3. Results

3.1. Identification of Halophytophthora thermoambigua CRESS virus 1

Transcriptomic screening of eight pooled strains of Halophytophthora spp. (Table S1) yielded 729 sequencing reads that could be assembled into a 979 nt-long contig (NODE_2192_length_979; Fig. S1) that displayed similarity to unclassified ssDNA viruses, including Tick-associated circular DNA virus (UTM74968.1; 27.6% identity, 76.3% query coverage, E = 2.69E-11) and Grus japonensis CRESS-DNA-virus sp. (QTE03368.1; 26.9% identity, 71.2% coverage, E = 1.17E-09). The hits to viruses encompassed a region encoding the Gemini_AL1 domain (Pfam: PF00799; E-value: 2.26e-07) corresponding to the catalytic domain of the geminiviral replication initiation protein (Rep), suggesting that the assembled contig corresponds to a previously unknown member of the phylum Cressdnaviricota. Notably, a BLASTx search queried with the contig sequence identified as the top hit a hypothetical protein (N0F65_000360) encoded by the mitochondrial genome of the oomycete Lagenidium giganteum (accession DAZ93511.1), with 67.81% identity and 45% query coverage (E-value = 4e-63). The second and third best hits were also to two mitochondrially encoded hypothetical proteins from P. sojae (YP_001165403 and YP_001165402), pointing to the association of the identified virus with different oomycete species.

To identify the Halophytophthora strains infected with the putative ssDNA virus, total nucleic acids were extracted from the eight strains originally subjected to transcriptomic analysis, and virus identification was performed by PCR amplification using a pair of specific primers designed based on the contig sequence (Table S2). The positive signal was obtained exclusively with the DNA extracted from H. thermoambigua BD651, the ex-type strain of the species isolated from brackish water in Parque Natural da Ria Formosa, Santa Luzia, Tavira (Portugal). To obtain the complete genome of the putative ssDNA virus, DNA extracted from isolate BD651 was amplified by PCR using abutting primers (Table S2). After DNA purification, cloning, and sequencing, a complete circular genome of 2793 nt was obtained (Figs 1A and S1) and denoted Halophytophthora thermoambigua circular Rep-encoding single-stranded DNA virus 1 (HtCRESSV1) (GenBank accession number PV883009, SRA Bioproject PRJNA1437999). Although our data do not support a multipartite genome, we cannot formally rule out that a highly divergent or very low-abundance additional segment(s) escaped detection by homology-based searches.

Figure 1: Genome organization and rep motif architecture of HthCRESSV1. (A) Schematic representation of the circular genome of HthCRESSV1 showing the two major ORFs encoding the replication-associated protein (Rep) and the hypothetical protein (HP), as well as the large (LIR) and small (SIR) intergenic regions. Predicted stem–loop structures and repeated elements are indicated. (B) Southern blot analysis of HthCRESSV1. M, digoxigenin-labelled DNA size marker (λ DNA digested with HindIII). The viral DNA forms are indicated as follows: double-stranded (ds) DNA; cr, circular single-stranded (ss) DNA. Both forms migrate faster than the linear full-length genome (~2.7 kb, determined by inverse PCR), consistent with their circular/single-stranded conformations rather than a difference in sequence length. (C) Comparison of Rep motif composition across representative members within the order Geplafuvirales and HthCRESSV1, including rolling-circle replication motifs (I–III) and helicase motifs (Walker A, Walker B, and motif C). Geplanaviridae: QDJ95229 (Capybara virus 11), NP_066185 (Horseradish curly top virus); Geminiviridae: ALE29795 (Lake Sarah-associated circular virus 45); YP_009506372 (Abutilon golden mosaic Yucatan virus), P14991.2 (Beet curly top virus), Q88888.1 (Tomato pseudo-curly top virus), P0CK40.1 (Bean golden yellow mosaic virus), YP_003778178 (Turnip curly top virus); Genomoviridae: YP_010798490 (Zizania latifolia genomoviridae), YP_010798492 (Artemisia carvifolia genomoviridae), YP_009181999 (Soybean associated gemycircularvirus 1), UIW22431 (Sclerotinia sclerotiorum hypovirulence associated DNA virus 1), QXP49794 (Botrytis gemydayirivirus 2). Numbers in grey boxes represent number of nucleotides deleted in the alignment. (D) Randomized Axelerated ML tree (RAxML) depicting the phylogenetic relationships of the predicted Replicase. Nodes are labelled with bootstrap support values ≥50%. Branch lengths are scaled to the expected underlying number of substitutions per site. Some of the classified families and genera are collapsed to simplify the tree. Family classification and the corresponding GenBank accession numbers are shown next to the virus names. Scale bars represent expected changes per site per branch. Presence of a capsid protein with a single jelly-roll fold (SJR CP), and an ORF2-encoded hypothetical protein (HP) as described in this study, are indicated by green and black circles, respectively. Grey circles indicate features that are unknown or could not be confidently assigned.

