Impact of preexisting virus-specific maternal antibodies on cytomegalovirus population genetics in a monkey model of congenital transmission

Human cytomegalovirus (HCMV) infection is the leading non-genetic cause of congenital birth defects worldwide. While several studies have investigated the genetic composition of viral populations in newborns diagnosed with HCMV, little is known regarding mother-to-child viral transmission dynamics and how therapeutic interventions may impact within-host viral populations. Here, we investigate how preexisting CMV-specific antibodies shape the maternal viral population and intrauterine virus transmission. Specifically, we characterize the genetic composition of CMV populations in a monkey model of congenital CMV infection to examine the effects of passively-infused hyperimmune globulin (HIG) on viral population genetics in both maternal and fetal compartments. In this study, 11 seronegative, pregnant monkeys were challenged with rhesus CMV (RhCMV), including a group pretreated with a standard potency HIG preparation (n = 3), a group pretreated with a high-neutralizing potency HIG preparation (n = 3), and an untreated control group (n = 5). Targeted amplicon deep sequencing of RhCMV glycoprotein B and L genes revealed that one of the three strains present in the viral inoculum (UCD52) dominated maternal and fetal viral populations. We identified de novo minor haplotypes of this strain and characterized their dynamics. Many of the identified haplotypes were consistently detected at multiple timepoints within sampled maternal tissues, as well as across tissue compartments, indicating haplotype persistence over time and transmission between maternal compartments. However, haplotype numbers and diversity levels were not appreciably different across HIG pretreatment groups. We found that while the presence of maternal antibodies reduced viral load and congenital infection, it has no apparent impact in the intrahost viral genetic diversity at the investigated loci. Interestingly, some haplotypes present in fetal and maternal-fetal interface tissues were also identified in maternal samples of corresponding dams, providing evidence for a wide RhCMV mother-to-fetus transmission bottleneck even in the presence of preexisting antibodies. Author summary Human cytomegalovirus (CMV) is the most common infectious cause of birth defects worldwide. Knowledge gaps remain regarding how maternal immunity impacts the genetic composition of CMV populations and the incidence of congenital virus transmission. Addressing these gaps is important to inform vaccine development efforts. Using viral samples collected from a monkey model of congenital CMV infection, we investigated the impact of passively-administered maternal antibodies on the genetic composition of the maternal virus population and that transmitted to the fetus. Our analysis focused on two CMV genes that encode glycoproteins that facilitate viral cellular entry and are known epitope targets of the humoral immune response. By identifying and analyzing variants across sampled maternal tissues, we found no impact in CMV genetic diversity by preexisting CMV-specific antibodies, despite the observation that such antibodies reduce viral load and confer some protection against congenital transmission. We further found that some minor variants identified in fetal and maternal-fetal interface tissues were also present in corresponding maternal tissues, indicating that a large number of viral particles passed from dam to fetus in observed cases of congenital transmission.


149
RhCMV viral load was quantified from each sample using qPCR, as described in [9]. For 150 all samples, multiple viral load measurements (3 to 18) were taken to ensure that samples with 151 relatively low levels of virus present were identified as being positive for RhCMV. Viral load on 152 the log10 scale was calculated as the mean of the individual log10 viral load sample 153 measurements. When viral load was below the limit of detection (100 viral copies per ml for 154 plasma and amniotic fluid and 100 viral copies per total DNA µg for urine and saliva), we set its 155 value to half of the detection limit.  176 containment facilities was required. The monkeys were maintained in a standard environment 177 enrichment setting which included manipulable items, swings, food supplements (fruit, 178 vegetables, treats), task-oriented feeding methods as well as human interaction with caretakers 179 and research staff. Dams were released into the colony after 2 or 3 weeks following C-section.

180
Anesthesia was considered for all procedures considered to cause more than slight pain 181 in humans, including routine sample collection. The agents used included: ketamine, butorphanol, 182 Telazol, buprenorphine, carprofen, meloxicam, and midazolam as needed. The criteria for end-11 209 which performs an iterative process of removing spurious, low abundance sequence groups by 210 adding them to more abundant, genetically similar sequence groups when the genetic mismatch 211 between groups occurs at nucleotide positions with low quality.

212
To obtain a set of haplotypes and their frequencies for a given sample, we combined 213 identified haplotypes across technical replicates. Specifically, for a haplotype to be considered 214 present in a sample, we required it to be detected in both sample replicates. Haplotypes that did 215 not meet this criterion were merged with their genetically-closest haplotype in the sample, and 216 the count of this genetically closest haplotype in the sample was increased accordingly. When 217 only a single replicate was available, we could not perform this step and therefore kept all 218 identified haplotypes present in the single available replicate. When more than two technical 219 replicates were available, we restricted our analyses to the two replicates that were the most 220 similar to one another genetically, based on correlation of haplotype frequencies (see below).

