise mechanism how the metabolism between plants and 29 bacteria becomes entangled to sustain the energy-intensive 30 process of nitrogen fixation has remained an open question. 31 The current model of nutrient exchange in rhizobia-legume 32 symbiosis postulates that, in exchange for fixed nitrogen, the 33 plant provides dicarboxylic acids such as succinate, which is 34 metabolized through the tri-carboxylic acid (TCA) cycle to 35 generate ATP and reduction equivalents needed for the nitro-36 genase reaction (17)(18)(19). However, multiple lines of evidence 37 argue against a simple exchange of succinate for ammonium 38 during symbiosis (20,21). While the absence of oxygen pro-39 motes nitrogenase activity, it also inhibits TCA-mediated 40 catabolism of succinate due to redox inhibition of key TCA-41 enzymes including citrate synthase, isocitrate dehydrogenase, 42 and 2-oxoglutarate dehydrogenase (22,23). Thus, the TCA 43 cycle probably operates below its full aerobic potential. Fur-44 thermore, if the metabolism of symbiotic nitrogen-fixing bacte-45 ria is based exclusively on the provision of succinate, then the 46 bacterial nitrogen requirement must be covered solely by nitro-47 genase. However, nitrogen-fixing root-nodule bacteria (termed 48 bacteroids), do not self-assimilate but rather secrete large quan-49 tities of ammonium (24)(25)(26) suggesting that the plant provides 50 the bacteroids with a nitrogen-containing nutrient to cover 51 their nitrogen needs. Finally, the degradation product of succi-52 nate in the TCA cycle is carbon dioxide. However, it has been 53 reported that nitrogen-fixing bacteroids also secrete the amino 54

Significance Statement
Symbiotic bacteria assimilate nitrogen from the air and fix it into a form that can be used by plants in a process known as biological nitrogen fixation. In agricultural systems, this process is restricted mainly to legumes, yet there is considerable interest in exploring whether similar symbioses can be developed in non-legumes including cereals and other important crop plants. Here we present systems-level findings on the minimal metabolic function set for biological nitrogen fixation that provides the theoretical framework for rational engineering of novel organisms with improved nitrogen-fixing capabilities.  nodule extracts (92% ±6% and 92% ±6%) for B. diazoefficiens 114 (Fig. 1A) and S. meliloti bacteroids respectively. Therefore, 115 we concluded that the co-feeding of arginine and succinate is 116 sufficient to stimulate nitrogenase activity in bacteroids.

117
The nitrogenase enzyme complex catalyzes one of the most 118 energy-consuming enzymatic reactions found in nature with 119 16 ATP molecules and 8 low-potential electrons required for 120 the reduction of a single nitrogen molecule. Nitrogenase is 121 irreversibly inactivated in the presence of oxygen, which re-122 stricts the reduction of atmospheric nitrogen to low-oxygen 123 conditions. Thus, to support nitrogen fixation, bacteroids 124 must produce substantial amounts of ATP under microoxic 125 conditions. The finding that succinate as the sole nutrient 126 did not result in nitrogenase stimulation suggested that succi-127 nate catabolism via the TCA cycle does likely not generate 128 sufficient ATP to support efficient nitrogenase reaction.  (Table 1). Upon the addition 150 of 13 C arginine, we observed a rapid increase in the labeled 151 intracellular arginine pool (99.43% 13 C), demonstrating active 152 arginine transport into nitrogen-fixing bacteroids.

153
Upon further incubation, we found 13 C isotope labels in  (Table 1). Also, citrulline, which represents the first 159 step of the arginine deiminase (ADI) pathway, was labeled with 160 60.22% 13 C. This finding suggests that bacteroids employ the 161 ADI pathway for anaerobic, enzyme-coupled ATP production 162 by the enzyme carbamate kinase.    We reasoned that TnSeq provides a unique opportunity to 187 6.26 ± 1.06 a 13 C Fractional labeling after 150 min incubation with 13 C L-arginine. Shown is the average and the standard error of the mean (SEM).
identify specific arginine transport and degradation genes that 188 become essential upon engagement in symbiosis.