To further confirm the presence of HtCRESSV1 in H. thermoambigua and to assess the existence of the single-stranded and double-stranded replicative intermediates expected of viruses that replicate by rolling circle mechanism (as suggested by the presence of the rep gene), purified DNA samples obtained from mycelia were subjected to Southern hybridization with HtCRESSV1-specific probes. Two HtCRESSV1-specific bands were detected: one displayed faster mobility than the 2027 bp marker of λDNA digested with HindIII, whereas the second one displayed slower mobility and exceeded 2322 bp (Fig. 1B). The distinct nature and size of the faster-migrating band are consistent with a circular ssDNA genome of ~2 kb, whereas the slower-migrating band is compatible with a dsDNA-replicative intermediate; both are essentially full-length relative to the 2793 bp genome. Their difference in electrophoretic mobility is most parsimoniously explained by a difference in conformation, a circular single-stranded genome (faster-migrating band) versus a circular, relaxed double-stranded replicative form (slower-migrating band), rather than by a difference in sequence length. A defective, deletion-bearing genome would instead be expected to migrate as a distinctly shorter band, which we do not observe, a pattern analogous to that reported for the ssDNA/dsDNA-replicative forms of Sclerotinia sclerotiorum hypovirulence-associated DNA virus 1 (Yu et al. 2010). The presence of two defined viral DNA forms, rather than a heterogeneous smear indicative of degradation, is consistent with the replication of the viral genome as an episomal, likely circular molecule.

3.2. Description of the genomic features of HthCRESSV1

The complete HtCRESSV1 genome contains two open reading frames (ORFs), both oriented in the same direction (Fig. 1A). ORF1 encodes the putative Rep of 380 aa (estimated 44.66 kDa), with the domain organization typical of members in the phylum Cressdnaviricota (Kazlauskas et al. 2019), whereas ORF2 encodes a hypothetical protein (HP) of 363 aa (estimated 43.19 kDa). The two ORFs are separated by two intergenic regions. The smaller intergenic region (SIR) comprises a 29-nt AT-rich sequence located between the stop codon of the HP and the start codon of the Rep and has the potential to form a stable secondary structure (Figs 1A and S2). The larger intergenic region (LIR) of 529 nt separates the 3′ end of Rep and the 5′ end of HP. A putative TATA-like promoter motif (TATATATATTACTA) is present at the 5′ end of the LIR, suggesting it may function as the transcriptional start site for the downstream HP gene. Notably, LIR features several tandemly repeated 72-nt units (TAGATTATAAAAAACCACAAGAAAGAATTGGTTTTTAGTTAAAAAGTTGGTTTTTAGGTTTTTTATATGGTT), encompassing fragments capable of forming stable stem-loop structures (Figs 1A and S2). The 3′ end of LIR includes a predicted stable hairpin with a 9-nucleotide loop, which could serve as the origin of replication, where the viral Rep initiates the rolling circle replication by introducing a nick. Notably, the sequence of the putative nonanucleotide, TGGCAGTGT, present at the apex of the hairpin is different from those predicted for other well studied cressdnaviricots, such as geminiviruses, genomoviruses, smacoviruses, or circoviruses (Varsani and Krupovic 2017, 2018, Fiallo-Olivé et al. 2021). Thus, the overall architecture of the LIR, including its AT-richness, promoter-like motifs, and stem-loop structure support its potential dual role in initiating both replication and transcription.

3.3. Comparative analysis of Rep and HP sequences

Multiple Rep sequence alignment demonstrated conservation of canonical rolling-circle replication (RCR) motifs I–III and helicase-associated Walker A, Walker B, and Motif C regions, confirming that HthCRESSV1 encodes a putatively functional rolling-circle Rep protein (Fig. 1C).

