Transcriptome landscape of a thermal-tolerant coral endosymbiont reveals molecular signatures of symbiosis and dysbiosis

Background Reef-building corals depend upon a nutritional endosymbiosis with photosynthetic dinoflagellates (family Symbiodiniaceae) for the vast majority of their energetic needs. While this mutualistic relationship is impacted by numerous stressors, heat stress from warming oceans is a predominant threat to coral reefs, placing the future of the world’s reefs in peril due to climate change. While some Symbiodiniaceae species exhibit tolerance to thermal stress, the molecular basis for this tolerance is largely unexplored. To identify the underpinnings of heat tolerance and symbiosis, we compared the in hospite and free-living transcriptomes of Durusdinium trenchii, a pan-tropical heat-tolerant Symbiodiniaceae, under stable temperature conditions and acute hyperthermal stress. Results We discovered that under stable conditions, in hospite cells exhibited lower transcriptional activity than free-living counterparts, suggesting the shutdown of genes uniquely required for a free-living lifestyle in a process we term transcriptome reduction. However, under hyperthermal stress the transcriptional response was larger in hospite than in the free-living state, indicating an exacerbated stress environment within the host cell. While we identified a core heat stress response shared between free-living and in hospite D. trenchii, this represents only a minority of the stress-induced transcriptional activity. Our findings also revealed distinct processes at work in free-living and in hospite cells under non-stress conditions, with free-living D. trenchii exhibiting evidence of sexual reproduction. We also found that in hospite cells display putative host evasion activity including components of apicomplexan parasite invasion machinery. Conclusions In this work, we unraveled the molecular signatures of symbiont heat tolerance within the host, which is a critical step to enable the development of engineered endosymbionts as a tool for restoration of coral reefs. Our findings also highlight the necessity of conducting experiments on Symbiodiniaceae in the appropriate symbiotic state, as the stress responses of free-living cultures may not be representative of Symbiodiniaceae within a holobiont. We suggest that the necessity of an intermittent free-living state, outside of its highly co-evolved symbiosis, prevents genome reduction as seen in obligate endosymbionts. We posit that transcriptomic reduction of non-essential genes in hospite may be an alternative adaptive strategy for endosymbionts in ancient symbiotic assemblages.

, and while 171 they have been insightful, they may present a reduced transcriptional picture of the 172 alterations that occur in Symbiodiniaceae in hospite. Only one other investigation 173 directly compared gene expression in hospite to free-living conditions (Rosic et al., 174 2010), but it focused on only two heat-shock protein genes. We propose that it is 175 imperative to consider the in hospite environment in the assessment of 176 Symbiodiniaceae response to thermal stress, and that culture-based studies require 177 careful consideration when extending findings to in hospite conditions. 178

