Abstract
Cross-species transmission of coronaviruses (CoVs) poses a serious threat to both animal and human health1-3. Whilst the large RNA genome of CoVs shows relatively low mutation rates, recombination within genera is frequently observed and demonstrated4-7. Companion animals are often overlooked in the transmission cycle of viral diseases; however, the close relationship of feline (FCoV) and canine CoV (CCoV) to human hCoV-229E5,8, as well as their susceptibility to SARS-CoV-29 highlight their importance in potential transmission cycles. Whilst recombination between CCoV and FCoV of a large fragment spanning orf1b to M has been previously described5,10, here we report the emergence of a novel, highly pathogenic FCoV-CCoV recombinant responsible for a rapidly spreading outbreak of feline infectious peritonitis (FIP), originating in Cyprus11. The recombination, spanning spike, shows 97% sequence identity to the pantropic canine coronavirus CB/05. Infection is spreading fast and infecting cats of all ages. Development of FIP appears rapid and likely non-reliant on biotype switch12. High sequence identity of isolates from cats in different districts of the island is strongly supportive of direct transmission. A deletion and several amino acid changes in spike, particularly the receptor binding domain, compared to other FCoV-2s, indicate changes to receptor binding and likely cell tropism.
Main
Following two epidemics, SARS-CoV 2002-4 and MERS-CoV 2012-ongoing, and a pandemic of previously unseen proportions, SARS-CoV-2 2019 onwards, coronaviruses no longer need lengthy introductions of importance and scale. They are not only present in the human population, they are also in wildlife13-15, companion animals16-19 and livestock13,20,21, and in all species these viruses have major impacts. The innate ability of coronaviruses to recombine with other coronaviruses continues to fuel their species-switching ability. It is therefore not surprising that both human and animal coronaviruses are linked in complex transmission and evolution cycles3,14,19,22.
FCoV is found across the globe. The virus exists in two biotypes with the main biotype, feline enteric coronavirus (FECV), showing low virulence and clinical signs are typically limited to mild enteritis. The second biotype of FCoV, proposed to originate each time from a mutation in an FECV-infected cat (reviewed in 23), is known as feline infectious peritonitis virus (FIPV). FIPV causes feline infectious peritonitis (FIP), which is a fatal disease if left untreated. Clinical signs include an abdomen swollen due to peritoneal fluid, fever, weight loss lethargy, anorexia, dyspnoea, ocular abnormalities and neurological signs8,16,24,25. Mutations in the spike gene or the accessory genes 3abc and of 7ab of FCoV8,16,23,26 are thought to result in changes to the virus’s tropism from cells in the enteric tract to macrophages, resulting in the different disease presentation seen with the two biotypes. This change in primary tropism also impacts the virus’s ability to transmit from cat to cat, with the main transmission pathway of FECV being faecal-oral and FIPV typically having relatively poor transmission potential. Antivirals, including remdesivir and GS-441524 have recently been successfully used to treat cats with FIP27.
FCoV and CCoV both belong to the Alphacoronavirus 1 species alongside the porcine transmissible gastroenteritis virus (TGEV) and, probably, the never fully sequenced rabbit enteric coronavirus28,29. Both FCoV and CCoV have evolved two different serotypes through complex recombination events between the two viruses with a suspected gradual evolution from CCoV-II to TGEV and the later spike deletion to porcine respiratory coronavirus (PRCV)23,30. Whilst recombination events between CCoV and FCoV have significantly contributed to the serotype evolution and have been frequently described, so far, none of them lead to enhanced transmissibility of FIP beyond closest contact10,23,31,32. Similarly, recombination events have been reported between CCoV and TGEV33, the latter has been found to recombine with the pedacoronavirus (alphacoronavirus genus) porcine epidemic diarrhoea virus6. These observations are particularly important in view of the Alphacoronavirus 1-related human infections recently observed19,22.