223
We performed additional tests and required additional criteria to be met to ensure the 224 quality of each sample that would undergo subsequent analysis. First, to reduce the number of 225 spurious haplotypes in a given sample, we set a frequency threshold that sample haplotypes were 226 required to exceed. This threshold was established as 0.436% based on analysis of plasmid 227 controls. Specifically, we constructed two synthetic plasmids, one containing the gB gene and the 228 other containing the gL gene. Two technical replicates from each plasmid were sequenced using 229 the same protocol as for the RhCMV samples. Because a single haplotype should be present in 230 these plasmid control populations, any shared low-frequency haplotype is likely a product of PCR 231 amplification error or sequencing error. We found 19 minor haplotypes in the gB plasmid control 232 sample after merging technical replicates. These haplotypes ranged in frequency from 0.01% to 233 0.59% ( Figure S1). We found 29 minor haplotypes in the gL plasmid control sample after merging 234 technical replicates. These haplotypes ranged in frequency from 0.03% to 0.42% ( Figure S1).

235
Our chosen frequency threshold of 0.436% was set at the 0.95 quantile of the combined minor 236 haplotype distributions from the gB and gL plasmids.

237
As a second quality assurance step, we performed chimera detection on the haplotypes 238 in each merged sample. A haplotype was classified as a chimera if there was a combination of 239 partial alignments to two observed (and higher frequency) haplotypes in the same sample.
240 Detected chimeras were discarded. These chimeras contributed to only a small fraction of the 241 total reads in each sample (ranging from 1.95% to 7.82% of the reads across all samples).

242
As a third quality assurance step, we restricted our analysis to those samples that had a   Table S1 shows the final set of maternal tissue and amniotic fluid samples that were 248 included in our analyses, for both the gB and the gL loci. Table S2 shows the set of samples from 249 the maternal-fetal interface (other than the amniotic fluid samples) and from fetal samples that 250 were included in our analyses. In addition to these samples, the genetic composition of the 251 inoculum was analyzed. Each of the three viral stocks comprising the inoculum (UCD52, UCD59, 252 180.92) was independently sequenced. Two successfully sequenced replicates were available 253 for each of the three stock samples.

276 Results
277 Maternal viral load dynamics, congenital transmission, and strain dominance.

278
As previously described [9], dams in the high-potency HIG pretreatment group had 279 reduced peak viral loads in maternal plasma relative to dams in the control group following primary 280 maternal infection ( Figure S2A). Viral kinetics in the saliva and urine were also delayed in the

297
Across the majority of analyzed samples, we found that the dominant in vivo UCD52 298 haplotype was the canonical UCD52 reference haplotype of the viral inoculum. This was the case 299 both for the gB locus and the gL locus, and across all groups and compartments studied.

300
Our analysis of amplified sequences from the gB locus identified a large number of minor 301 haplotypes in maternal and fetal compartments that differed from the canonical UCD52 gB 15 302 haplotype by typically only a single nucleotide (Figure 1, Figures S4-S11). These minor 303 haplotypes ranged in frequency from just above the sequencing error cut-off frequency of 0.436% 304 up to 43.27%, with a median frequency of 0.80%. Maternal samples differed in the number of 305 identified gB haplotypes they contained, ranging from 1 to 33, with a median of 5 haplotypes per 306 sample. The number of haplotypes identified in a sample was not positively correlated with the 307 sample's viral load (Figure S12), indicating that the numbers of observed haplotypes were not 308 restricted by sample viral load. Given our constrained cut-off for haplotypes detection, these minor 309 haplotypes are potentially produced de novo as RhCMV spreads within each monkey. Within 310 individual dams, we observed that some of the minor haplotypes were shared across timepoints 311 from the same compartment and/or across compartments (Figure 1, Figures S4-S11). This 312 finding indicates that some of these minor haplotypes persist over a timespan of weeks in a given 313 compartment and that some of these minor haplotypes are likely transmitted across anatomic 314 compartments. Of the minor haplotypes that were shared across compartments, most were 315 shared between the plasma and one other compartment (Figure 1, Figure 2, Figures S4-S11).
316 This pattern may reflect plasma being a source of viral haplotypes for other compartments; 317 alternatively, it may simply be due to a larger number of plasma samples being successfully 318 sequenced relative to those from other compartments (Table S1). Interestingly, in 6 out of the 8 319 monkeys that had both urine and saliva sequences available, there were also minor gB 320 haplotypes that appeared to be shared exclusively between urine and saliva samples. These 321 haplotypes were generally found first in urine and then in a later week in the saliva, suggesting 322 potential oral auto-inoculation from virus shed in urine.

323
To assess whether the number of identified gB haplotypes differed by pretreatment group, 324 we calculated the median number of minor UCD52 haplotypes in each available tissue for each 325 of the 11 dams. We found no significant differences in the median number of minor gB haplotypes 326 by tissue across any pair of pretreatment groups (all Mann-Whitney U tests > 0.1; Figure 3A). To 327 determine whether certain tissues tended to allow more non-synonymous variation than other 16 328 tissues, or whether the extent of nonsynonymous variation differed by pretreatment group, we 329 further calculated the proportion of minor gB haplotypes that differed from the canonical haplotype 330 by a nonsynonymous mutation, by tissue and monkey. No major differences were found between 331 tissues or between pretreatment groups (for data on haplotypes, see Appendix S1).

