Persistence and Intra-Host Genetic Evolution of Zika Virus Infection in Symptomatic Adults: A Special View in the Male Reproductive System

We followed the presence of Zika virus (ZIKV) in four healthy adults (two men and two women), for periods ranging from 78 to 298 days post symptom onset. The patients were evaluated regarding the presence of the virus in different body fluids (blood, saliva, urine and semen), development of immune responses (including antibodies, cytokines and chemokines), and virus genetic variation within samples collected from semen and urine during the infection course. The analysis was focused primarily on the two male patients who shed the virus for up to 158 days after the initial symptoms. ZIKV particles were detected in the spermatozoa cytoplasm and flagella, in immature sperm cells and could also be isolated from semen in cell culture, confirming that the virus is able to preserve integrity and infectivity during replication in the male reproductive system (MRS). Despite the damage caused by ZIKV infection within the MRS, our data showed that ZIKV infection did not result in infertility at least in one of the male patients. This patient was able to conceive a child after the infection. We also detected alterations in the male genital cytokine milieu, which could play an important role in the replication and transmission of the virus which could considerably increase the risk of ZIKV sexual spread. In addition, full genome ZIKV sequences were obtained from several samples (mainly semen), which allowed us to monitor the evolution of the virus within a patient during the infection course. We observed genetic changes over time in consensus sequences and lower frequency intra-host single nucleotide variants (iSNV), that suggested independent compartmentalization of ZIKV populations in the reproductive and urinary systems. Altogether, the present observations confirm the risks associated with the long-term replication and shedding of ZIKV in the MRS and help to elucidate patterns of intra-host genetic evolution during long term replication of the virus.


METHODS
Quantitative reverse-transcription PCR (qRT-PCR). RNA sample was extracted from 300µl samples of saliva, 400µl of urine or serum and 100 µl of semen, with NUCLISENS ® easyMag ® (BioMerieux) on the platform "off board"-pre lysed sample. Briefly, lysis buffer was added in a proportion of 3:1 to sample. The mixture was incubated for 10 minutes at room temperature and after this time we added 30 µl of magnetic silica and take the Easymag machine. The RNA was eluted in 40 µl of elution buffer and stored at-80 0 C until the time of use. All extracts were tested for the presence of human RNase P gene by RT-PCR to confirm sample quality. The PCR reaction was carried out with RNA from each sample with a set of primers and probes with FAM as dye reporter for the probe. Primers/probes for Zika virus (ZIKV), Dengue virus (DENV) and Chikungunya (CHIKV) were previously described (10,12,18) . All assays were performed using the AgPath-IDTM One-Step qRT-PCR reagents (Applied Biosystems). Succinctly, we used 5 µl of the RNA extracted in 1µl of the mix from primers/probe [10pM/µl] and 19 µl of the reagent mix from AgPath-IDTM One- Step RT-PCR kit following manufacturer's instructions. To ensure RNA integrity and sample quality, all extracts were also tested for human RNase P (RNP) gene by qRT-PCR. All extracts showed robust RNase P values ranging from cycle threshold (Ct) values, which ranged from 16.3 to Ct 33.5 (see Supplementary Appendix Table S1 for details) demonstrating the integrity of the material collected. Assays for DENV (18) and CHIKV (12) were performed as controls for co-infection, as they have been reported in Brazil during the ZIKV outbreak. IgM antibodies were detected using capture ELISA with a specific viral antigen for ZIKV (14) .  Table 5B -TS5). The probes were designed to cover the entire ZIKV genome and to encompass the genetic diversity present on GenBank on January 14, 2016. In total, 26 ZIKV sequences were used during probe design (TS5).

Virus isolation in cell
Extracted RNA was fragmented at 94˚C for 0-60 seconds and each sample was BEAST v1.8.3 (5) was used to estimate dN/dS and the rate of ZIKV evolution within the male reproductive system (MRS), HyPhy v2.3 was used to identify codons with evidence for positive, diversifying selection, and for samples with >50x average coverage, we examined intra host genetic variation using FreeBayes v1.0.2 (7) .
Genome assembly and analysis. Consensus genomes based on alignments to a reference were generated as previously described by Blackley (16) . Ancestral states that define each clade (i.e., synapomorphic changes) were then further investigated by replacing them onto structural models for ZIKV peptides built with YASARA energy minimization and structural alignments were carried and variations in structure due to the selected amino acid changes were recorded. The structures were then evaluated for linear and discontinuous epitopes using the online servers IEDB and BePRO (17) .
BEAST v1.8.3 (5) was used to estimate dN/dS and the rate of ZIKV evolution within the male reproductive system (MRS). For these analyses, we utilized the ORF sequences for the 14 genomes we assembled from semen samples from patient ZIKV17. ZIKV17 experienced the most prolonged MRS infection in this study. We partitioned sites into codon positions 1, 2 and 3 and used independent HKY substitution models. To estimate substitution rate, we included Γ (4)distributed rate heterogeneity and tested multiple combinations of molecular clocks (strict and relaxed with lognormally distributed rate categories) and coalescent priors (constant size, Bayesian Skygrid with 10 parameters and Bayesian Skyline with 10 groups). (3-5, 8, 9) We compared models using marginal likelihood comparison with path sampling and stepping-stone estimation approaches (1) . For estimating dN/dS, we used the renaissance counting (11,13) option with the simplest molecular clock and coalescent prior combination (strict and constant size). All of these Bayesian analyses were run for 100 million Markov chain Monte Carlo steps, sampling parameters and trees every 10,000 generations.

ZIKV EVOLUTION DURING PROLONGED INFECTION.
For patient ZIKV17, we obtained near complete genome sequences from fourteen sequentially collected semen samples and two urine samples. For patient ZIKV19 five semen and two urine samples yielded near complete genomes (Table S2). With the exception of a few low frequency insertions/deletions associated with homopolymer repeats (Table S3), patterns of genetic variation (iSNVs and consensus-level changes) were distinct between urine and semen samples from the same patient ( Figure 6; Table S3), 6 consistent with independent compartmentalization of ZIKV populations in the reproductive and urinary systems. Using time-structured phylogenies, we estimated the rate of evolution during the prolonged infection of the MRS in ZIKV17. Our estimates were highly consistent across multiple models and were indistinguishable from published rates for the entire ZIKV outbreak in the Americas ( Figure S3, Table   S4). ZIKV evolution within the MRS was dominated by synonymous substitutions, consistent with strong purifying selection across most of the ZIKV genome. We observed one nonsynonymous and six distinct synonymous substitutions in the MRS of ZIKV17 (dN/dS via robust counting = 0.06) and only a single synonymous substitution in ZIKV19. We also observed a significant difference between the ratio of nonsynonymous:synonymous variants present at different frequencies in samples with a high depth of coverage. Synonymous changes were most prevalent among variants that reached high frequencies (≥50%) during the course of infection, while nonsynonymous changes were more common in variants that were maintained at low frequencies ( Figure 5D; Fisher's exact test p-value = 0.01). This pattern is consistent with incomplete purifying selection acting on low frequency variants (15) . In addition,     (6) with the 95% HPD shaded in grey.