205
Functional classification revealed that the large majority of 206 symbiosis genes comprise cellular functions such as metabolism 207 (507 genes, 51.89%), gene regulation (196 genes, 20.06%) and 208 other cellular processes (228 genes, 23.34%) (Fig. 2C, Data SI). 209 The identified gene set included well-characterized symbiosis 210 factors involved in nodulation (34 genes, 3.48%) as well as 211 functions associated with the nitrogenase enzyme complex 212 (12 genes, 1.23% ). Collectively, a set of 177 symbiosis genes, 213 corresponding to over one-third of the 507 metabolic symbiosis 214 genes (34.91%), was associated with nitrogen metabolism 215 including genes for the transport (59 genes), biosynthesis 216 (78 genes) and degradation (40 genes) of amino acids and 217 other nitrogen-containing compounds. While only 3 out of 218 78 essential biosynthesis genes (3.85%) were involved in the 219 synthesis of arginine, we found a large fraction of 18 out of 59 220 essential transport genes (30.51%) and 22 out of 40 essential 221 catabolic genes (55.00%) annotated as being involved in the 222 uptake and catabolism of arginine and its derivatives. In sum, 223 these findings highlight that the provision of arginine and its 224 consecutive degradation is of fundamental importance to drive 225 symbiotic nitrogen fixation in legumes.

234
In addition, two arginine/agmatine antiporter genes adiC 235 (SMa0684) and adiC2 (SMa1668) encoded on pSymA were 236 also essential during symbiosis. Interestingly, all identified 237 transport systems participate in urease, and ADI pathways 238 that mediate acid tolerance. Cross-comparing expression pro-239 files using previously published RNA-seq data sets (39), we 240 found that artABCDE was the only transport system to be 241 constitutively expressed during all stages of symbiosis, while 242 the expression of all other transporters was specifically in-243 duced during development into nitrogen-fixing compartments 244 (symbiosomes).

245
To gain further insights, we searched for additional symbio-246 sis genes related to acid tolerance and indeed found multiple 247 essential components in the TnSeq dataset (Data SI). From 248 the urease pathway, we identified two arginases (argI1 and 249 argI2 ) and the urease components ureA and ureE. From the 250 ADI system, we found the arcABC operon to be essential for 251 symbiosis. Both systems catalyze the conversion of arginine 252 into ornithine leading to the production of ammonia as part 253 of the acid tolerance mechanism. Furthermore, the ADI sys-254 tem also provides ATP via the enzymatic step of ornithine 255 carbamoyltransferase arcB (40). Interestingly, two additional 256 copies of ornithine carbamoyltransferase were also essential 257 (argF1, and arcB2 ), emphasizing the importance of genetic 258 redundancy in ADI dependent ATP synthesis during symbiosis. 259 The urease and ADI acid tolerance mechanisms rely on the 260 efflux of ammonium (41). Indeed, the ammonium efflux pump 261 encoded by amtB was among the top-ranked symbiosis genes. 262 These findings underscore the importance of ammonium se-263 cretion as a compulsive property of bacteroids independent of 264 the nitrogenase reaction.

Arginase gene deletions show nitrogen starvation pheno-266
types during plant infection assays. To validate the impor-267 tance of the identified arginine-dependent acid tolerance sys-268 tems for symbiosis, we constructed a panel of deletion mutants 269 of the urease and ADI pathways and assessed nitrogen star-270 vation phenotypes during plant infection assays. Out of the 271 8 mutants evaluated, all displayed symbiotic defects. On 272 the level of the arginine transport systems, we found that 273 artABCDE and satABC showed a reduction in nitrogenase 274 activity of 47.06% ± 7.27% and 55.45% ± 10.99%. Similarly, 275 gene deletions in the urease pathway such as the arginases 276 mutants argI1 and argI2 exhibited a reduction of 71.18% ± 277 5.21% and 70.97 ± 5.16% for single deletions and 80.89% ± 278 3.15% for the double deletion mutant. Deletion of the urease 279 ureGFE and the ammonium efflux system amtB resulted in a 280 64.68% ± 5.50% and 80.90% ± 4.81% reduction in nitrogenase 281 activity.

