Nitrogen conservation, conserved: 46 million years of N-recycling by the core symbionts of turtle ants

Nitrogen acquisition is a major challenge for herbivorous animals, and the repeated origins of herbivory across the ants have raised expectations that nutritional symbionts have shaped their diversification. Direct evidence for N-provisioning by internally housed symbionts is rare in animals; among the ants, it has been documented for just one lineage. In this study we dissect functional contributions by bacteria from a conserved, multi-partite gut symbiosis in herbivorous Cephalotes ants through in vivo experiments, (meta)genomics, and in vitro assays. Gut bacteria recycle urea, and likely uric acid, using recycled N to synthesize essential amino acids that are acquired by hosts in substantial quantities. Specialized core symbionts of 17 studied Cephalotes species encode the pathways directing these activities, and several recycle N in vitro. These findings point to a highly efficient N-economy, and a nutritional mutualism preserved for millions of years through the derived behaviors and gut anatomy of Cephalotes ants. Category Biological Sciences-Evolution

came from such specialized groups we extracted 16S rRNA fragments >200bp from each 146 metagenome library. Top BLASTn hits were downloaded for each sequence, and jointly used in 147 a maximum likelihood phylogenetic analysis. In the resulting tree (Fig. S5), 94.4% of our 335 148 Cephalotes symbiont sequences grouped within 10 cephalotine clades that included sequences 149 from prior in vivo studies. Inferences on metagenome content have, hence, been made using 150 partial genomes from the dominant, specialized core taxa. 151 152 Classification of assembled scaffolds took place using USEARCH comparisons against public 153 reference genomes in IMG and the KEGG database (see Materials & Methods,and 154 Supplementary Methods for more detail). Results from these analyses paralleled our 16S rRNA-155 based discoveries of a highly conserved core microbiome (Fig. 3). In all metagenomes, 156 Cephaloticoccus symbionts 44 from the Opitutales were the most dominant, with scaffolds from 157 these bacteria typically forming a single "cloud" differentiated from others by depth of coverage 158 and %GC content. Xanthomonadales scaffolds were ubiquitous and typically abundant, with 159 multiple scaffold clouds often evidencing co-existence of distinct strains, with up to ~10% 160 average %GC divergence. Less abundant, though still ubiquitous across metagenomes were 161 clouds of scaffolds from the Pseudomonadales, Burkholderiales, and Rhizobiales. Multiple 162 scaffold clouds at differing depths of coverage were consistent with multiple strain co-existence, 163 for each taxon, within most microbiomes. Unlike these core groups, Flavobacteriales,164 Sphingobacteriales, and Campylobacterales were common but not ubiquitous. For instance, 165 presence/absence calls for N-metabolism genes (Table S5) suggested the absence or extreme 166 rarity of these symbionts in several metagenomes. 167 Consistent with our acetylene reduction experiments using C. varians, IMG/ER based annotation 170 recovered no N-fixation genes in any of the 18 metagenome libraries. This absence encompassed 171 genes encoding the molybdenum-containing nitrogenase system (i.e. nifD, nifH, nifK), and those 172 from the iron-only (anfD, anfG, anfH, anfK) and vanadium-containing (vnfD, vnfG, vnfH, vnfK) 173 systems (Table S5) Table S5). The presence of complete gene sets for the core protein subunits of 181 the urease enzyme in all sampled microbiomes suggests that symbionts from most Cephalotes 182 species can make ammonia from N-waste urea. Taxonomic classification for urease gene-183 encoding scaffolds suggested that abundant Cephaloticoccus symbionts (order: Opitutales) 184 encoded all three core urease genes. Complete copies of each gene were found on a single 185 Opitutales-assigned scaffold within 15 of 18 metagenomes. Genes encoding the urease accessory 186 proteins (ureF, ureG, and ureH) were often found on these same scaffolds, with a strong trend of 187 conserved architecture for this gene cluster (Fig. S6). 188 189 Urease genes were occasionally assigned to other bacteria ( Fig. 4; Table S5), suggesting that 190 more than one symbiont can participate in this recycling function. Notable were cases from C. rohweri (Xanthomonadales), C. grandinosus (Burkholderiales), and C. eduarduli (unclassified 192 Bacteria), which hosted additional bacteria encoding complete sets of urease core and accessory 193 proteins. In these cases, urease genes mapped to single scaffolds with identical gene order to that 194 seen for Cephaloticoccus (Fig. S6). Urease function was also inferred for Rhizobiales bacteria in 195 several Cephalotes species. Rhizobiales-assigned scaffolds encoding urease genes differed from 196 those of Cephaloticoccus with respect to gene order, the presence of the ureJ accessory gene, 197 and the existence of a gene fusion between ureA and ureB (Fig. S6). 198 199 A maximum likelihood phylogenetic analysis of UreC proteins encoded by the sampled 200 microbiomes identified two distinct Cephalotes-specific lineages (Fig. S8). The first (bootstrap 201 support = 99%) consisted of Rhizobiales-assigned UreC proteins, with relatedness to homologs 202 from various families in the Rhizobiales. The second (bootstrap support = 93%) was comprised 203 of proteins assigned to Opitutales,Burkholderiales,Xanthomonadales,and unclassified Bacteria,204 showing relatedness to homologs from bacteria in the Rhodocyclales (Betaproteobacteria). 205

