The root of the transmissible cancer: first description of a widespread Mytilus trossulus-derived cancer lineage in M. trossulus

Two lineages of bivalve transmissible neoplasia (BTN), BTN1 and BTN2, are known in blue mussels Mytilus. Both lineages derive from the Pacific mussel M. trossulus and are identified primarily by the unique genotypes of the nuclear gene EF1α. BTN1 is found in populations of M. trossulus from the Northeast Pacific, while BTN2 has been detected in populations of other Mytilus species worldwide but not in M. trossulus itself. The aim of our study was to examine mussels M. trossulus from the Sea of Japan (Northwest Pacific) for the presence of BTN. Using hemocytology and flow cytometry of the hemolymph, we confirmed disseminated neoplasia in our specimens. Cancerous mussels possessed the unique BTN2 EF1α genotype and two mitochondrial haplotypes with different recombinant control regions, similar to that of common BTN2 lineages. This is the first report of BTN2 in its original host species M. trossulus populations in West Pacific may be the birthplace of BTN2 and a natural reservoir where it is maintained and whence it spreads worldwide. A comparison of all available BTN and M. trossulus COI sequences suggests a common and recent, though presumably prehistoric origin of BTN2 diversity in populations of M. trossulus outside the Northeast Pacific.


29
Clonally transmissible cancer (CTC) is a neoplastic disease passed from individual to individual 30 by physical transfer of cancer cells 1-3 . The first inkling of a transmissible cancer came from a study 31 of canine transmissible venereal tumor, CTVT, dating back to 1876 4 . Since then CTC has been 32 confirmed for CTVT 5  The finding that CTC is the cause of DN in six different bivalve species is fairly recent, and it is 42 reasonable to anticipate that further discoveries will turn up. There are also other lines of indirect 43 evidence pointing to a widespread occurrence of transmissible cancers in bivalves 13 . At the same 44 time, the data are not sufficient to ensure that CTC is the only cause of DN in those six species, or 45 whether CTC is the usual cause of DN in bivalves in general. 46 DN is a fatal leukemia-like cancer affecting many marine bivalves 14 . It was first described in 47 recorded anywhere before and whether they demonstrate an affinity to particular mitochondrial 129 lineages of the host. Finally, we verified the purebred M. trossulus ancestry of mussels from  infected population by genotyping them by three additional taxonomically diagnostic markers 131 ("Species confirmation") and identified their sex histologically and/or genetically ("Sex 132 identification"). 133 134 Sample collection and preprocessing 135 Mussels were collected by scuba diving at three localities of the SOJ in July-September 2019: 136 Vladivostok city public beach "Vtoraya Rechka" (43°10'02"N, 131°58'07''E, depth 5 m, a natural 137 bottom habitat, sample size N=39, mean shell size L=25 mm, sample "R"), "Vostok" Marine 138 Biological Station (42°89'35''N, 132°73'33''E, depth 5 m, experimental mussel plantation, N=20, 139 L=50 mm, sample "V") and the Gaydamak Bay (42°52'04"N, 132°41'27"E, fouling of a mooring 140 buoy anchored at a depth about 5 m about 50 m from the shore, N=226, L=37 mm, sample "J"). 141 "Vostok" station is a nature reserve and the least polluted of the studied localities, while Gaydamak 142 is an industrial harbor heavily polluted by sewage. After sampling, the mussels were transported 143 to the laboratory and stored for 2-4 days alive, each sample in a separate aquarium, before the 144 experiments. used to estimate ploidy levels of aneuploid peaks relative to the diploid peak of the same specimen 159 and the rate of aneuploid cells in the sample. Mussels were diagnosed as healthy ("DN-rejected") 160 if the scatter plot SSC-A vs FSC-A and PB450-A histogram of fluorescence indicated only non-161 proliferating cells (agranulocytes and granulocytes) with one peak for diploid phase or with 162 admixture of very few (<5%) proliferating cells (minor peak for tetraploid cells). If a cell 163 population with additional peaks was detected, the individuals were considered as "DN-164 suggested". 165