282
Plants inoculated with the argI1,argI2 single and double 283 deletion mutant harbored a typical phenotype of nitrogen 284 starvation. The aerial part of infected plants was smaller than 285 those inoculated with WT strain (Fig. 3A, S1). Nodules 286 induced by the argI1,argI2 double deletion mutant displayed 287 the yellowish color of non-functional M. truncatula nodules 288 (Fig. 3B). Furthermore, observations of argI1,argI2 nodule 289 sections by scanning electron microscopy showed that nodules 290 were hollow and remaining bacteroids exhibit aberrant cell 291 morphology (Fig. 3C, S2). Collectively, our results suggest 292 that the identified arginine-dependent acid tolerance system 293 is a prerequisite for the faithful establishment of symbiosis.

294
Identification of AspC as an arginine:pyruvate transaminase. 295 In our isotope labelling studies with bacteroids, we detected 296 fractional labelling of 90.40% for GBA suggesting the presence 297 of a functional ATA pathway. However, in the S. meliloti 298 genome, corresponding genes have not been assigned. Among 299 the essential symbiosis genes identified by our TnSeq analy-300 Leghemoglobin content 500 µm  In the lower part of the reaction network, a linear pathway 340 through transamination and subsequent dehydrogenase steps 341 leads from GABA to succinate (Fig. 4A). Upon heterologous 342 expression and protein purification, we biochemically profiled 343 a panel of 16 candidate enzymes. In addition to the argi-344 nine deiminases ArcA1 and ArcA2 and the arginase ArgI1, 345 we found two agmatinase (ArgI2 and SpeB) and one ureohy-346 drolase (SpeB2) acting on 4-guanidinobutyraldehyde (GBL), 347 GOP and GBA (Fig. 4B). The highest level of pathway re-348 dundancy resides on the level of dehydrogenases. Besides 349 the five known GabD1-5 proteins from S. meliloti, we identi-350 fied four additional isoenzymes Gab6-9. Thereof GabD6 and 351 GabD7 share a dehydrogenase profile identical to GabD1 for 352 4-aminobutyraldehyde (ABL), succinic semialdehyde (SSA) 353 and GBL. GabD8 and GabD9 exhibited substrate specificity 354 for GBL (Fig. 4B)    transaminases, we identified three additional enzymes (DatA, 356 AatB, and ArgD) that exhibited substrate preferences for or-357 nithine, putrescine and agmatine and two enzymes (GabT2, 358 and GabT3) with preferences for GABA (Fig. 4B). Simi-359 larly, we also profiled two additional decarboxylase enzymes 360 (OdcA, and OdcB) that either catalyzed the decarboxylation of 361 ornithine or AOP (Fig. 4B). Collectively, these results demon-  (Fig. 4C). We observed that 90% of the arginine 369 was rapidly metabolized within 30 minutes. As expected, or-370 nithine and GOP appeared as the first intermediates and then 371 concomitantly decreased with the appearance of the second 372 level of intermediates putrescine, AOP and ABL that ulti-373 mately converted into GABA, SSA, and succinate. During 374 the process, alanine steadily increased demonstrating that 375 pyruvate transamination couples the conversion of arginine 376 into succinate (Fig. 4D). In sum, these findings demonstrate 377 the synthetic reconstruction of the transamination network 378 that permits the co-catabolism of arginine and succinate.