Earlier this year, we alerted the veterinary field to an outbreak of FIP in Cyprus, where there had been a concerning increase in cases11. Cases were recorded as FIP only if they had compatible clinicopathological signs and a positive RT-qPCR for FCoV in one of the following samples; peritoneal fluid, pleural fluid, cerebrospinal fluid, fine needle aspiration biopsies, or tissue biopsies from granulomatous lesions. In comparison, in 2021 and 2022, there were three and four RT-qPCR confirmed FIP cases recorded in Cyprus, respectively, to date in 2023 (January-August), 165 cases have been confirmed. This represents more than a 40-fold increase. The outbreak emerged in January 2023 in Nicosia, the capital of Cyprus. An increase in cases was first observed in January 2023 and by February Nicosia recorded the peak number of cases in any district (Figure 1A & B). The next highest number of cases was observed in Famagousta, which peaked in March, and by then the outbreak had spread to all districts of the Republic of Cyprus. (Supplementary Tables 1-5). In June and July there was a decline in RT-qPCR confirmed cases, which coincided with a large media awareness campaign to alert veterinarians to the spread of FIP11. This fall in RT-qPCR confirmed cases is likely due to most veterinarians diagnosing cases based on clinicopathological findings without performing PCR testing due to the additional financial cost. On August 3rd, the Republic of Cyprus minister’s cabinet approved the use of the stocked human coronavirus medications to be used for cats with FIP. In order for veterinarians to have access to this medication, amongst others, a PCR confirmation was required, reflecting the increased cases seen during August 2023. In order for veterinarians to have access to this medication a PCR confirmation was required, reflecting the increased cases seen during August 2023 (Figure 1A & B). The number of unreported cases in Cyprus is very high, not least due to the high number of feral cats. Estimates from the Pancyprian Veterinary Association indicate around 8,000 deaths due to FIP up to mid-July 2023. Furthermore, in October 2023, a first imported case of FIP was confirmed in the UK.
The most common clinical form of FIP was effusive (69.7%; Figure 1C), followed by neurological FIP (27.9%) and the non-effusive form (2.4%) (Supplementary table 4). Where peritoneal fluid was assessed by cytology, non-degenerative neutrophils admixed with macrophages and small lymphocytes were seen in a protein rich background (Figure 1D). Five cases were assessed by histopathology, including intestinal mass (n=1), lymph node (n=2), and kidney (n=2). All showed similar histological features, with multifocal to coalescing, pyogranulomatous to necrotising and lymphoplasmacytic inflammation (Figure 1E). The angiocentric nature can be seen in some areas, whilst in others there is total effacement of the tissue. Immunohistology for FCoV antigen demonstrated a heavy viral load within intralesional macrophages (Figure 1E).
RNA samples were obtained from 91 confirmed FIP cases between 2021 and 2023, representing a mixture of geographic origin, sex, and clinical presentation (Supplementary Tables 6-11). They were sequenced using cDNA/PCR-amplification-based Nanopore sequencing to better understand the outbreak and to determine if cat-to-cat transmission of FIPV is occurring. Additional samples were submitted from two cats presenting with FIP following recent import from Cyprus to the UK. Initial focus laid on amplification of the spike, which was successful in 43 of the Cypriot samples and in the two imported UK cat samples. The other samples were degraded or contained too few viral copies. None of the seven samples from before 2023 amplified (Supplementary Tables 12-17). Other regions of the genome were amplified in several individuals; however, none of the cats had the entire genome amplified due to intentional conservation of limited samples until a working primer scheme is designed for the entire genome.
The sequenced spike region of the FCoV samples produced three distinct versions of the spike sequence. BLAST was used to identify close relatives of these spike sequences. The first is a spike sequence most closely related to an FCoVI (genbank ID MT444152.1) with 79% similarity, which occurred in one Cypriot and one UK-import sample, and the other two are almost entirely CCoVII spike sequence, flanked by FCoVI sequence. The CCoVII sequence is most closely related to the NA/09 strain (GB JF682842), a hypervirulent pantropic canine coronavirus (pCCoV)34, at 97% sequence identity. The spike sequence is also closely related to other pCCoV spike sequences with only partial spike sequences available (Figure 2, Supplementary Figure 1). Two of the samples, one Cypriot and one UK-import case, showed high similarity with FIPV-1 spikes. However, all other samples, including one UK-import case align with a pCCoV spike. This is likely a defining feature of the virus circulating in the outbreak in Cyprus. There are two versions of this spike gene, one of which has a deletion of approximately 630bp near the beginning of the spike sequence in the N terminal domain, and this deletion version is present in the majority, 35/43 of the pCCoV spikes sequenced.