206
Metagenomic analyses: ammonia assimilation and amino acid synthesis genes found across 207 numerous core taxa 208 The above results provide genetic mechanisms to explain symbiont-mediated urea recycling in 209 C. varians, suggesting a broad distribution for this function across the Cepalotes genus. Also 210 necessary to explain our experiments are: 1. symbiont genes to assimilate the ammonia made 211 from urea degradation, and 2. symbiont genes using this assimilated N to synthesize amino acids. 212 Assessment of our metagenomes met these expectations in C. varians and all 16 other host 213 species. But in contrast to our findings for a small number of urea recyclers, genes involved in 214 these processes assigned to all core symbiont taxa, suggesting extensiove metabolic overlap. 215 216 Within all metagenomes, numerous taxa encoded complete gene sets for ammonia assimilation 217 (e.g. Figs. 4, S7, S9, S10; Table S5). Similarly, gene sets for the synthesis of each essential and 218 non-essential amino acid were complete in all metagenomes. With the exception of histidine 219 synthesis, complete only for Campylobacterales, each amino acid biosynthesis pathway was 220 complete within multiple bacterial taxa (Figs. 4, S6, S7, S9, S10; Table S5). Xanthomonadales 221 and Burkholderiales bins (outside of C. angustus) encoded all genes to synthesize the remaining 222 eight essential amino acids. This paralleled findings for Opitutales, Rhizobiales, and 223 Pseudomonadales: the former typically showed an incomplete pathway for lysine, while the 224 latter two often seemingly lacked a single gene for methionine biosynthesis. Sphingobacteriales 225 and Flavobacteriales lacked required genes in the lysine and methionine pathways. And 226 pathways for threonine, valine, leucine, isoleucine, tryptophan, and phenylalanine were 227 occasionally missing all or most genes within the Flavobacteriales. Many of the core taxa also 228 possessed complete gene sets for the synthesis of non-essential amino acids. 229