379
The catabolism of succinate and arginine is interlinked. Since 380 the ATA network consumes two equivalents of pyruvate but 381 generates only a single equivalent during its operation, we 382 reasoned that the network strictly depends on the provision 383  5. Model of the CATCH-N cycle operating in N2-fixing bacteroids. Arginine (blue) and succinate (brown) are co-feed to bacteroids in equimolar ratio. Co-catabolism is interlinked through an arginine-pyruvate transamination step yielding alanine. Alternatively, an arginine-oxaloacetate transamination step yields aspartate. Enzymatic conversion of arginine releases ammonium (blue) independent from the nitrogenase reaction (green). NADH (magenta) produced through the co-catabolism is regenerated over a bifurcated electron transport chain onto terminal acceptors oxygen and nitrogen.
of additional pyruvate, which must be formed by simulta- Based on these findings, we devised a model that restores 418 electron flow from NADH to QH2 by electron bifurcation to 419 nitrogenase and the high-affinity terminal oxidase (Fig 5). If 420 this is the main pathway that permits regeneration of NADH, 421 then all components must be essential in symbiosis. Indeed, 422 we found genes encoding for components of the nitrogenase 423 nifHDK, the electron bifurcation complex fixABCX and the 424 alternative complex IV fixNOQP1-3 among the top-ranked 425 symbiosis genes in the TnSeq dataset (Supplementary Dataset 426 SI). These genetic evidences suggested that a bifurcated elec-427 tron transport chain operates during nitrogen-fixing symbiosis. 428 Estimation of the ATP balance of the bifurcated electron 429 transport chain. The endergonic branch of the electron bi-430 furcation reaction generates low-potential reducing equivalents 431 in the form of flavodoxin hydroquinone (Fld hq ) for nitrogenase 432 catalysis (Fig 5). For every Fld hq the nitrogenase consumes 433 two additional ATP molecules. However, the exergonic branch 434 of the electron bifurcation reaction translocates only three 435 protons corresponding to a single ATP that is generated per 436 electron passing from QH2 to coenzyme Q onto oxygen. Thus, 437 the electron bifurcation of each NADH appears to be asso-438 ciated with a net loss of one ATP. In addition to NADH, 439 the CATCH-N cycle also provides QH2 via succinate dehydrogenase ( Fig 5). Thereby, up to two ATP are generated Here, we report on the CATCH-N cycle operating on the co-485 feeding of arginine and succinate as part of a specific metabolic 486 network that drives the process of symbiotic nitrogen fixation. 487 The CATCH-N cycle shares aspects with the plant mitochon-488 drial arginine degradation pathway (50, 51), however, it de-489 livers up to 25% higher yield in nitrogen in the form of two 490 alanines and three ammonium secreted for each co-feed argi-491 nine and succinate. Thus, from the plant's perspective, the 492 CATCH-N cycle multiplies the nitrogen releasing capacity of 493 arginine. On the level of bacteroids, the CATCH-N cycle pro-494 vides an elegant solution for maintaining an active respiratory 495 chain under the highly acidic and microoxic conditions present 496 within the lysosomal compartment of the symbiosome. Thus, 497 the CATCH-N cycle also functions as an effective mechanism 498 to promote the survival of bacteroids within infected plant 499 cells. Equimolar arginine and succinate serve as substrates 500 and a molar ratio of nitrogen to the oxygen of 1:4 is required to 501 operate the CATCH-N cycle. Therefore, nitrogen-fixation still 502 depends on oxygen as terminal acceptor, while harnessing ele-503 mentary nitrogen as the second electron acceptor for reducing 504 equivalents generated by the metabolism. Also, the CATCH-N 505 cycle requires a constant flux of 8 protons into the symbiosome 506 to maintain the pH balance of the reaction. These protons 507 must be translocated by the action of plant ATPases as part of 508 the symbiosome acidification process. Thus, the operation of 509 the CATCH-N cycle depends on the presence of an active plant 510 metabolism. From the nitrogen balance standpoint, a feedback 511 loop exists between the nitrogenase function of bacteroids and 512 the availability of arginine within the host plant. Ammonium 513 released by bacteroids is rapidly incorporated by plant cells 514 into glutamate, glutamine, and aspartate that all serve as 515 precursors for the biosynthesis of ornithine and subsequently 516 for arginine occurring within chloroplasts. The output of the 517 CATCH-N cycle results in a net gain of assimilated nitrogen 518 that subsequently amplifies the plant's arginine biosynthesis 519 capacity as part of a positive feedback mechanism. As human-520 ity faces global challenges with population growth and climate 521 change, we need to rethink how tomorrows agriculture will 522 look like. Thereby systems-biology approaches to broaden 523 our understanding of plant-microbe interactions, as well as 524 the design of synthetic nitrogen-fixing microbes that mimic 525 natural symbiosis with plants, hold significant promise. Our 526 integrated model of the CATCH-N cycle provides new insights 527 into the principles underlying legume symbiosis and comprises 528 an important stepping stone for the rational biotechnological 529