Maximum likelihood phylogenetic analysis placed HthCRESSV1 within a clade of unclassified viruses assigned to the order Repensiviricetes. This assemblage of unclassified viruses is more closely related to the family Geplanaviridae than to Genomoviridae or Geminiviridae (Fig. 1D), highlighting its distinct evolutionary position within Cressdnaviricota. In addition to viral homologues, BLAST searches identified Rep-like sequences embedded in the mitochondrial genomes of several oomycetes, including L. giganteum, P. sojae, and Peronospora tabacina (Table S3). These sequences cluster with HthCRESSV1 in the phylogenetic tree, indicating shared evolutionary ancestry between the viral Rep and mitochondrial Rep-like elements in oomycetes.

The BLASTn search revealed high nucleotide identity (76.88%) between HtCRESSV1 and an unannotated region of the mitochondrial genome of P. palmivora (culture ICMP:17709; NC_056125.1; 80% query coverage, E-value = 0.0). Similar matches were detected in mitochondrial sequences of other oomycetes, including P. colocasiae, P. botryosa, and Pseudoperonospora cubensis (Table S4). These homologous regions included the coding regions of both HtCRESSV1 Rep (Table S4) and HP (Table S5). Alignment of the HtCRESSV1 HP with the homologous mitochondrial sequences revealed strong aa conservation across several regions (Fig. S4).

In all characterized cressdnaviricots, the second of the two major ORFs encodes a capsid protein with the jelly-roll fold. Unexpectedly, sequence analysis showed that HP encoded by ORF2 does not have the potential to adopt the jelly-roll fold and instead is likely to be a membrane protein. The protein is predicted to be predominantly α-helical (70.2%), with β-stranded regions and unstructured coils contributing 6.6% and 23.2%, respectively. Five transmembrane domains were predicted in the N-terminus of the protein, whereas ~50 C-terminus proximal amino acids were predicted to form a coiled coil (Fig. 2A). These features are unprecedented among proteins encoded by cressdnaviricots. Indeed, BlastP searches queried with the HP sequence against viral databases at NCBI and IMG/VR yielded no significant hits. Homologues of HP were exclusively encoded within the above-mentioned putative EVEs identified in oomycete mitochondria, suggesting specific association with HthCRESSV1-like viruses. To gain further insight into the topology and structure of HP, we attempted to model its structure using AlphaFold3 (Zimmermann et al. 2018). Given that AlphFold3 does not explicitly model membrane proteins in their natural membrane environment, we opted for modelling only the HP ectodomain first, i.e. the domain that follows the transmembrane helices and is predicted to extend away from the membrane. To determine the likely oligomeric state of HP, we modelled multimers with progressively increasing oligomeric states from monomer to decamer and compared their predicted template modelling (pTM) and interface predicted template modelling (ipTM) scores (Table S6). The pTM and ipTM scores gradually improved from monomer to heptamer (best scores: ipTM = 0.77, pTM = 0.77), and started to decrease again for higher oligomeric states, suggesting that HP forms heptamers (Figs 2B and S4). In the modelled heptamer structure, the subunits arranged into an elongated tube-like structure, where the central α-helix formed contacts with the neighbouring subunits. The predicted coiled-coil region was predicted to fold on itself, forming intramolecular contacts at the distal part of the ectodomain. We next modelled oligomers of the full-length HP. As in the case of the ectodomain, the highest pTM and ipTM scores were obtained for the hexameric and heptameric states, although the overall scores were considerably lower than for the ectodomain (Table S6). Nonetheless, the ectodomain model could be superposed with the full-length model (root-mean-square deviation [RMSD] of 2 Å over 132 residues), supporting the relevance of the full-length heptameric model (Fig. 2C and D). Thus, structural modelling suggests that HP forms an extended, membrane embedded spike-like complex. Structure-based searches with the monomer or heptamer against structure databases using DALI (Holm 2022) and FoldSeek (van Kempen et al. 2024) did not reveal structural homologues, further emphasizing the distinctiveness of the HP compared to the previously characterized cellular and viral proteins.

Figure 2: Structural modelling of the ORF2-encoded hypothetical protein. (A) Sequence features of the HthCRESSV1 hypothetical protein (HP). The sequence is represented by a rectangle, with the positions of the five transmembrane domains (T1–5) and the coiled-coil region indicated. (B) Structural model of the heptamer of the HP ectodomain generated with AlphaFold3. The model is coloured based on the predicted local distance difference test (pLDDT) score, a per-residue measure of local confidence, from red to blue (the colour scale is provided at the bottom of the figure). (C) Structural model of a heptamer of the full-length HP. surface representation is shown, with each subunit coloured differently. For one subunit, the surface is set to be semitransparent to show the secondary structure elements. (D) A monomer of the full-length HP. One of the subunits modelled in the context of a heptamer is shown separately and coloured to depict the membrane-spanning domains and the coiled-coil region, as indicated in panel (A).