A Small Core Thermal Stress Response is Shared Between Free-living and in 179
hospite D. trenchii 180 181 Within the conserved transcriptional changes under thermal stress shared 182 between the free-living and in hospite states of D. trenchii (Fig. 3, Fig. 4), 57 transcripts 183 were upregulated and 28 downregulated. Examination at the transcript level reveals that 184 the magnitude of expression within these core transcripts is, overall, higher in free-living 185 D. trenchii than in hospite (Fig. 4). With a relatively small number of transcripts 186 representing the core response, we were not able to perform meaningful GO term 187 enrichment analysis. Instead, our interpretation of this small conserved core response is 188 based upon transcript annotations. 189 Though the core set of transcripts differentially expressed in response to thermal 190 stress ( Fig. 3) represents the minority of transcriptional alterations under stress in 191 hospite and in free-living cells, it sheds light on key processes conserved in the D. 192 trenchii stress response independent of symbiotic state. These processes include 9 hospite and free-living stress responses, these associated transcripts are potentially 196 important candidates in the identification of the genetic basis for thermal tolerance in D. 197 trenchii. 198 Maintenance of membrane viscosity is critical for cell homeostasis as 199 hyperthermal stress increases the fluidity of membranes, potentially leading to 200 disintegration of the lipid bilayer (Allakhverdiev et al., 2008). In Symbiodiniaceae, 201 thylakoid membrane lipid composition is a key determinant of thermal-stress sensitivity 202 (Tchernov et al., 2004). Within the core heat stress response, we identify an ortholog to 203 a fatty acyl-CoA desaturase, an enzyme involved in the remodeling of cell membranes 204 for homeoviscous adaptation (Hazel, 1995) that is potentially critical in modifying fluidity 205 to accommodate increasing temperatures. 206 The upregulation of apurinic endonuclease-redox protein (ARP) both in hospite 207 and in free-living conditions suggests that repair to oxidative DNA base damage occur 208 in both conditions. The apurinic/apyrimidinic (AP) endonucleases are important DNA 209 repair enzymes involved in two overlapping pathways: DNA glycosylase-initiated base 210 excision and AP endonuclease-initiated nucleotide incision repair (Akishev et al., 2016). 211 In the context of heat stress, this may allow D. trenchii a faster response to DNA 212 damage than other Symbiodiniaceae species. 213 Anion channels, spermidine synthase, and retrotransposons are disparate 214 entities which can each serve as heat stress response mediators with many 215 downstream effectors. In plants, anion channels are a master regulator of stress 216 response (reviewed by Rob et al., 2012). In the core heat stress response, we find the 217 upregulation of a transcript putatively coding for aluminum-activated malate transporter 218 9 (ALMT9). In plants, ALMT9 is involved in the release of chloride currents in guard 219 cells and malate homoeostasis. In response to stress, the release of chloride and 220 malate leads to the rapid stomatal closure (Dreyer et al., 2012). Though the precise role 221 of ALMT9 in stress response in Symbiodiniaceae species is unknown, it may be 222 involved in downstream stress response signaling. A transcript coding for spermidine 223 synthase is also upregulated across heat stress treatments. Spermidine synthase in 224 Arabidopsis dramatically improves stress tolerance when overexpressed via its effect on 225 multiple downstream targets (Kasukabe et al., 2004), while exogenous application of the 226 protein increases antioxidant tolerance in creeping bentgrass (Li et al., 2015). An 227 additional transcript of the core response potentially acting as an upstream effector of 228 numerous targets codes for LINE-1 reverse transcriptase. Recent work by Chen et al. 229 (2018) identified Ty1-copia-type LTR retrotransposons in Symbiodinium microadriaticum 230 which are activated by heat stress, suggesting that this may be a mechanism for the 231 generation of genetic variation under heat stress. Diatoms also possess stress-activated 232 retrotransposons, which are postulated to serves as generators of genetic variability 233 (Maumus et al., 2009). The involvement of retrotransposons in stress response has also 234 been observed across taxa. In plants, retrotransposons are activated by stress and go 235 on to activate numerous host regulatory pathways (Grandbastien, 2015(Grandbastien, , 1998. 236 D. trenchii response to heat stress depends on the symbiotic state 237 238 The majority (88% in hospite and 76% free-living) of the transcriptional response 239 to thermal stress is specific to the symbiotic state of the dinoflagellate (Fig. 3B). Both 240 free-living and in hospite D. trenchii have distinct responses to a thermal challenge of 241 34°C not only at the level of transcript identity (Fig. 3), but also with regard to enriched 242 gene ontology terms which are unique to each symbiotic state (Fig. 5). In hospite D. 243 trenchii have transcriptional activity with parent GO terms such as mRNA transport, N-244 terminal protein amino acid acetylation, and cellular response to temperature stimulus, 245 all of which are not enriched in the free-living transcriptomic response to heat stress 246 (Fig. 5). There is also a striking in hospite-specific decrease in GO terms within the 247 category "binding of sperm to zona pellucida" under heat stress. In the case of the free-248 living state, negative regulation of viral genome replication is highly downregulated in 249 contrast to the symbiotic state (Fig. 5). While the analysis reveals that the majority of 250 ontologies affected by heat stress are disparate depending upon symbiotic state, 251 decreased activity of vesicle-mediated transport is a common theme in the heat stress 252 response of both in hospite and free-living D. trenchii (Fig. 5). Despite this particular 253 instance of overlap in function indicated by gene ontology analysis, the transcripts 254 underlying vesicle-mediated transport in free-living and symbiotic treatments are distinct 255 in the symbiotic and free-living conditions (Supplement_2_DEGs_Annotations.xlsx). 256 Remarkably, despite the more highly responsive transcriptome of in hospite D. 257 trenchii to heat stress than free-living cells, free-living D. trenchii mount a more dramatic 258 response to photoxidative stress than in hospite at both 28°C and 34°C (Fig. 6). Gene 259 ontology terms for thermoception and response to oxidative stress are also highly 260 represented in free-living D. trenchii at 34°C, whereas no enrichment is detected in 261 hospite (Fig. 6). This suggests that the intracellular environment of the host provides 262 some level of protection for the endosymbiont from some aspects of photothermal 263 stress, potentially by a modified light environment within the host. This is supported by 264 earlier work by Bhagooli and collaborators (Bhagooli et al., 2008), who found that 12 multiple genera of Symbiodiniaceae exposed to photothermal stress in hospite had 266 unaffected photosynthetic output and less impact to maximum quantum yield in 267 comparison to cultured cells across genera. Coral hosts possess multiple mechanisms 268 for sheltering endosymbionts including photoprotective host chromoproteins (Dove et 269 al., 2001;Quick et al., 2018;Salih et al., n.d.;Smith et al., 2013) and other host-270 mediated modifications to the internal light environment, such as scattering by the coral 271 skeleton (Enríquez et al., 2005;Wangpraseurt et al., 2012). While this sheltering in 272 hospite may, on its surface, appear contradictory, we posit that the in hospite state has 273 a more complex transcriptional response to thermal stress overall, but that the host may 274 reduce light stress. 275 By examining ontologies from differentially expressed transcripts exclusively 276 associated with free-living and in hospite treatments, we found that the in hospite heat 277 stress response is highly enriched for mRNA transport and cellular response to 278 temperature stimulus (Fig. 5). We suggest that host sheltering of in hospite cells light 279 may contribute to a symbiont's capacity to maintain some cellular functions which may 280 be heavily impacted in unsheltered free-living cells. This may result in the mounting of a 281 more effective transcriptional response to heat stress. Under acute thermal stress, yeast 282 block mRNA transport of transcripts uninvolved in the heat stress response (Saavedra 283 et al., 1996). If unsheltered free-living cells are under more severe stress, experiencing 284 the unfiltered brunt of light in combination with heat stress, mRNA transport may be 285 highly inhibited, as suggested by our analyses. 286 The in hospite state also exhibits a dramatic reduction in GO term enrichment 287 within the biological process binding of sperm to zona pellucida, a response not 13 observed in the free-living comparison (Fig. 5). One of the transcripts underlying this 289 enrichment (TRINITY_DN30024_c0_g1_i6) has high identity to sperm surface protein 290 Sp17, a protein possessed by sperm which binds the zona pellucida during mammalian 291 fertilization, critical for cell adhesion (O'rand et al., 1995;Richardson et al., 1994). The 292 expression of the putative Sp17 ortholog may function in the mediation of normal 293 symbiosis, potentially involved in adhesion between the symbiont and host, while the 294 observed decrease in expression may be a part of dysbiosis. 295 Taken together, the intracellular environment experienced by in hospite 296 symbionts represents a complex set of interactions between the host, the external 297 environment, and the endosymbiont. The disparate transcriptomes of stress in hospite 298 and free-living likely result from a combination of protection from some aspects of stress 299 by the host, the complexity of an intracellular environment compared with that of a free-300 living cell, and further, the transcriptional activity involved in the mediation of the 301 symbiosis, as discussed in sections that follow. 302