230
Metagenomic analyses: uric acid synthesis and degradation, and other means of urea production 231 Uric acid is a major waste product of many insects. This compound is also found in bird excreta, 232 a common Cephalotes food. To analyze capacities to recycle N from uric acid we examined gene 233 content in the pathway converting this compound into urea. Scaffolds assigning to Hymenoptera, 234 thus likely originating from Cephalotes genomes, often contained a subset of genes involved in uric acid degradation, including one encoding the canonical uricase enzyme (uaZ) and one 236 encoding breakdown of 5-HIU (uraH; Genes synthesizing urea from uric acid mapped to numerous scaffolds across several 252 metagenomes. However, in seven libraries, they mapped to just one or two Burkholderiales-253 assigned scaffolds. Synteny was conserved in these cases, and such scaffolds also possessed 254 additional genes encoding the subunits of xanthine dehydrogenase (Fig. S6), an enzyme 255 converting xanthine into uric acid. Core symbionts appear to produce xanthine via purine 256 recycling, as genes for guanine deaminase enzymes (Fig. S13) classified to Burkholderiales in 16 metagenomes (Table S5). Adenine deaminase enzymes were similarly encoded by bacteria, 258 namely Rhizobiales, across 10 metagenomes. 259 260 Further analyses of our metagenomes revealed that bacteria aside from Burkholderiales can 261 produce urea through other mechanisms ( Fig. 4; Table S5). For instance, across most hosts, 262 microbes from the Burkholderiales, Rhizobiales, Xanthomonadales, and/or Pseudomonadales 263 possessed arginase genes, catalyzing a reaction that converts arginine to urea and ornithine. In 264 several metagenomes, arginase genes also binned to Hymenoptera, suggesting their presence in 265 Cephalotes genomes. Genes for a separate, two-step pathway converting arginine to urea ( Multiple strains for many of the aforementioned core taxa often co-exist within a single gut 271 community 43 . So despite pathway "completeness" assessed at the level of host order, it remains 272 unclear whether individual symbiont strains encode complete pathways for key aspects of N-273 metabolism. We addressed this through genome sequencing of cultured symbiont strains from 274 five of the eight core bacterial taxa across two host species (i.e. C. varians and C. rohweri). The 275 14 strains prioritized for sequencing were chosen based on 16S rRNA gene identity (or near 276 identity) in comparison to core symbionts previously sampled through culture-independent 277 techniques ( Fig. S5). Alignments of C. varians isolates to scaffolds from conspecific 278 metagenomes ( Fig. S15) indicates that these strains or very close relatives are present in vivo, 279 supporting the relevance of in vitro findings from these strains to the natural gut community..  Table S7) included the discovery of a 282 Burkholderiales strain (Cv33a) with a capacity to convert uric acid into urea. This strain lacked 283 urease genes, but three cultured symbionts encoded all genes necessary for urease function, 284 including Cephaloticoccus isolates from C. varians (Cv41) and C. rohweri (Cag34) and a 285 Xanthomonadales symbiont from C. rohweri (Cag60). Thirteen out of fourteen isolates encoded 286 the glutamate dehydrogenase gene (gdhA) converting ammonia into glutamate, and most 287 encoded complete pathways for synthesizing most amino acids. As expected from metagenomic 288 analyses, all genomes lacked nitrogenase genes. 289

290
The fastidious nature of some symbionts limited our ability to infer strain functions for common 291 core taxa. Insights for these groups were gained through draft genome assembly from our best 292 sampled metagenome (i.e. C. varians colony PL010) using the Anvi'o platform (version 1.2.3) 46 293 in conjunction with the CONCOCT differential coverage-based binning program 47 . The 11 near 294 complete draft genomes, where >87% of universal single copy genes were detected, spanned 295 seven of our eight core symbiont taxa ( Fig. 5; Tables S8, S9, S10). Gene content analyses 296 supported findings from metagenomics and cultured isolate genomes (Supplementary Results). 297 In short, the dominant symbiont strains individually encoded up to 17 complete amino acid 298 biosynthesis pathways. Incomplete pathways were often missing just one to two genes. Nearly 299 all draft strain genomes showed capacities to assimilate ammonia into glutamate. N-recycling 300 pathways appeared incomplete within predicted N-recycling Burkholderiales and 301 Cephaloticoccus strains. This suggests the occasional absence of this function from these taxa 302 (but see Fig. 6) or, possibly, incomplete genome assembly due to challenges of scaffold binning 303 To test whether genetic signatures reflect actual N-recycling capacities, and to study the 306 conservation of this role within key taxa, we performed a series of in vitro assays, expanding the 307 number of targeted cephalotine species. Urease activity was qualitatively assessed by the 308 generation of ammonia in the presence of urea, and we obtained positive results for seven of 309 fifteen tested isolates (Fig. 6). All Cephaloticoccus (Opitutales) were positive, as was one of six 310 Xanthomonadales isolates. Results for four isolates with sequenced genomes accurately reflected 311 predictions from the presence/absence of urease genes. 312 313 Production of urea from allantoin served as a proxy for activity of the xanthine/uric acid pathway 314 ( Fig. 6). Urea was produced from allantoin for 11 of the 17 assayed Burkholderiales isolates 315 (Table S11), suggesting function for at least part of this pathway. Coding capacity from the five 316 isolates with sequenced genomes accurately predicted the results of these assays. 317

