Two percent (95% CI 2.1–2.4%) of seronegative women will acquire primary cytomegalovirus (CMV) infection during their pregnancy (1). At birth, 10–15% of congenitally infected neonates will have symptoms; of those 6% will die and up to 90% will develop sequelae such as sensorineural hearing loss (SNHL), developmental delay, cognitive impairment, neuromuscular dysfunction (cerebral palsy), epilepsy, impaired vision function, and possibly autism spectrum disorder. Asymptomatic infants remain at significant risk of developing SNHL before 2 years of age. Congenital CMV infection is responsible for more long-term sequelae than either Down syndrome or Fetal Alcohol Spectrum Disorder (2). Zika virus (ZIKV) is a mosquito borne flavivirus that has also been causally implicated in brain injury and microcephaly following in-utero infection (3). The estimated risk of clinically detected central nervous system (CNS) injury following first trimester ZIKV infection ranges from 1 to 13% (4). In an epidemiologic study based on Brazilian data, the peak risk of microcephaly appears to correlate with ZIKV infection between gestational week 14 and 17 (5), but findings of ventriculomegaly and cerebral calcifications have been seen with ZIKV infections as late as the 3rd trimester (6).

Significant barriers exist to the effective and timely diagnosis and treatment of fetal brain injury secondary to in-utero viral infections such as CMV and ZIKV. Many maternal viral infections are asymptomatic; therefore, detection of infection would require surveillance programs. However, enthusiasm for costly surveillance programs has been dampened by a combination of challenges in the accurate diagnosis of fetal brain injury, lack of effective in-utero interventions such as hyperimmune globulin (7), and limited means to monitor treatment effects in real time. In fetuses destined to develop microcephaly, head circumference measured on prenatal ultrasound may be normal prior to 22 weeks gestation age (weeks GA) and fetal MRI is of limited utility in diagnosing other intracranial abnormalities prior to 18 weeks GA resulting in late diagnosis in the absence of early surveillance programs (8, 9). Finally, definitive fetal diagnosis requires amniocentesis with amniotic fluid viral PCR. Although emerging data suggests that earlier amniocentesis may be accurate, the traditional recommendation is for amniocentesis to be performed 6 weeks from maternal exposure and at >21 weeks' gestation for CMV (10). This calculus may be shifting as data emerges regarding the efficacy of other early anti-viral therapies. A recent small randomized, double-blind, placebo-controlled trial demonstrated 71% reduction in the fetal infection rate with high dose oral valcyclovir in women with a primary CMV infection in the first trimester (3–12 weeks gestation) (11). Valganciclovir treatment was also associated with resolution of hydrops fetalis and minimal sequelae in a recent case report (12). The missing element to any first trimester strategy for the diagnosis and treatment of in utero viral infections is a non-invasive first trimester assay similar to cell-free DNA based tests for aneuploidy. An ideal first trimester test would be able to detect fetal injury, especially neurologic injury, prior to seroconversion and have improved sensitivity. Serial testing would ideally be available to determine if treatment was effective and to distinguish between those fetuses at low and high risk for adverse neurologic outcomes prior to mid-trimester ultrasound findings.

Exosomes/extracellular vesicles (ECVs) are membrane-bound endosome-derived nanovesicles that allow the intercellular transfer of genetic material, cytoplasmic and membrane proteins, lipids, miRNAs and other intracellular materials (13). Exosomes freely cross the blood-brain barrier and the placenta through unknown mechanisms, carrying proteins and micro RNAs (miRs) into the peripheral blood while protecting them from degradation (14, 15). We have published a series of foundational investigations demonstrating novel techniques to isolate ECVS derived from the fetal CNS from maternal serum using the protein Contactin-2/TAG1 (CTN-2/Tag-1) that is expressed almost exclusively by developing neurons in the hippocampus, cerebral cortex and cerebellum (16– 18)¹. A small proportion of CTN-2/TAG-1 ECVs may arise from tissues such as His-Purkinje cells (19) but the relatively small contribution of these tissues is unlikely to hinder assay performance. Based on comparisons of concentrations between pregnant and non-pregnant subjects, we have demonstrated that 97% of CTN-2/TAG-1+ ECVs isolated from maternal serum are fetal in origin (16); the 3% contamination by maternal neural ECVs reflects low levels of ongoing neural plasticity in adults. We have demonstrated that the placenta is not a source of CTN-2/TAG-1 (17). Taken together, our work demonstrates that CTN-2/TAG-1 can be used to isolate fetal CNS derived ECV (CNSEs) from maternal blood to generate novel biomarkers. We hypothesized that protein markers of neurologic injury quantified in CTN-2/TAG-1+ ECVs purified from maternal blood could identify fetal CNS injury secondary to in-utero viral infection with CMV or ZIKV.