Alterations required for the transition between free-living and symbiotic states 303
Comparing the in hospite and free-living transcriptomes of D. trenchii provides a 304 window into the alterations necessary for symbiotic living. Several processes are 305 exclusively enriched in hospite under non-stress conditions at 28°C, in comparison to 306 the free-living state (28°C symbiotic vs. 28°C free-living) (Fig. 6). Late endosome to 307 vacuole transport and mRNA cis splicing via spliceosome are both highly enriched in 308 hospite along with response to prostaglandin F. The roles of increased lysosome activity 309 and enhanced cis splicing in symbiosis are unknown, but complexities in 14 canonical intron splicing (Aranda et al., 2016;Lin, 2011;Mendez et al., 2015). The 312 enrichment of response to prostaglandin F is extremely curious as the highest 313 concentrations of prostaglandins are found in corals, where these compounds are 314 produced by the host (Valmsen et al., 2001;Weinheimer and Spraggins, 1969). 315 Hydrogen peroxide catabolism is enriched in hospite at both 28°C and 34°C in 316 comparison to the free-living state, with higher activity at 34°C. Increased scavenging of 317 hydrogen peroxide in hospite could potentially be adaptive and necessary for 318 maintaining stable symbiosis, as while the release of peroxide in the free-living state is 319 likely of minimal detriment, the release of this form of ROS in hospite is held to induce 320 an immune response on the part of the host (reviewed by Weis, 2008). 321 In addition to the aforementioned enrichment of photooxidative stress terms in 322 the free-living state under thermal challenge (34°C), the transcripts upregulated in the 323 free-living state at 28°C are also enriched for numerous terms associated with response 324 to photooxidative stress. While this is amongst the highest activity of parent GO terms in 325 the free-living condition at both thermal treatments, no increase in enrichment is found 326 in the in hospite condition at either 28°C or 34°C. Dipeptide transport as well as 327 heterochromatin organization are also highly enriched in the free-living condition at 328 28°C. 329