Regions of POL1ab and ORF3c/E/M were amplified in several of the cats in addition to the spike genes. These sequences and the spike sequences were individually aligned to the CCoVII, FCoVI and FCoVII whole genome sequences available on NCBI. Figure 2 shows maximum likelihood trees all three different regions (also Supplementary Figures 2-4). While the POL1ab and ORF3c/E/M amplicons all cluster with FCoVI, the spike sequences cluster with CCoVII, most closely with the pantropic strains NA/09 and CB/05. The spike amplicon from one UK-import cat clusters among the spike sequences from the outbreak in Cyprus.
A representative genome of FCoV-23 was assembled from amplicons amplified from multiple Cypriot samples (Supplementary File 1), a single gap of 1,221bp remains in this assembly at the second recombination breakpoint. Supplementary Figure 5 shows a neighbour joining tree for the assembled coronavirus along with other members of Alphacoronavirus 1 and a distantly related Canine Respiratory Coronavirus as an outgroup. The assembled genome clusters with representatives of FCoVI, similar to the clustering of the amplicons outside of the spike region.
A recombination analysis was carried out comparing the FCoV-23 genome with FCoVI, FCoVII and CCoVII strain CB/05, since NA/09 is not available as a complete genome. Figure 3A shows visualisations of the Bootscan35 analysis, the RDP536 analysis and the pairwise distances between the sequences. Significant p-values supporting the recombination were reported by multiple methods as listed in Supplementary Table 19. The MaxChi37 breakpoint matrix is available in Supplementary Figure 6. Despite the second breakpoint being missing from the analysis due to the remaining gap in the genome, the recombination is very clear and includes a small region of the POL1b gene and the majority of the spike gene, but none of the genes to the 3’ end of spike. Some reads, while not sufficiently high coverage to create a consensus for this gap region, suggest that the recombination break point occurs within the spike gene, close to the 3’ end (Supplementary Figure 7). Figure 3B shows the historical break point between FCoVI and CCoVII that created FCoVII alongside the recombination identified here.
The main determinant in disease development and transmission of FCoV-23 appears to be the spike recombination. One of the main suggested determinants of biotype changes, the furin cleavage site (FCS) at the S1/S2 interface23,26 is absent in FCoVIIs and also FCoV-23. An interesting observation, however, is that the majority of samples show a deletion in domain 0, strongly resembling the deletion observed in TGEV and porcine respiratory coronavirus (PRCV) (Figure 4A). In other coronaviruses, including TGEV38 and CCoV-HuPn-201839, domain 0 was shown to bind sialosides. Modelling the structure of spike against the closely related, experimentally confirmed CCOV-HuPn-2018 spike39 shows a much more compact confirmation for the domain 0 deletion spike and similarity to a structural prediction based on a “swung-out” or a “proximal” confirmation template (Figure 4B, Supplementary Figure 8). A number of amino acid changes were observed between “classical” FECV-2 and FIPV-2 spikes. In particular domains A and B, the receptor binding domain (RBD) show a number of class changing amino acid changes distinct from FCoV-2 spikes (Figure 4A). Modelling the RBD changes against the structure highlights changes at positions 546 and 595, as well as 556, 603, and 636 as being potentially strongly influential to receptor binding properties (Figure 4C).
Previously indicated key proteins for biotype switch spike, Orf3abc and 7b were compared to the recently published computational analysis of mutations observed in FECV versus FIPV by Zehr et al.26. Unfortunately, the suggested key determinant of FIPV in the FCoV-2 spike, position 1404, is not yet resolved in FCoV-23. Other positions in spike show new mutations or are suggestive of FIPV (Supplementary Table 20). Whilst a new mutation each was identified in Orf3a and b, no specific indications of pathogenesis could be determined (Supplementary Table 21). Similarly, in Orf7b, two new mutations but no indication of pathogenesis could be identified (Supplementary Table 22).
Discussion
Recombination within the Alphacoronavirus 1 species has been frequently observed previously and has even given rise to the FCoVII serotype5,23,24(Figure 3B) in cats. Even complex recombination between FCoVI and CCoV has previously been observed10. However, in this work, we have identified a new subtype of FCoV that has recombined with a hypervirulent strain of pCCoV. The recombinant, which we propose naming FCoV-23, shows clearly distinct properties from previously observed FCoV infections. The sequence similarity found between FCoV-23 and CCoVII is higher than would be expected without an additional, recent recombination event.