3.4. Visualization of possible virus-like particles

The presence of viral particles has been demonstrated for a number of fungal cressdnaviricots. To verify whether HthCRESSV1 is capable of forming VLPs, despite lacking the gene for canonical capsid protein, we attempted to purify VLPs using standard purification protocols and visualized them by TEM (Fig. 3A). Following fractionation of the putative VLP preparation on the continuous sucrose density gradient, 15 fractions were collected and analysed for the presence of HthCRESSV1 genome using PCR with specific primers. HthCRESSV1 genomic DNA was abundant in fractions 4 and 5 (15%–20% sucrose) (Fig. 3B). TEM examination of these fractions revealed numerous spherical, pleomorphic particles that co-fractionated with HthCRESSV1 DNA. Consistent with the absence of a canonical capsid protein gene in the HthCRESSV1 genome, no icosahedral particles were observed (Fig. 3C). In the absence of biochemical characterization of the VLPs, we refrain from referring to the observed virus-associated particles as bona fide HthCRESSV1 virions. Further infectivity assays and more detailed ultrastructural characterization of the particles will be required for establishing their provenance and role during infection.

Figure 3: Purification attempts of HthCRESSV1 virus-like particles (VLPs). (A) Overview of virus-associated particle purification from strain BD651. Left: Mycelium homogenized in liquid nitrogen was suspended in isotonic solution and subjected to differential centrifugation at 1000, 8000, 20 000, and 100 000 g. Pellets (Ppt) and supernatants (Sup) were collected at each step and analysed. Right: The P100 000 pellet was subjected to continuous sucrose density gradient centrifugation (10%–60% w/v). Following ultracentrifugation, 2 ml aliquots were collected from the top, yielding 15 fractions. (B) PCR detection of HthCRESSV1 in the 15 fractions obtained by density gradient centrifugation. M, DNA size marker. (C) Transmission electron microscopy (TEM) of pooled fractions 4 and 5. Samples were negatively stained with 1% uranyl acetate. Scale bars, 50 nm.

3.5. Analysis of HtCRESSV1 mitochondrial localization

Exclusive association of HtCRESSV1-like sequences with oomycete mitochondria prompted us to verify the mitochondrial localization of HtCRESSV1. To this end, the crude mitochondrial fraction (P12 000) obtained by cellular fractionation was subjected to two-step sucrose density gradient centrifugation. Following centrifugation, 15 fractions (F1–F15) containing purified mitochondria were collected (Fig. 4A), total nucleic acids were extracted from each fraction and analysed by PCR with HtCRESSV1-specific primers.

Figure 4: Mitochondrial localization of HthCRESSV1. (A) The P12 000 pellet was subjected to discontinuous sucrose density gradient centrifugation (15%, 23%, 32%, and 60% w/v). For the first purification, fractions collected at the 32%/60% interface (approximately fractions 12–13 from the top) were pooled and subjected to a second round of purification. Fractions (2 ml each) were collected from the top of the gradient. Total nucleic acids were extracted, ethanol-precipitated, and resuspended in TE buffer containing 1 mg ml−1 RNase A. (B) Agarose gel electrophoresis of total nucleic acids from the second mitochondrial purification fractions. Total nucleic acid samples extracted from each fraction were subjected to agarose gel electrophoresis from left (lane 1; 10%) to right (lane 15; 60%). The arrow indicates the HtCRESSV1 band. The image is shown with inverted contrast. M, DNA ladder (C) and (D) PCR specific for HthCRESSV1 detection and measurement of Cytochrome c oxidase (COX) activity for the second mitochondrial purification fractions. (C) HthCRESSV1-specific PCR of total nucleic acid samples extracted from each fraction. M, DNA ladder, P.C., positive control; N.C., negative control (water). (D) Cytochrome c oxidase (COX) activity in fractions from the second purification. COX activity was measured using 100 μl from each fraction, combined with 950 μl of assay buffer and 50 μl of substrate solution (reduced cytochrome c) for a total volume of 1 100 μl. Absorbance at 550 nm was measured for 1 min, and the decrease in absorbance (ΔA) was recorded. Activity was calculated in units of μmol cytochrome c/min/ml using ΔA/min and the molar extinction coefficient difference of cytochrome c (21.8 mM−1 cm−1).