318
In summary, genomic inferences on N-recycling seem to accurately reflect the metabolism of 319 core symbionts. And importantly, the phylogenetic placement of strains with in vitro assay data 320 reveal sporadic distributions of N-recycling, with notable enrichment in two clades 321 (Cephaloticoccus and a specific, unnamed lineage of Burkholderiales) suggesting long-standing 322 roles in the efficient use of N by the Cephalotes holobiont. 323 324

Discussion
Our findings show that ancient, specialized gut bacteria of Cephalotes ants recycle waste N 326 acquired through the diet (urea) and, likely, through ant waste metabolism (urea and uric acid). 327 Workers acquire large amounts of symbiont-recycled N in the form of essential and non-essential 328 amino acids. Symbionts encode genes to derive their own uric acid and urea, suggesting a third 329 potential origin for the influx of waste N into this system. Across a broad range of Cephalotes 330 species, gene content for N-metabolism varies little within core taxa and N-recycling roles 331 appear conserved within specific symbiont lineages. These findings depict an efficient N-332 economy retained across 46 million years of Cephalotes evolution. They also support the 333 hypothesis that this multi-partite gut microbiome plays an adaptive role within an N-poor dietary 334 niche. 335

336
The magnitude of symbiont contributions to host N-budgets has rarely been calculated. But, 337 findings from wood-feeding termites implicate N-fixing bacteria in providing up to 60% of the N 338 in termite colonies 10 . Measures from the leaf-cutter ant system suggest that N-fixing bacterial 339 symbionts provide 45-61% of the fungus garden's N-supply 48 . Our estimate in C. varians that 340 15-36% of the free amino acid pool was derived from symbiont-recycled N, within five weeks of 341 feeding, was notable, though not directly comparable to either of these estimates. Reduced 342 survival of antibiotic-treated workers, on diets where urea was the only source of N (Fig. S2), do 343 however suggest the importance of symbiont N-metabolism in adults. A similar importance of 344 bacteria was previously suggested for Cephalotes atratus 49 . In carpenter ants Blochmannia have 345 noticeable impacts on worker performance, larval and pupal development, and colony growth; 346 and the detriments of Blochmannia removal can be partially alleviated by the addition of 347 essential amino acids to the diets of aposymbiotic ants 19 . While the impacts of Cephalotes worker microbiomes on larvae have not been measured, adult N-stores are implicated in larval 349 nourishment for several ants 50 . These results suggest a large potential for symbionts of adults to 350 shape performance at all stages within the colony. 351

352
The ubiquity of N-recycling Blochmannia across the Camponotini 19, 51 combine with our 353 findings to support the hypothesized importance of nutritional symbionts in canopy-dwelling, 354 herbivorous ants 14 . A trend of "convergent associations" 52 has, thus, emerged: canopy foraging 355 for N-poor or N-inaccessible foods has evolved separately in association with unrelated, yet 356 functionally similar symbionts. Future work on other ant herbivores and their conserved 357 symbionts 23, 29 will assess the generality of such functional convergence. Also of interest will be 358 studies of symbiont-independent strategies for navigating N-poor diets 7, 53 . 359