Preliminary data suggest there may be no need for a biotype change to result in FIP and there is compelling evidence supportive of direct transmission of FCoV-23. However, more work is needed, including investigation into asymptomatic carriers. Of support, onset of FIP appears rapid and shows little discrimination in age of the infected cats. Most importantly, remarkably high sequence identity from isolates collected at different times and locations, is highly supportive of direct transmission of FCoV-23. The scale of this FIP outbreak has not been observed previously. Concerningly, Cyprus has a high population of unowned cats that are frequently relocated to other parts of Europe, and the wider world. The risk of spreading this outbreak is significant as evidenced by the first confirmed UK case40. This is exemplified by the recent confirmation of a first UK-imported case with further investigations into other cases ongoing.
CCoV infections in dogs are typically self-limiting, producing mild or asymptomatic enteritis. Previous work on the FCoV-23 close relative strain, CCoV NA/09 and CB/05, found the virus was hypervirulent and infected a range of organs in canine hosts. CCoV typically infects cells of the enteric tract, but pCCoVs were shown to spread to a range of internal organs, including intestine, lung, spleen, liver, kidney, lymph nodes, brain, and even T and B cells 34,41-43. CB/05 has also been identified as responsible for small outbreaks41.
The clinical signs of FCoV-23 are similar to those observed in classical FIP cases. However, FIP infection itself already shows a strong pantropism, affecting many organs in infected cats25. What is remarkable, is the high number of FCoV-23 positive cells observed in immunohistochemistry samples (Figure 1C). This is indicative of high viral loads, but must be validated through quantification. Furthermore, a relatively high percent (Supplementary Table 4) of confirmed cases in Cyprus was presented with neurological signs (28%), which is double of what would be expected with classical FIP (14%)27. This may be due to increased awareness of presentation or be inherent to FCoV-23 neurotropism; however, further investigation is required.
Significant changes to the spike protein in FCoV-23 may provide some clues as to the enhanced pathogenicity of this virus. FCoV-23, like other FCoVIIs and CCoVII, does not contain an FCS at the S1/S2, and the related virus CCoV-HuPn-2018 was experimentally found to be uncleaved39. Uncleaved spike has been shown to be more stable than cleaved one and could enhance faecal shedding and stability in the environment44. Conversely, the cleavage at S1/S2 may facilitate the movement of the N-terminal domains and allow the RBD to adopt the receptor-binding-competent form. However, the deletion of the N-terminal domain 0 may compensate for the increased rigidity. Whilst a similar deletion between TGEV and PRCV was initially indicated as the determinant between a primarily enteric and primarily respiratory tropism, respectively45,46, a recent study shows a different result in vivo47. Sialoside binding is an important feature of coronavirus infections and may contribute to intracellular spread48. It is then all the more surprising that FCoV-23 loses the sialoside-binding domain 0. However, whilst binding to sialosides can enhance virus attachment and entry into host cells, binding to sugars can lead to increased retention of virus on producer cells. Whilst some viruses, like influenza, solve this issue by encoding their own sugar-cleaving enzyme, neuraminidase, coronaviruses must find a balance through modifying glycoprotein binding49. Losing domain 0 could therefore be a trade-off to enhance virus release. This may be compensated by the earlier discussed enhanced flexibility of spike and increased binding efficiency at the RBD through mutation. These are manifold in the FCoV-23 pCCoV recombinant and in particular mutations at position 546 (already possibly associated with FIP) and 595 are likely to have a strong impact on aminopeptidase N receptor binding (Figure 4C).
Recombination between FCoVI and pCCoVs is not surprising in that i) such recombinations have been observed before, and ii) pCCoVs have evolved in Greece and south eastern Europe34,42. However, the emergence of pCCoV in Greece happened over 18 years ago, so the question as to why this happened now is unclear yet. One possible explanation is the “right mutation, right time, right place” theory. Recombination between feline and canine coronaviruses happen frequently but as previous reports show, they rarely spread. Introduction of a more successful, spreading variant to a dense population, like the cats in Cyprus, may be sufficient to allow this virus to cause an outbreak. Whilst alphacoronaviruses show great cross-neutralizing activity39, neutralising immune evasion of FCoV-23 may be a contributing factor as indicated by the infection of cats of all age groups. However, further investigation is required to determine shedding routes and virulence. Preliminary RT-qPCR results and high virus-infected cell positivity in the colon are strongly indicative of faecal shedding (data not shown and figure 1D). Analysis of historic samples, (asymptomatic) cats and dogs may help further evaluate the origins of this virus.