D. trenchii exhibits transcriptome reduction in hospite 330
Genome reduction is a common phenomenon in obligate microbial suggest that chromatin remodeling may play an important role in the switch between the 358 free-living state and endosymbiosis, contributing to the substantial downregulation 359 associated with the symbiotic life phase. We consider this resulting transcriptome 360 reduction analogous to the genome reduction observed in numerous obligate symbiotic 361

systems. 362
The potential for sex in free-living Symbiodiniaceae 363 In many species, sex and reproduction are inextricably linked, but for some, sex 364 represents a special occasion. For instance, consider a member of a taxon sister to the 365 dinoflagellates, the apicomplexan Plasmodium falciparum. This malaria-causing 366 parasite only engages in sexual reproduction within its mosquito host (Kooij and 367 Matuschewski, 2007;Talman et al., 2004). Be it frequent or rare, the function of sex, 368 despite its apparent ubiquity, is not a settled case. One hypothesis regarding the 369 evolution of sex is that recombination mediated by sexual reproduction arose as a 370 mechanism to repair DNA damage in an individual (Bernstein et al., 1985). A competing 371 hypothesis holds that the advantage of meiotic recombination resides in the genetic 372 diversity afforded to offspring (Burt, 2000;Otto, 2009). Recombination through sexual reproduction may be critical to the maintenance of 395 Symbiodiniaceae genomes. Muller's Ratchet dictates that asexual reproduction without 396 other forms of recombination has the potential for the accumulation of deleterious 397 mutations over generations (Felsenstein, 1974;Muller, 1964Muller, , 1932. This effect has 398 support from RNA viruses lacking recombination (Chao, 1997) and may be one of the 399 driving forces behind genome reduction in obligate endosymbionts (Andersson and 400 Kurland, 1998), but obligate symbionts may have other mechanisms to avert genomic 401 meltdown. Selection at the level of the holobiont has been suggested as one 402 mechanism of endosymbiont genome maintenance (Rispe and Moran, 2000), with genomic degradation and losses observed likely to be neutral or nearly neutral 404 (Pettersson and Berg, 2007). We posit that recombination in the free-living phase of 405 Symbiodiniaceae may be instrumental in genomic maintenance of functions necessary 406 for the free-living state. 407

Evasion of host immunity by symbiont-derived factors 408
In order to maintain a stable endosymbiotic existence, members of the family 409 Symbiodiniaceae must evade the host immune system. A considerable number of 410 studies have shown that Symbiodiniaceae suppress and/or circumvent host immunity in 411 order to colonize and persist in hospite; multiple mechanisms of circumvention appear 412 to occur, including the suppression of apoptosis (Matthews et al., 2017;Rodriguez-413 Lanetty et al., 2006;Tchernov et al., 2011) as well as the manipulation of markers of 414 endosome age (Chen et al., 2005(Chen et al., , 2004(Chen et al., , 2003. More recent work has identified 415 endosymbiont miRNAs which target host transcripts involved in immunity (Baumgarten 416 et al., 2018(Baumgarten 416 et al., , 2013. We identified numerous transcripts involved in the suppression of 417 innate immunity and anti-inflammatory pathways, including a putative NRCL3 ortholog 418 and NF-κB inhibitor, which are upregulated in the symbiotic state. However, in order for 419 this to plausibly occur, the symbiont must be able transmit these signals to the host. 420 Based on transcriptome evidence, we postulate mechanisms for this below. with the longstanding support for glutamine synthetase activity as the mechanism by 465 which Symbiodinacaea assimilate ammonium from seawater and re-assimilate waste 466 ammonium from the cnidarian host (Anderson and Burris, 1987). 467 In the in hospite state, five transcripts annotated as ABC transporters are 468 upregulated in comparison to the free-living state at 28°C (Table S5) substrates, including ions, nutrients, amino acids, peptides, proteins, lipids, metals, 475 lipids, oligonucleotides, and sugars (Locher, 2016;Wilkens, 2015). 476