These exploratory data suggest that human fetal transmission of CMV can be detected non-invasively using fetal CNSEs isolated from maternal blood as early as 11 weeks gestation and that fetal CNS injury can be detected in a non-human primate model within 2 days of infection. Fetal CNSEs were useful at gestational day 53–55 in the non-human primate model described, which is roughly equivalent to the end of the first trimester in humans. Despite the sample size limitations, these data are the first to suggest that first trimester non-invasive diagnosis of fetal viral infection is possible. If confirmed, this finding could transform the paradigm of prenatal diagnosis for fetal viral infection. Our findings that human Contactin-2 Tag-1 antibody could be used to purify fetal CNSEs in a non-human primate model suggests that this methodology could be used to augment studies of perinatal viral infection in appropriate animal models. Of note, our samples likely represent a mixture of exosomes, other extracellular vesicles and some organelles and we did not attempt to confirm purity due to small sample volumes. However, as opposed to studies where purity is critical (mechanistic and functional exosome studies), biomarkers studies are judged on their ultimate test performance in discriminating between disease and non-disease states and purity is less critical. In future studies, if larger serum/plasma volumes are available, we plan to perform flow cytometry studies to investigate whether or not viral infection alters the proportion of exosomes that care Contactin-2 Tag-1 and/or CD-81.

One surprising finding from our human CMV study was that all fetuses with evidence of in-utero infection appear to have significantly reduced levels of injury markers. We hypothesize that all fetuses infected with CMV take an initial “hit” to the CNS but that some are able to rebound secondary to neuroplasticity, and that some are not and are born with neurologic sequelae. Therefore, the most important value of the early samples may be to discriminate between pregnancies with and without fetal transmission to determine which fetuses would benefit from potential interventions but would not provide sufficient information regarding prognosis. Serial sampling may be needed to discriminate between fetuses that are responding to treatment and/or recovering due to innate defenses and those that have ongoing neuronal injury. An addition benefit of fetal CNSE-based prenatal diagnosis is that both initial and serial sampling will be more acceptable to patients since amniocentesis is not necessary. Non-invasive testing would also negate the unavoidable delays in diagnosis using amniocentesis—where the highest sensitivity and specificity are reached after 21 weeks' gestation. At this gestational age, some CNS injury may be irreversible. If the positive and negative predictive value of CNSE-based prenatal diagnosis is confirmed, it may shift the cost-benefit analysis for CMV surveillance in pregnancy. A recent cost-benefit study concluded that surveillance might be cost effective if anti-viral interventions were more than 30% effective (29), however more precise targeting of those pregnancies that require intervention and serial evaluation would reduce excess costs of treatment and testing from unnecessary treatment.

In addition to markers of neuronal injury, CNSEs may carry viral products or markers of CMV infection that reflects viral specific mechanisms. The similarities between the biogenesis of extracellular vesicles and the release of viral particles from infected cells are intriguing as both processes involved the ESCRT pathway (30). The CMV virus utilizes EVs to enhance viral spread, therefore it is not surprising that EVs can contain viral products (31). CMV specific viral products include the viral envelope proteins gB and gH as well as CMV generated miRs (32, 33). We have attempted to detect gB in CNSEs in our prior CMV studies and have found the results to be inconsistent. A recent study has demonstrated that only 15% of EVs released from CMV infected cells were positive for gB, 5.3% were positive for gH and 3.74% were positive for both gB and gH. The factors that determine the proportion of viral protein positive EVs are not known. Therefore, viral proteins appear to be imperfect biomarkers of CMV infection (34). We were unable to isolate viral genetic products from fetal CNSEs, which was disappointing; we had been hopeful that CNS “viral load” could be estimated non-invasively.