332
The UCD52 haplotype patterns observed using the gL locus are consistent with those at 333 the gB locus. Specifically, minor gL haplotypes generally differed from one of the two dominant 334 gL haplotypes present in the inoculum by a single nucleotide (Figure S13-S22). Similar to the 335 frequencies observed for gB haplotypes, minor gL haplotypes were present at frequencies as low 336 as 0.44% and up to 48.16%, with a median frequency of 1.05%. Samples differed in the number 337 of identified gL haplotypes they contained, ranging from 2 to 29 with a median of 6 haplotypes per 338 sample. Again, no correlation was found between the number of haplotypes identified in a sample 339 and the sample's viral load (Figure S23). Some of the identified minor gL haplotypes appeared 340 to persist within the same tissue over time, and some were shared across tissue compartments. 341 Similar to our findings at the gB locus, most of the minor haplotypes that were shared across 342 compartments were shared between the plasma and one other compartment (Figure S24). Minor 343 gL haplotypes shared between urine and saliva compartments again suggested auto-inoculation.
344 Finally, consistent with the findings from the gB locus, the median number of gL minor haplotypes 345 observed in any tissue did not differ between pretreatment groups (Figure 3). We again found 346 no significant differences between tissues or pretreatment groups in the proportion of minor gL 347 haplotypes that were nonsynonymous (for data on haplotypes, see Appendix S2), consistent with 348 the lack of pattern at the gB locus.

349
We next assessed whether HIG pretreatment had an impact on RhCMV genetic diversity, 350 as measured by pairwise nucleotide diversity  for each sample's UCD52 viral population. Levels 351 of viral genetic diversity varied significantly between monkeys, compartments, and across weeks 352 (Figure 4 for gB, Figure S25

428
In our analyses, despite high-potency pretreatment monkeys having significantly lower 429 peak viral loads compared to standard and control group monkeys, we found no strong evidence 430 for a relationship between HIG pretreatment and maternal plasma UCD52 haplotype number or 431 nucleotide diversity. Together, these results suggest that preexisting antibodies can reduce 432 overall viral load but do not appear to restrict replication of specific UCD52 viral variants or limit 433 viral diversity. We also found no significant differences in saliva or urine virus haplotype number 434 or nucleotide diversity between the three groups at either gB or gL loci. While we previously 21 435 assessed and reported lower maximum plasma viral diversity levels in monkeys pretreated with 436 HIG compared to the control group [9], here, we included a more in-depth analysis across 437 timepoints to report the median viral diversity levels across monkeys, and did not find any lasting 438 impact of preexisting antibodies on maternal viral diversity.

439
Our identification of shared, minor UCD52 haplotypes between maternal plasma samples, 440 amniotic fluid, placental tissue, and fetal tissues is consistent with previous studies investigating 441 the population genetics of HCMV in newborns [12,32], which together point towards a loose 442 vertical transmission bottleneck between mother and fetus. While previous studies have 443 estimated transmission bottleneck sizes for HCMV and other viruses, in this study we were unable 444 to quantify transmission bottleneck sizes between mother and fetus due to: 1) low levels of 445 haplotype diversity and 2) haplotype frequencies near the limit of detection. Nevertheless, based 446 on the identification of minor UCD52 haplotypes across maternal, maternal-fetal interface, and 447 fetal tissues, our analysis suggests that diversity in a given tissue is generated through a 448 combination of multiple viral haplotypes being passed to that compartment, along with de novo, 449 local generation of viral mutations.  (Table S1). Second, this study did not employ full genome sequencing, but instead sequenced 467 only two gene regions (gB and gL) to allow for studies of virus population in samples with low viral 468 load. Subsequently, any effect of antibody selection over the non-sequenced regions of gB and 469 gL or over other viral proteins will not be observed. Third, because RhCMV is a DNA virus with a 470 low mutation rate, our conclusions were limited by the low levels of genetic diversity observed in 471 the samples. We also used highly conservative haplotype-calling and error reduction methods to 472 ensure that the haplotypes we identified were not false positives. As a result, however, we likely 473 excluded many true haplotypes, which reduced the diversity levels we characterized and limited 474 our ability to make inferences regarding transmission bottleneck sizes. Finally, given that our 475 animal model of congenital CMV transmission involves maternal CD4 + T cell depletion, which 476 results in consistent placental transmission, our results might not be applicable to 477 immunocompetent individuals.

478
Despite these limitations, however, we were able to conclude that minor haplotypes 479 persisted over time within single maternal tissue compartments and that these minor haplotypes 480 were occasionally shared between anatomic compartments. Moreover, there was not a strict 481 bottleneck for the viral major and minor haplotypes that appeared in placenta, amniotic fluid, and 482 fetal tissues. All these observations are consistent with those from human congenital CMV cases 483 [12,15,35-37]. Patterns of viral diversity within and across compartments, however, did not 484 appear to differ between HIG pretreatment groups. These findings indicate that, although potently-485 neutralizing CMV-specific antibodies can effectively reduce viral population size and prevent