Conclusion 477
In summary, our results demonstrate that symbiotic state has major impacts on 478 hospite cells displaying reduced differential expression in a process we term 491 transcriptome reduction, as well as putative host evasion activity. We posit that the 492 observed transcriptomic reduction of non-essential genes in hospite may be an 493 alternative adaptive strategy to genome reduction for endosymbionts that must also 494 maintain the capacity for a free-living life phase. We intend for this to serves as the 495 basis for future gene editing efforts to understand the genetic basis of symbiosis and 496 dysbiosis, with the ultimate aim of engineering of stress-tolerant coral endosymbionts. 497 These efforts will for provide additional options in future reef protection and restoration 498 efforts desperately needed for the persistence of coral reefs into the 21 st century. 499 500

501
Aposymbiotic Exaiptasia pallida (strain CC7) were inoculated with D. trenchii 502 (strain CCMP2556) and maintained with feeding of Artemia franciscana twice weekly for 503 12 months. Free-living D. trenchii were cultured in Prov50 (Smith and Chanley, 1975) 504 medium and maintained in log phase via regular passage. To identify thermal 505 treatments that elicited an acute heat stress response, Durusdinium trenchii (strain 506 CCMP2556) cultures were exposed to thermal stress and assessed with PAM 507 fluorometry (Fig. S1). Based upon the observed rapid decline in Fv/Fm at 34 °C in 508 contrast to the maintained photosynthetic efficiency at 28 °C (Fig. S1), these treatments 509 were selected and implemented for comparisons following 6 hours at the target 510 temperatures. Symbiotic anemones and free-living D. trenchii cultures were held at 511 28°C under light intensity levels of 80 µE (12:12 light:dark cycle) and ramped up to the 512 34°C challenge at a rate of 2°C per day for thermal stress treatments, with increases of 513 temperature performed at the beginning of the photoperiod. Free-living (cultured) D. 514 trenchii and in hospite D. trenchii were frozen in liquid nitrogen mid-photoperiod for 515 sampling. Heat stress treatments were exposed to 34°C for a total of six hours. The

Symbiosis state comparison and thermal challenge 608
To inform the design of this experiment, maximum quantum yield in cultured D. 609 trenchii was assessed through time, at temperatures from 28 to 36° C (n=3) using a 610 PAM fluorometer (Walz, Germany). We found that the rapid decline in maximum 611 quantum yield began at 34°C at 2 hours of thermal stress. In an 8 hour timespan, we did 612 not find significant decline at the lower thermal treatments.  filtered using all bacterial, archaeal, and viral genomes present in RefSeq (as of Sep. 8, 652 2017) using Kraken, an ultra-fast taxonomic sequence classification program (Wood 653 and Salzberg, 2014). The 526,545,945 remaining PE reads were then assembled with 654 Trinity v2.4.2 (Grabherr et al., 2011) with a minimum contig length of 200 bp. 655 Redundant contigs were removed using uCLUST (Edgar, 2010)  For annotation, the assembly was searched against the NR protein database 665 using DIAMOND, a fast search algorithm that employs double-indexing and spaced 666 seeds of various weights and shape to make it ~20,000x faster than BLASTX with 667 comparable level of sensitivity and accuracy (Buchfink et al., 2015). These results were 668 input into with Trinotate using BLASTP and BLASTX searches against the SwissProt 669 and Uniref90 databases with an e-value cutoff of < 1 x 10 -5 . Open reading frames 670 (ORFs) were predicted using Transdecoder under the default settings. Gene ontology 671 classifications were extracted from search results against the SwissProt database. The 672 searches were conducted using DIAMOND, a fast search algorithm that employs 673 double-indexing and spaced seeds of various weights and shape to make it ~20,000x 35 faster than BLAST with a similar sensitivity and accuracy (Buchfink et al., 2015). 675 Additional annotation was performed using Blast2GO (Conesa et al., 2005) and viewed using the web-based REVIGO program (Supek et al., 2011). Supplementary Figure 14. REVIGO