Although viral activity may not be able to be measured directly in fetal CNSEs—there are other potential markers that may reflect perinatal outcomes. EVs from infected cells are known to contain soluble DC-SIGN, a C-type lectin family molecule that increases the susceptibility of recipient cells to CMV but DC-SIGN is found largely in dendritic cells rather than neurons (35). Another study examined total EVs isolated from neonates with active CMV infection. EV levels of the CMV miRs US25-1-5p and UL112-3p were significantly higher in the setting of active CMV infection and were significantly associated with the degree of liver damage secondary to viral hepatitis as assessed by neonatal liver function tests (36). In addition, when free plasma levels of miRs from 13 infants with congenital CMV were compared to controls, levels of miR 183-5p and miR-210-3p were significantly higher (37). This suggests that CNSE miR levels may also be potentially useful in assessing degree of neurologic injury. In addition to the candidates above, there are several other miRs and proteins that have some premise from the literature. Human miR-21 is known to inhibit CMV viral gene expression by targeting Cdc25a, a cell cycle regulator, in neural progenitor cells and may act as an innate anti-viral defense (38). BclAF1 restriction factor, a cellular anti-viral protein, is down regulated by CMV miR-UL112-1 and is also found in ECVs released from human brain cells (39, 40). CMV miR UL112-1 down regulates the antiviral protein Interleukin-32 (IL-32), aiding in evasion of the host immune response (41). In turn, IL-32 is an important regulator of TNF-α, which may adversely affect microglial-mediated neuroinflammation and potentiate fetal brain injury. While TNF-α has not widely been reported in exosomes, tumor necrosis factor-α 1 receptors (TNFR1) levels have been used as a biomarker of response to inhibitors of neuroinflammation in neuronal ECVs (42). Human miR-221 in neural precursor cells inhibits CMV replication by promoting type I interferon and upregulating the activation of NF-kB (43).

In summary, fetal CNSEs appear to have potential in the early identification of viral mediated fetal brain injury. Fetal CNSEs may be useful as early as 11 weeks gestation—and therefore may identify brain injury prior to irreversible damage. Fetal CNSEs offer an advantage as testing can be performed non-invasively, avoiding both the risk and the delay in diagnosis associated with amniocentesis. Fetal CNSEs may also be a useful adjunct to animal models of perinatal viral infection—where the timing of infection is more precisely known. However, larger studies are required to confirm the potential of fetal CNSEs both for endemic viral infections such as CMV and for episodic pandemic infections such as ZIKV and unknown future viral strains.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by IRB, Shaare-Zedeck Medical Center. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by IACUC, Oregon National Primate Research Center.

LG conceived of the scientific experiments related to exosome/ECVs as biomarkers, supervised the assays, and wrote the manuscript. AS and AM assisted in the scientific design, performed the Zika assays, and edited the manuscript. YS, YE-Y, and MH designed the CMV cohort study, collected biologic samples and outcome data, participated in the scientific design, and edited the manuscript. AH, DS, AF, VR, and NH designed and performed the primate studies, collected biologic samples and outcome data, participated in the scientific design, and edited the manuscript. ND and NM assisted in the scientific design, performed the CMV assays, and edited the manuscript. All authors agree to be accountable for the content of the work.

This study was supported by Bill and Melinda Gates Foundation OPP1119489 (LG) for development of non-invasive assessment of fetal neurodevelopment and for open access publication fees, National Institutes of Health: R01HD069238 (LG) for support of LG's laboratory and technician time, National Institutes of Health P51 ODO11092 (Haigwood) for support of the Oregon National Primate Research Center, National Institutes of Health R01 HD096741 (DS) for support of Non-Human Primate Experiments.

LG holds a patent for the methods for isolating fetal central nervous system extracellular vesicles/exosomes. This technology is not currently licensed for commercial use. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.