This paper reports the emergence of a new alphacoronavirus 1 feline/canine coronavirus that shows high spreading potential with the associated pathology of lethal FIP if left untreated. Further investigations into the properties of this new virus are now essential. Whilst antivirals, including GS-441524 and molnupiravir were successfully used to treat some cats affected by FCoV-23, early treatment is essential but associated costs are often prohibitive. Prevention of spread and the development of new vaccines are important to stop this virus epizootic from becoming a panzootic.
Author contributions
Conceptualization, C.A., A.W., D.G-M., S. M. and C.T-B.; Methodology, C.A., A.W., S.M., and C.T-B.; Validation, C.A., A.W., S.M. and C.T-B.; Formal Analysis, A.W., S.M. and C.T-B.; Investigation, C.A., A.W., D.E., M.O., S.F., A.M., M.L., R.H., C.T-B.; Resources, C.A., A.W., D.E., M.L., A.H., A.Z., S.L., M.G. and C.T-B.; Data Curation, A.W., S.M., and C. T-B.; Writing – Original Draft Preparation, C.A., A.W., and C.T.B.; Writing – Review & Editing, C.A., A.W., S.F., D.G-M., S.M. and C.T-B.; Visualization, A.W., S.M. and C.T-B.; Supervision, C.A. and C.T-B.; Project Administration, C.A. and C.T-B.; Funding Acquisition, C.A., and C.T-B.
Funding
This work was supported by EveryCat Health Foundation award number EC23-OC1 (C.A., D.G-M., C.T-B.), Vet Dia Gnosis Ltd (C.A.). This research was funded by the BBSRC Institute Strategic Programme grant funding to the Roslin Institute, grant numbers BBS/E/D/20241866, BBS/E/D/20002172, and BBS/E/D/20002174 (C.T-B.).
Competing interests
The authors declare no competing interests.
Methods
Enrolment of FIP cases in Cyprus
The electronic records of Vet Dia Gnosis in Limassol, Cyprus were searched for any positive FCoV RT-PCR cases from September 2021 up to August 2023. The cases to be enrolled needed to have compatible clinicopathological findings for FIP as recently outlined by the European Advisory Board on Cat Diseases Guidelines25 as well having a positive RT-PCR for FCoV in one of the following samples: peritoneal fluid, pleural fluid, cerebrospinal fluid and granuloma fine needle aspiration biopsies or tissue biopsies. The original samples were submitted to the Vet Dia Gnosis commercial laboratory (Limassol, Cyprus) by local veterinarians and then submitted to LABOKLIN commercial laboratory (Bad Kissingen, Germany). Ethical approval for this study was granted by the Pancyprian Vet Association. According to the terms and conditions of the Vet Dia Gnosis, as well as the Cypriot legislation [The Dogs LAW, N. 184 (I)/2002], no special permission from animal owners or the animal welfare commission is needed for additional testing on residual sample material once diagnostics are completed. According to the terms and conditions of the LABOKLIN laboratory, as well as the RUF-55.2.2.2532-1-86-5 decision of the government of Lower Franconia, no special permission from animal owners or the animal welfare commission is needed for additional testing on residual sample material once diagnostics are completed.
Sample from cat imported to the UK
Two UK veterinary practices contacted the Royal Dick School of Veterinary Sciences (R(D)SVS) with suspected cases of FIP in cats recently imported from Cyprus. Peritoneal fluid samples were taken from the cats and sent to R(D)SVS for further testing and for sequencing at The Roslin Institute with the consent of the cats’ owners. The veterinary practices and the cats’ owners were kept informed at all stages.
Histopathology and immunohistochemistry
A sub-set of cases (n=4) from the FIP cats diagnosed after January 2023 with available tissue specimens were selected. Tissue specimens were carefully obtained and immediately fixed in 10% buffered formalin. Following fixation, the tissues were processed by embedding, 4μm sectioning, and subsequent staining with hematoxylin and eosin (H&E). The inclusion criteria were defined based on comprehensive histopathological assessments. Specifically, emphasis was placed on identifying the characteristic FIP-associated histological hallmarks, which encompassed vasculitis, phlebitis, and periphlebitis. Consecutive tissue sections were mounted onto charged slides. Following pre-treatment for 5 min at 110°C in 0.01M pH 6 citrate buffer, slides were incubated for 30 min at room temperature with primary mouse monoclonal anti-feline coronavirus antibody at 1:400 (BioRad, MCA 2194). The EnVision anti-mouse system was used for visualization according to the manufacturer’s instructions (Agilent).