Hemocytology and immunochemistry
166 All procedures for cell fixation and staining were described in detail in a previous study 41 . In brief, 167 the cells were stained with TRITC-labeled phalloidin (Molecular Probes) and DAPI (Vector 168 Laboratories) for visualization of actin cytoskeleton and nucleus. We used primary mouse 169 monoclonal antibodies against anti-α-acetylated tubulin (clone 6-11B-1, Sigma Aldrich and 170 proliferating cell nuclear antigen (PCNA, clone PC10, Abcam) for detecting mitotic spindles and 171 proliferating cells, respectively. These antibodies were previously characterized as labeling 172 dividing cells in bivalves [41][42][43][44]  EF1α PCR products from the hemolymph and the foot tissues of the cancerous and the control 197 mussels were subjected to molecular cloning. CR PCR products were cloned from the hemolymph 198 of the cancerous mussels only. Molecular cloning procedures were subcontracted to Evrogen JSC 199 (Russia). Quick-TA kit (Evrogen JSC) was used for cloning, and the plasmids were transformed 200 into competent E. coli (Evrogen JSC). In all the cases at least 16 colonies were sequenced using 201 M13 primers. Some sequences were detected only in one colony. They were probably artificial 202 mutations generated by PCR and molecular cloning procedures such as polymerase errors and 203 random crossing-over of incomplete PCR extension products of original alleles 48,49 and therefore 204 were excluded from the following analyses. 205 In COI, signals from multiple alleles were resolved by sequencing only. If "piggybacks", that is, 206 overlapping peaks at some positions, were observed on chromatograms, the major peaks were 207 attributed to the presumable cancer allele in the hemolymph samples and to the presumable host 208 allele in the foot samples. 209 Sequence chromatograms were analyzed in MEGA X 47 . Sequences were aligned with MUSCLE 210 algorithm with some manual adjustment. 211 8 Phylogenetic analysis 212 Nucleotide sequences of CR, COI and EF1α from four cancerous mussels were aligned together 213 with the corresponding sequences from a previous BTN study (data on 11 mussels with confirmed 214 BTN) 10 and mitochondrial sequences of the Baltic mussel 62mc10 37 . The 62mc10 genome was 215 previously shown to be similar to cancerous ones 10 . Alignments by MUSCLE and maximum 216 likelihood phylogenetic trees were generated using MEGA X 47 , with 100 bootstrap replicates, 217 treating gaps in the alignment as missing data. The same substitution models as in Yonemitsu et 218 al. 10 were used for tree generation. The trees were visualized using iTOL tool 50

231
According to a recent survey 60 , the blue mussel populations in the SOJ are overwhelmingly 232 dominated by M. trossulus, but the presence of a "cryptic" species, M. galloprovincialis and its 233 hybrids with M. trossulus cannot be entirely ruled out. Mussels from the "J" sample (N=21), 234 including four target mussels with DN, were genotyped for three nuclear markers routinely used 235 for discriminating M. trossulus and M. galloprovincialis 61 , ME15/16 62 , ITS 63 and MAL-I 64 , using 236 the primers and protocols in the original articles. The DNA extracted from the hemolymph and 237 from the mantle of each mussel were analyzed in parallel. The mussels were also sexed, at first by 238 a microscopic examination of fresh tissues of the mantle, where the gonads in blue mussels are 239 partly localized. It turned out, however, that most mussels were in post-spawning condition and 240 lacked gametes. Therefore we identified their "mitochondrial" sex by the presence or the absence 241 of M-mtDNA 16S haplotypes, following the approach of Rawson and Hilbish 65 and using DNA 242 extracted from the mantle. 243