360
The conserved nature of symbiont community composition across cephalotines is remarkable 361 compared to patterns for many arthropods 54, 55, 56 , adding to a trend across eusocial insects. 362 Within the termites, for instance, many core symbionts hail from host-specific lineages, 363 revealing ancient, specialized relationships 3 . Among the corbiculate bees, some relationships 364 with gut symbionts date back to 80 million years 57 . But even for these hosts, occasional symbiont 365 turnover takes place-in association with dietary shifts, for termites 58 , and among major 366 phylogenetic divisions in bees 57 . 367 368 Evolved behaviors have likely preserved partner fidelity in these groups. Among eusocial bees, 369 symbiont transfer takes place within the hive, through a combination of trophallaxis, 370 coprophagy, or contact with nest materials 59, 60, 61 . Termite siblings transmit symbionts through 371 oral-anal trophallaxis 62, 63 . A similar mode of passage has been noted for Cephalotes and 372 Procryptocerus ants and for other ant herbivores as well 21,25,26,27 . Of likely further importance 373 for cephalotines is a fine-mesh filter, enveloping the proventriculus, which can bar the passage of 374 particles as small as 0.2 μM beyond the crop. This filter develops shortly after young adults 375 solicit trophallactic symbiont transfers 25 . Symbionts acquired during early adulthood will, thus, 376 be sealed off within the midgut, ileum, and rectum, with minimal opportunities for subsequent 377 colonization by additional, ingested microbes. These dual drivers of partner fidelity 64 may 378 collectively explain the preservation of an ancient nutritional mutualism and sustained 379 exploitation of N-poor foods by successful canopy herbivores. 380 381 382

Materials & Methods 383
Collections and experimental assays 384 Details on ant collections and the uses of ants from particular locales are presented in Fig. 1 and 385 Table S1. For many of these protocols, additional details can be found in the Supplementary 386

Methods. 387
388 Experiments on live ants were performed on Cephalotes varians colony fragments collected 389 from the Florida Keys. Acetylene reduction assays were used to assess the capacity for N-390 fixation. To achieve this, we incubated adult workers (and also, in some instances, queens, 391 larvae, and pupae) in air-tight syringe chambers loaded with acetylene very shortly after 392 collection in the field. After incubation, samples were analyzed with a gas chromatography-393 flame ionization detector to quantify levels of acetylene and ethylene. 394