RNA extraction and cDNA synthesis
RNA from specimens from Cyprus underwent automated total nucleic acid extraction using the MagNA Pure 96 DNA AND Viral NA Small Volume kit (Roche Diagnostics). RT-PCR for FCoV was performed at Laboklin Bad Kissingen, Germany50.
RNA samples from cats imported to the UK from Cyprus were extracted using the QIAamp viral RNA extraction kit (Qiagen) according to the manufacturer’s instructions.
cDNA synthesis on all RNA samples was carried out using LunaScript RT SuperMix Kit (NEB) with 16 μl template RNA in 20 μl reactions according to the manufacturer’s instructions.
Primer design and PCR amplification
A set of primers for 1kb tiled amplification were designed initially using primalscheme51 with available FCoV genomes on NCBI52, and then by manual redesigns. Currently there is not a working scheme for the entire genome, however, through multiple redesigns on multiple samples it was possible to amplify sections which, when combined, cover the entire genome with one gap. It should be noted that sequences obtained from this process do not all originate from the same sample, and the genome assembled from these is representative at a population level rather than an individual level, and therefore may not accurately represent virus present within an individual. Efforts are ongoing to create a working tiled amplification scheme. The primer sequences used here are in Supplementary Table 18. The PCRs for these multiplexed amplifications used the same conditions described below, but with multiplexed primers.
DNA was amplified using primers targeting the majority of the spike region and incorporating a flanking region of known FCoV sequence, with an expected amplicon size of around 3.7 kb. These were designed based on previously amplified flanking sequences during the tiled amplification design process: forward primer 5’-GACGCAGACTTCAGTGTTA-3’; and reverse primer 5’-ACCATTATGCCATTRTARTA-3’. These primers do also produce offtarget amplification of the feline growth hormone receptor. Additional amplicons designed during the tiled amplicon designing process were used to amplify regions of POL1ab and Orf3c/E/M from the samples (Supplementary Table 18).
PCR amplification was carried out on each sample using VeriFi Hot Start Mix (PCR Biosystems), 1.25μl 10μM forward and reverse primers and 2μl cDNA in a 25μl reaction. Amplification was performed with the following PCR conditions: initial denaturation at 95°C for 1 minute, then 35 cycles of denaturation at 95°C for 15 seconds, annealing at 52°C for 15 seconds, extension at 72°C for 4 minutes, followed by a final extension at 72°C for 5 minutes. Amplified DNA was purified using AMPure XP beads (Beckman Coulter). The amplicons targeting POL1ab and NSP3c/E/M were amplified in a pool with other primer pairs using the same PCR conditions but with 4μl cDNA in the reaction instead of 2μl.
Nanopore sequencing
Amplicons were quantified using Qubit (Thermo Fisher) High Sensitivity assays and diluted to 150ng DNA per sample in 12μl nuclease-free water. Libraries were prepared using Oxford Nanopore Technologies’ (ONT) NBD112.96 ligation kit following the manufacturer’s protocol for amplicon sequencing with some modifications. Due to unavailability of NBD112.96 kit reagent AMII H, following ligation of barcodes a second end-prep was carried out using the Ultra II End Repair module (NEB). The rest of the protocol was carried out from the adapter ligation stage as per manufacturer’s protocol for LSK112 using the AMX-F adapter supplied in the early access Q20+ version of kit 112. The library was loaded onto a MinION R10.4 flow cell and sequencing was carried out on a GridION sequencing device.
In order to identify the virus present in one of the UK cases as quickly as possible, only the spike amplicon was amplified and the sample sequenced using ONT RAD004 rapid sequencing kit on an R9.4 flow cell following manufacturer’s protocol. The other UK case was amplified at an earlier time and was included in a sequencing run with the Cypriot samples, the data from that sample was treated in the same way as the Cypriot samples and was the sample later found to have the non-recombinant FCoVI spike.