DN diagnostics 245
Flow cytometry 246 Flow cytometry of hemolymph revealed two distinct patterns (Fig. 2). Most of the individuals had 247 one peak of hemocytes, interpreted as diploid (Fig. 2a), or a diploid peak with a small admixture 248 of tetraploids (Fig. 2b). These mussels were diagnosed as healthy. The second pattern was revealed 249 in 10 "J" individuals (4% of "J" sample), which had an additional population of aneuploid cells 250 ( Fig. 2c-f). Their ploidy, calculated relatively to the 2n peak, varied among individuals from 3.7n 251 to 5.2n. The proportion of these cells in the hemolymph of different mussels varied from 12 to 252 98%. These mussels were classified as "DN-suggested". We did not reveal any significant signal 253 from proliferating neoplastic cells, which could be expected as a third peak at the right side of the 254 aneuploid peak in the histogram. Two "DN suggested" mussels with a moderate proportion of 255 neoplastic hemocytes, J54 (proportion 44.4%) and J111 (26.4%), and two mussels with a high 256 proportion of neoplastic hemocytes, J161 (91%) and J181 (80%) (Fig. 2c- Hemocytological study revealed striking differences between the control mussels on the one hand 261 and all the four DN-suggested mussels on the other hand. DN-suggested individuals had, in 262 addition to normal adherent hemocytes with a low nucleus-to-cytoplasm ratio, also anomalous 263 round non-adherent hemocytes. They looked like hedgehogs on the slides, the resemblance being 264 due to an altered cytoskeleton with prominent actin "spikes". Their nuclei were pleomorphic, 265 larger than those of normal hemocytes (Fig. 3). We considered these cells as neoplastic and the 266 mussels they belonged to as DN-confirmed. 267 PCNA staining revealed few neoplastic hemocytes at the DNA synthesis stage (Supplementary 268 1) from the hemolymph and from the foot samples were the same. In contrast, in all the four DN-275 confirmed mussels the 146 bp fragment was overrepresented in the hemolymph samples in 276 comparison with the foot samples ( Fig. 1, Supplementary Fig. S2). At the same time, this fragment, 277 as well as the 144 bp fragment, characteristic of all the cancerous mussels, was also recorded in 278 some healthy mussels, making the results inconclusive.  Hereafter the alleles identified in our study will be designated by letters if initially recorded in the 283 study of Yonemitsu et al. 10 , and by numerals if newly found, except in a few cases when we wanted 284 to emphasize the similarity between the alleles (e.g. EF1α-G1, CR-1', see below). It should be 285 noted, however, that the alignments in Yonemitsu et al. 10 were slightly longer than here. 286 The sequencing and molecular cloning of EF1α PCR products revealed no differences between 287 sequences from the hemolymph and from the foot tissues of the control individuals 288 (Supplementary Table S2 and Fig. 1). In contrast, all EF1α sequencing chromatograms from the 289 hemolymph and the foot samples of DN-confirmed mussels showed a mixed signal. Molecular 290 cloning of these PCR products revealed 2-4 different sequences in the foot samples and 2-6 in the 291 hemolymph samples (Fig. 1, Supplementary Fig. S3). Some sequences were common, that is, 292 represented in relatively many colonies, and differed by more than one substitution, while others 293 were rare. All the mussels had two common sequences, EF1α-G and EF1α-H. They differed by 21 294 substitutions, which made them the most dissimilar common sequences, and were always more 295 frequent in the hemolymph than in the foot samples. Other common sequences were usually 296 specific of individual mussels and were present in both tissues (Fig. 1, Supplementary Table S2  297 and Supplementary Fig. S3). Putatively, EF1α-G and EF1α-H represented a heterozygous cancer 298 genotype. Other common sequences probably represented the diploid host genomes. Rare 299 sequences, including EF1α-G1 different from EF1α-G by one substitution (Supplementary Fig.  300 S3), were probably methodological artifacts. 301 Direct sequencing of COI and CR revealed identical alleles in the hemolymph and in the foot 302 samples of individual control mussels, in homoplasmic condition (J17 and J38, Supplementary 303 Table S2). On the contrary, the chromatograms from analyses of different tissues of mussels with 304 DN looked very different. 305 No heteroplasmy was observed in the CR chromatograms of cancerous mussels. In foot samples, 306 unique alleles were identified. In the hemolymph samples, two alleles were identified: CR-1 in J54 307 and J161 and CR-2 in J111 and J181. CR-1 and CR-2 were very different both from each other 308 (31 substitutions, 5.0% difference) and from the other alleles (2.7-6.6%), considering that the 309 differences between all the other alleles were within the range of 0.3-1.5%. Molecular cloning 310 confirmed the results of direct sequencing, but also revealed additional rarer sequences invisible 311 on sequencing chromatograms (Supplementary Table S2, Fig. 1). In the hemolymph of J111 and 312 J181 the same alleles as in foot were found, while in the hemolymph of J161 an allele CR-1', 313 supported by five clones, was found, differing from the major CR-1 allele by four substitutions. 314 13 No COI heteroplasmy was observed in the foot samples of cancerous mussels while several 315 positions with overlapping peaks of a very different height were observed on chromatograms of 316 the hemolymph samples (cf. Fig. 1). The heteroplasmy was readily identified as representing 317 presence of the "foot allele" of the same individual in minority (lower peaks) in addition to a 318 dominant hemolymph allele. Two major COI sequences were identified in the hemolymph 319 samples: COI-1 in J54 and J161 and COI-2 in J111 and J181. These alleles differed from each 320 other by 6 substitutions (0.95%) and by 0.60-0.95% from all the other alleles. 321 Thus, the sequence analyses revealed genetic chimerism of mussels with DN, with the hemolymph 322 and the foot tissues being dominated by different genotypes of both the nucleus and mitochondrion. 323 However, while EF1α genotyping revealed the same cancer genotype in all the diseased mussels, 324 mtDNA genotyping revealed two different cancer genotypes, marked by different combinations 325 of COI and CR alleles, in different mussels. The conclusive evidence that DN in mussels from the 326 SOJ is BTN came from a genetic comparison with cancers from previous studies. 327