Agarose gel electrophoresis analysis of the total nucleic acids extracted from the 15 fractions revealed that fraction F8 contained the highest concentration of high-molecular-weight DNA, likely corresponding to the mitochondrial DNA. Notably, this fraction also contained a faint band of ~2500 bp in size, matching the expected size of the viral genome (Fig. 4B). Consistently, semi-quantitative PCR analysis showed that HtCRESSV1 genome was enriched in fraction F8 (Fig. 4C).

To evaluate the distribution of mitochondria in the gradient fractions, cytochrome oxidase (COX) activity, a mitochondrial marker, was measured in all 15 fractions. Strong COX activity was detected in fractions F7 and F8 (0.05 μmol cytochrome c/min/ml) indicating the presence of intact mitochondria at high concentrations in these fractions (Fig. 4D), suggesting that HtCRESSV1 localizes within mitochondria.

3.6. HtCRESSV1 is transmitted via zoospores

To verify whether HtCRESSV1 can be transmitted via zoospores we analysed single-lesions formed by zoospore-derived isolates. All 25 strains derived from zoospores isolated from single lesions formed on cork oak leaves retained HtCRESSV1 infection (Fig. 5A), indicating 100% zoospore transmission rate of the virus.

Figure 5: Biological effects of HthCRESSV1. (A–C) Comparison of growth between virus-infected (BD651) and virus-free H. thermoambigua strains (BD630, BD631, BD668). (A) Colonies grown on s. V8A medium for 3 days. From left to right: BD651 (virus-infected), BD630 (virus-free), BD631 (virus-free), BD668 (virus-free). (B) PCR detection of HthCRESSV1 in each strain. M, DNA size marker. (C) Colony diameters after 3 days of growth on s. V8A medium. Black bar represents the virus-infected strain BD651; hatched bars represent virus-free strains. The virus-infected strain showed significantly reduced growth compared to virus-free strains (*P < .01). (D) Comparison of growth rates of BD651, BD630, BD631, and BD668 at different temperatures (4°C, 8°C, 15°C, 20°C, 25°C, 28°C, 30°C, and 33°C). Growth rates were calculated as colony diameter per day on s. V8A medium. Symbols: circles, BD651 (virus-infected); triangles, BD630 (virus-free); diamonds, BD631 (virus-free); squares, BD668 (virus-free). Note that the optimal growth temperature was 28°C for BD651, 20°C for BD630 and BD631, and 25°C for BD668.

To obtain virus-free strains, treatment with the antiviral agents, ribavirin and cycloheximide, a method frequently employed for mycovirus clearance (Uchida et al. 2021), was performed. However, no virus-free strains were successfully isolated among the 20 oomycete strains regenerated from protoplasts (Fig. S5). Similarly unsuccessful was the attempt to obtain virus-free strains by single-zoospore isolation, an approach previously used to cure fungal hosts from the cressdnaviricot Diaporthe sojae circular DNA virus 1 (Wang et al. 2024).

3.7. Evaluation of host impact

To investigate whether HtCRESSV1 infection affects the growth of the host oomycete, similar to what has been reported during the SsHADV-1 infection of the fungus S. sclerotiorum, we compared the growth of naturally virus-infected strain BD651 and noninfected strains BD630, BD631, and BD668. First, mycelial colonies were compared after 3 days of incubation at 25°C on sV8A medium (prepared with 50% seawater). The colony diameter of the virus-infected strain BD651 was significantly smaller compared to the virus-free strains (Fig. 5B and C). The virus-infected strain also showed slightly higher hyphal density consistent with slower growth.

Next, we compared the daily hyphal extension rates (cm/day) at temperatures ranging from 4°C to 33°C. The optimal growth temperatures (temperatures with the highest hyphal extension rates) for BD651, BD630, BD631, and BD668 were 28°C (1.17 cm/day), 20°C (1.48 cm/day), 20°C (1.42 cm/day), and 25°C (2.07 cm/day), respectively (Fig. 5D). Thus, virus-infected strain BD651 had the highest optimal growth temperature. Notably, however, the virus-infected strain BD651 was characterized by slower growth rate compared to the noninfected strains at lower (suboptimal) temperatures in the range of 4°C–25°C. These results suggest that HtCRESSV1 infection negatively impacts growth rates of the oomycete across a wide temperature range, most likely reducing its ability to colonize plant tissues.