Bioinformatic analysis
Basecalling and demultiplexing was carried out on the GridION sequencing device (ONT) using Guppy (v2.24-r1122) on super accurate mode, specifying --require_barcodes_both_ends and using appropriate super accurate basecalling models for each of the different sequencing methods used. Following basecalling, amplicon_sorter53 (v2023-06-19) was used to identify consensus amplicons between 3kb and 5kb without using a reference. Spike amplicons were identified from the output via alignment with minimap254 (v2.22) to the spike from the first sample we sequenced (Cypriot_G7_Gi_6739) for which a consensus was made using the same process, with the correct amplicon identified via BLAST55. The identified amplicons were polished with medaka (v1.8.0, ONT). For the UK case which was sequenced with the rapid kit, because the reads were fragmented by the library prep process, reads were used to polish the Cypriot_G7_Gi_6739 sequence using medaka. The reads were aligned to the polished sequence with minimap2 and visualized with IGV56 (v2.11.1) to visually confirm the reads supported the consensus sequence and that it had not been biased by the reference used. The amplicons in POL1ab and Orf3c/E/M were assembled and polished using LILO57. Multisequence alignments were carried out for each of the regions using mafft58 (v7.49) against all complete genomes of CCoV and FCoV genomes available on NCBI using the --adjustdirection flag. Alignments were visualized in Mega759 and sequences with poor sequencing quality were removed. All downloaded genomes were trimmed down to each of the target regions amplified from our samples. Maximum likelihood trees were constructed using IQ-TREE60 (v2.0.5). TempEst61 (v1.5.3) was used to determine the best fitting root for the trees, and visualizations and annotations of the trees were done using iTOL62 (v6.8.1). The trees were combined into figures using BioRender.
To assemble a representative genome for the population, all of the samples for which we had sequencing data were run through LILO. The polished amplicons were mapped to a single fasta file containing an FCoVI (MT239440.1) and a CCoVII (KP981644.1) using minimap2 and visualized with IGV, with only the amplicons from the spike region aligning to CCoVII. Representative amplicons covering the entire genome were selected and scaffolded using scaffold_builder63 with MT239440.1 as a reference. The raw reads were trimmed by 50bp to remove adapters and barcodes and they were used to polish this scaffold using medaka. Mafft was used to align this genome to representative genomes from Alphacoronavirus 1 with Canine Respiratory Coronavirus as an outgroup, and Mega7 was used to create a neighbour join tree which was visualized and annotated with iTOL. BLAST was used to align reads that span the gap to existing CCoV and FCoV genomes to determine the likely breakpoint (results in Supplementary Figure 6).
A multisequence alignment between the genome and MT239440.1, LC742526.1 and KP981644.1 was carried out using ClustalW64 in Mega7, and recombination analysis carried out using RDP536 (v.5.45) on default settings.
Structure prediction of spike deletion variants
SWISS-MODEL structure prediction and analysis (https://swissmodel.expasy.org/, accessed October 2023)65 were used to model the partial, high confidence N-terminal sequence of a full-length spike (1008aa) and a deletion spike (797aa). To model the G8 Cypriot full-length spike, we used the alphacoronavirus 1 experimentally resolved CCoV-HuPn-201839 structure as a template. Modelling against the 7usa.1.a, swung out confirmation, yielded GMQE 0.60 and Global QMEAND of 0.68±0.05 with a sequence identity of 88.97% and against 7us6.1.A, proximal confirmation, yielded GMQE of 0.62 and Global QMEAND of 0.58±0.05 with a sequence identity of 81.03%. Modelling of C10 spike, C-terminal deletion, against 7usa.1.A, swung out confirmation, yielded GMQE 0.74 and Global QMEAND of 0.69±0.05 with a sequence identity of 88.66% and against 7us6.1.A, proximal confirmation of HuPN, yielded GMQE of 0.76 and Global QMEAND of 0.70±0.05 with a sequence identity of 88.42%.
The receptor binding domain was modelled using the 7u0I.1B structure, again CCoV-HuPn-201839 complexed with canine APN as a template. GMQE was 0.89 and Global QMEAND 0.85±0.07 with a sequence identity of 92.41%. PDB files are available in the supplementary materials.
Acknowledgements
The authors like to thank the Pancyprian Veterinary Association and all the veterinarians in Cyprus for sample submission. We would like to thank Dr Sam Lycett, The Roslin Institute, The University of Edinburgh for advice on sequence analysis.
Footnotes
Typo in the title corrected.