Maximum likelihood trees 329
A general inspection of phylogenetic trees (Fig. 4) shows that the SOJ mussels are infected with 330 the BTN2 cancer lineage. The EF1α-G and EF1α-H alleles identified in these mussels are 331 previously known BTN2-specific alleles from other Mytilus hosts. For both mtDNA fragments the 332 cancerous alleles clustered together with the major BTN2-specific alleles and the 62mc10 (Fig. 4). 333 The alleles COI-1 and COI-2 differed from the major COI-B BTN2 allele by 6 and 8 substitutions, 334 respectively. CR-1 allele differed from the closest CR-D allele by 16 substitutions, and from the 335 62mc10 by 11 substitutions. CR-2 differed from the closest CR-C allele by 5 substitutions. The 336 alleles of M. trossulus (i.e. host alleles) from the SOJ were randomly scattered across the 337 14 339 An analysis of recombination identified the same breakpoints in CR-1/CR-1' and CR-D alleles but 340 different breakpoints in CR-2 and CR-C alleles ( Supplementary Fig. S5). However, CR-2 and CR-341 C differed by only two substitutions between the suggested breakpoints. We will therefore adhere 342 to the hypothesis that CR-1 and CR-D, as well as CR-2 and CR-C, represent the same 343 mitochondrial lineages originated through singular recombination events. To remember, 344 individual J161 was heteroplasmic for the close CR-1 and CR-1' alleles. Such "additional" 345 heteroplasmy occasionally occurs in BTN2 in different parts of its geographical distribution (Mch-346 41, Castro-26, Fig. 5). 347 In addition to the genetic similarity of BTN2 worldwide, Fig. 4 also illustrates geographic 348 differences within this cancer. In the SOJ there are basically two cancer mtDNA haplotypes, 349 comprising two 1% divergent COI alleles and two very divergent CR alleles (CR-1, CR-2), while 350 elsewhere there are two cancer haplotypes, comprising basically the same COI alleles of the "B 351 group" and two CR alleles (CR-C, CR-D) similar to that in the SOJ. Another difference is that the 352 two cancer haplotypes are apparently homoplasmic in the SOJ and in Europe but heteroplasmic in 353 Argentina (both alleles present in the same cancerous mussels). 354

Haplotype network 355
Among the COI data sets included in the analysis, that from Vancouver Island, British Columbia 54 356 demonstrated a remarkably strong polymorphism in comparison with other regional sets, including 357 another Northeast Pacific set. Therefore we provide separate networks performed without and with 358 15 the data of Crego-Prieto et al. 54 in Fig. 5 and Supplementary Fig. S4, respectively. The analyses 359 revealed a complex star-like haplotype network familiar from the previous phylogeographic 360 studies 34,36,53 . The network consisted of several major clades, each with a common core haplotype 361 and many rare haplotypes radiating from it. Some clades were "cosmopolitan" and included 362 samples from both the Atlantic and the Pacific oceans, while others were nearly restricted to the 363 Pacific. The unique BTN1 allele belonged to one of the cosmopolitan clades, which included 364 sequences from all the four macroregions considered but was dominated by samples from the 365 Northeast Pacific. The BTN2 alleles, including that from the SOJ, and the Baltic 62mc10, were 366 attached to another clade. This clade included samples from all the macroregions except the 367 Northeast Pacific (Fig. 5), with one notable exception. A single sequence identical to the cancerous 368 COI-1 from the SOJ was identified in the data from Bamfield locality on Vancouver Island 369 (KF931805 54 , Supplementary Fig. S4). 370 371 Species confirmation and sex identification 372 Only M. trossulus alleles were recorded in a subsample of mussels (N=21, including J17, J38, and 373 cancerous J54, J111, J161 and J181) studied by three additional nuclear markers, and the genotypes 374 retrieved from different tissues of the same mussels were always the same. Twelve mussels were 375 identified as "mitochondrial males" by 16S locus, among them J181, and the other mussels, as 376 females. 377