395
Controlled lab experiments were performed to quantify microbial contributions toward N-396 upgrading of non-essential dietary amino acids and, separately, N-recycling, and subsequent 397 upgrading, of dietary N-waste. Adult workers from each experimental colony (n=3 colonies per 398 each of three experiments) were divided into three groups with equal number. In the first 399 treatment, workers were fed antibiotics to suppress or eliminate their gut bacteria for three 400 weeks. After this time, workers transitioned to the trial period where they were continuously fed 401 antibiotics in addition to a diet of glutamate (with 15 N or 13 C) or a diet with urea containing 15 N. 402 Feeding for this trial period lasted four to five additional weeks. For the second and third 403 treatment groups, bacterial communities were not disrupted. Diets for workers in these groups 404 were identical to those of treatment group one, save for antibiotics, during the three week pre-405 trial period. For the four to five week trial period, workers from the second treatment group were 406 fed on the aforementioned heavy-isotope diets; those from the third group were fed otherwise 407 identical diets in which glutamate or urea consisted of standard isotope ratios (i.e. biased toward 408 lighter isotopes). 409 410 During the trial period we recorded worker mortality, noting an elevation in the 15 N urea feeding 411 group treated with antibiotics, but not in antibiotic-treated workers from the one examined 412 glutamate experiment (Fig. S2). Efficacies of antibiotic treatments were quantified via qPCR on 413 bacterial 16S rRNA genes and, for a subset of specimens, amplicon sequencing of 16S rRNA 414 ( Fig. S1; Table S14). Worker hemolymph was harvested at the end of the four to five week trial, 415 from three to ten surviving workers per replicate. Hemolymph was then pooled, used for amino acid derivitization, and subjected to GC-MS to quantify proportions of free amino acids 417 containing the heavy isotopes (see details in Table S12) from Cephalotes, such as Rhodocyclales), seven genomes from cultured bacterial isolates 435 belonging to core Cephalotes-associated taxa were used to obtain more accurate information of 436 phylogenetic binning. IMG/M-ER was used to annotate gene content from our scaffolds and 437 taxonomic bins. Based on these annotations, we focused on N-metabolism, using KEGG 66 and 438 Metacyc 67 as guides to manually construct degradation pathways for N-waste products and 439 biosynthetic pathways for amino acids. We examined the completeness of the N-waste 440 degradation pathways based on Fig. 6A and  Homologs from N-recycling pathways and 16S rRNA genes were extracted from each dataset 445 and used in phylogenetic analyses with closely related homologs from the NCBI database. To 446 further aid in understanding taxonomic composition and to illustrate depth of coverage for the 447 taxa in our libraries, we generated "blob plots" based on read mapping to classified scaffolds 448 using BWA 68 and modified scripts from a prior publication 69 . These graphs showed the depth of 449 coverage for each scaffold in relation to our classifications, along with the %GC content, a 450 taxonomically conserved genomic signature that further aided us in our efforts to visualize the 451 diversity of symbionts within microbiomes (Fig. 3 Contents were then plated on tryptic soy agar plates, and plates were incubated at 25°C under an 465 atmosphere of normal air supplemented with 1% carbon dioxide in a CO2-controlled water-466 jacketed incubator. After colony sub-cloning, pure isolate cultures were maintained under these 467 same conditions on the aforementioned plates or in tryptic soy broth. DNA extracted from these 468 cultures was subsequently used for 16S rRNA PCRs to compare isolates to bacteria previously 469 sampled through culture-independent studies. Isolates from C. varians and C. rohweri (both 470 previously well-studied through culture-independent means) were prioritized for genome 471 sequencing when their 16S rRNA sequences were identical or nearly identical to those of from 472 prior in vivo studies. Extracted bacterial DNA was used for library preparation and Illumina or 473 PacBio SMRT sequencing. Assembled genomes were uploaded to IMG/ER for annotation, with 474 N-metabolism pathway reconstruction and extraction of genes for phylogenetics occurring as 475 described above. Alignments of isolate genomes to metagenomes were done with Icarus 70 as 476 implemented in MetaQuast 71 and visualized in Circos 72 477 478 A subset of cultured isolates was subjected to assays to detect ammonia production from urea. 479 Several were also tested to determine whether allantoin, a derivative of uric acid breakdown, 480 could be used to synthesize urea. Methodological details on these assays are described in the 481 Supplementary Methods. As described above for genome sequencing prioritization, strains 482 prioritized for assays were those deemed highly related to specialized core    were inferred from 14 fully sequenced cultured isolate genomes (symbionts from C. varians and C. rohweri) and 11 draft genomes (assembled from C. varians colony PL010 metagenome; identified by the term "Bin" within their names). The maximum likelihood phylogeny of symbiotic bacteria on the left was inferred using an alignment of amino acids encoded by seven phylogenetic marker genes obtained from symbiont genomes, and branch colors are used to illustrate distinct bacterial orders. Red asterisk for urea recycling in the Cephaloticoccus-like Opitutales bin (7-1) indicates that urease genes from the PL010 metagenome binned to Opitutales, but not to the draft genome for the dominant strain. When combined with the likely presence of just one Opitutales strain within the PL010 microbiome, it is likely that a completely assembled genome would encode all urease genes. The black asterisk next to Arg denotes that inferences on pathway completeness for arginine biosynthesis were based on the pathway starting with glutamate (see Fig. S9), as opposed to other metabolic mechanisms. In vitro assayed N-recycling pathways:

≥80% bootstrap support
Core Cephalotini bacteria from culture independent efforts Xan puuD xdhABC urea production assay 1 kb urea degradation assay ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ Urea is produced without allantoin but allantoin boosts urea production. in the NCBI database, which are enclosed within taxon-specific colored boxes. Most cultured isolates had 16S rRNA sequences that were identical or most closely related to one obtained from a Cephalotes ant through culture-independent means. Such isolates also showed high identity to abundantly represented scaffolds from our metagenomes (Fig. S15). Nodes for cultured symbionts are connected to relevant rows within data tables, where the results of assays for urea production (from allantoin-part of the uric acid pathway) and urea degradation (to ammonia) assays are illustrated. Asterisks highlight isolates with a sequenced genome; for each of these, in vitro results matched expectations derived from gene content. Additional symbols used in the urea production table indicate whether allantoin boosted urea production and whether production was completely allantoin-dependent. Urea production without complete allanotindependence is likely to stem from arginine metabolism.