378
In this study we showed that Mytilus trossulus mussels from the SOJ, Northwest Pacific, were 379 affected with disseminated neoplasia (DN) and confirmed that it was caused by clonally 380 transmitted cancer (CTC), by demonstrating genetic chimerism of mussels with DN and a striking 381 similarity among their "extra" genotypes. The cancer alleles found in our study did not match those 382 of the BTN1 lineage from M. trossulus populations on the American coast of the Pacific Ocean 383 but matched the alleles of BTN2 lineage, which has been previously diagnosed in M. edulis from 384 Europe and in M. chilensis from Chile and Argentine but not in M. trossulus, from which it is 385 originally derived from. So, we conclude that M. trossulus from the SOJ are infected by BTN2. 386 This finding implies that this species, contrary to the hypothesis of Yonemitsu et al. 10 , has not 387 evolved resistance to this disease. Below we will first discuss the pathology and epidemiology of 388 BTN2 in mussels from the SOJ and then its genetic properties. 389 The features of the neoplastic hemocytes in our study-a rounded shape, a large nucleus, a high 390 nucleus-to-cytoplasm ratio and an increased ploidy-agree with the previous descriptions of DN 391 in Mytilus 14,22,23 . What seems unusual is their low proliferation level. Neoplastic hemocytes of 392 mussels are generally assumed to have a high proliferation activity 20 . However, a low proliferation 393 rate of neoplastic hemocytes in mussels with DN was reported in two other studies: of BTN 394 (supposedly BTN2) in France 22 and of DN 66 in the same Argentinean population where BTN2 was 395 recognized later 10 . A possible explanation is that in case of BTN2 the proliferation site of the 396 neoplastic cells is located not in the hemolymph. This hypothesis is inspired by the study of Burioli 397 et al. 22 , who observed a high mitotic rate of neoplastic cells in the vesicular connective tissue of 398 BTN-infected mussels. 399 Our data complement the results of another study of DN in mussels from the SOJ 40 , where only 400 one individual with DN was found in a histological examination of 40 mollusks from various 401 localities. We only found DN in one of the three populations examined. The DN prevalence in this 402 Gaydamak Bay population (4.0%) was probably underestimated, being based on flow cytometry, 403 which is not very sensitive at the early stages of the disease 19,22 . Still, our estimate of DN level in 404 the population from the Gaydamak Bay is close to the mean prevalence reported for Mytilus 405 populations worldwide 39,67-70 and much lower than the maximal prevalence of 56% reported for 406

M. trossulus population from British Columbia 71 . 407
Noteworthy, the mussels in the Gaydamak Bay were collected from the surface of a mooring buoy 408 in a heavily polluted area, where no mussels were recorded at the sea floor (our observations). 409 Since BTN is presumably transmitted by cancer cells through the water column (see Caza et al. 72 410 for more discussion), we suspect that the mussels fouling the buoy contracted the infection from 411 those that had fouled ships moored to it. We point out that mussel populations on mooring buoys, It is also possible that the apparent homoplasmy of the SOJ mussel cancers is spurious. Yonemitsu 433 et al. 10 have shown using qPCR that the levels of various haplotypes in heteroplasmic cancer 434 clones may be very different. Since we employed molecular cloning of CR and sequenced a limited 435 numbers of colonies (though more than Yonemitsu et al. 10 did to demonstrate heteroplasmy), we 436 18 might have overlooked rare alleles. Further, if we assume that the cancer clones from Gaydamak 437 are homoplasmic, it follows that the mussels fouling the same mooring buoy were infected with at 438 least two independent clones, which is unlikely, though not impossible. 439 One novel finding in our study is the remarkable divergence between the two BTN2 haplotypes in 440 the SOJ. Two COI alleles differed from each other and from "B alleles" on average by six 441 substitutions (~1%). This diversity has probably accumulated after the emergence of BTN2. In a 442 study of CTVT mtDNA evolution, Strakova et al. 33