Characterization of early viral populations in infants acquiring HIV through perinatal and breastmilk transmission: A review of what is currently known and the gaps needed to be addressed to guide passive HIV immunization of breastfeeding infants

Abstract

Newborns represent only 1% of the population, yet HIV vertical transmissions represent 10% of all new infections globally, even though antiretroviral therapy (ART) has been shown to reduce the risk of vertical transmission to less than 2%. While vaccines still represent the most efficient and cost-effective intervention to eradicate new infections, HIV immunogens that can effectively elicit broad spectrum protection are still at least a decade away. In contrast, passive immunization with broadly neutralizing antibody (bnAb) combinations has the potential to provide a more immediate pathway to HIV prophylaxis. Early phase infant trials are underway to establish the safety and pharmacokinetics of bnAb combinations selected for their potency against viruses acquired via adult transmissions. However, specific characteristics and phenotypic differences of vertically transmitted viruses in infants compared to adult viruses remain uncertain, including their susceptibility to known bnAbs. We review the current knowledge of vertically transmitted HIV viruses, including their genetics and phenotypic features. Differences in immunity between adults and infants lead us to hypothesize that distinct selection and evolutionary pressures act on the virus at the time of transmission and during the early phases of infection, and these may in turn affect the choice of bnAb combinations needed for protection against vertical transmission of HIV.

Introduction

While antiretroviral therapy (ART) in pregnant and breastfeeding persons living with HIV (PLWH) has substantially decreased the incidence of vertical transmission [1], infants continue to be disproportionally affected. In 2023, an estimated 120,000 new pediatric HIV infections occurred, representing 10% of all new HIV infections globally, although newborns represent only 1% of the global population [2]. Persistence of pediatric HIV transmissions can be attributed to numerous factors, including poor access to care, late diagnosis of maternal infection, incomplete adherence to treatment, resistance to ART, and maternal infection during pregnancy or breastfeeding [1, 3]. We note here that not all birthing or lactating parents identify as female; however, for simplicity, we will refer to birthing and/or lactating parents as “women” and “mothers”.

In the absence of ART, the rate of vertical HIV transmission is estimated to be 30–45% among breastfed infants [3, 4]. Postnatal transmission via breastfeeding accounts for almost half of pediatric infections [5] and is correlated with the duration of breastfeeding [6], with an estimated transmission rate of 1% per month of breastfeeding [7]. A randomized, controlled trial conducted in Nairobi, Kenya, estimated that in the absence of ART the rate of HIV transmission via breast milk over the first 24 months of life was 16.2%, with the majority of infections occurring early during breastfeeding [8]. While replacing breastfeeding with formula feeding is associated with a substantial reduction in HIV acquisition risk in the absence of antiretroviral treatment [8, 9], breastfeeding has critical short- and long-term benefits for the infant, including reduced morbidity, reduced mortality, and improved growth and development. Additionally, there are benefits of breastfeeding for the mother, such as reduced postpartum bleeding, improved birth spacing, and reduced risk of breast and ovarian cancer [10].

High maternal viral load (VL) is strongly and directly associated with vertical transmission risk [11, 12] and infant disease progression [13]. Viral suppression attained through antiretroviral (ARV) therapeutic use during pregnancy and postnatally in conjunction with infant ARV prophylaxis has been shown to reduce the risk of vertical HIV transmission to less than 2% [14, 15]. However, a recent cross-sectional study of 10 sub-Saharan countries showed that ART coverage and adherence among pregnant and breastfeeding women remains sub-optimal [16]. In high incidence countries, newly diagnosed maternal HIV infections account for an increasing proportion of infant HIV infections [17, 18]. Therefore, interventions that either replace or complement daily maternal ART and/or infant ARV prophylaxis are needed to eliminate vertical transmission of HIV, including addressing the need for infant prophylaxis during pregnancy and breastfeeding. In addition, while several HIV candidate vaccines are under study, an effective vaccine is still likely a decade or more in the future. On the other hand, passively administered HIV broadly neutralizing antibodies (bnAbs) have shown promise for the prevention of sexually acquired HIV in adults. Notably, the landmark Antibody Mediated Prevention (AMP) trials [19] established that passive infusion of the CD4 binding site bnAb VRC01 can prevent acquisition of HIV viruses sensitive to this antibody, with an estimated vaccine efficacy of 75% against viruses with a VRC01 IC80 <1 μg per milliliter [20]. This result pointed to the need for bnAb combinations, especially with bnAbs that target distinct epitopes on the HIV envelope (env), to protect against the diversity of circulating HIV. Accordingly, dual and triple combination bnAb regimens are under early phase study in adults [21–26], including several studies within the HIV Vaccine and Prevention Trials Networks (HVTN/HPTN).

BnAbs as a prophylactic and therapeutic intervention against HIV are being evaluated in early phase trials in infants to establish their safety and pharmacokinetics. A recent prospective clinical trial conducted in Botswana tested the use of the dual bnAb combination VRC01LS and 10–1074 in children living with HIV [27] and found that infusion of two bnAbs may have prolonged virologic control following interruption of ART in infants with low initial viral DNA levels [27]. On the prophylaxis side, a phase I trial of single bnAb administration, either VRC01, VRC01LS or VRC07LS in HIV-exposed infants is underway (NCT02256631) [28, 29]. Initial results indicate that VRC01 and VRC01LS have a good safety and pharmacokinetic (PK) profile [28, 29]. Trials testing bnAb combinations are also under way: IMPAACT 2037 will evaluate one and two doses of PGT121.414.LS and VRC07–523LS in HIV-exposed infants (impaactnetwork.org/studies/impaact2037). PedMAB1/2 is evaluating VRC07–523LS and CAP256V2LS alone and in combination at birth and at age 3 months in HIV exposed infants. SAMBULELO is a phase 2 study evaluating VRC07–523LS in HIV exposed and infected newborns. And an additional IMPAACT study under development will evaluate VRC07–523LS, ePGT121v1-LS, and PDGM1400LS, alone and in combination, to HIV-exposed and unexposed infants. Most studies have evaluated subcutaneous administration; however, IMPAACT 2048 will also evaluate intramuscular administration.

Many stakeholders have called for additional programmatic investment in bnAbs to prevent vertical transmission of HIV [30, 31], highlighting attributes that make bnAbs especially suitable in the pediatric setting: the short window of HIV exposure, the existing infrastructure for getting infants into care during the first 1–2 years of life, and the small doses needed for infant vs. adult HIV prophylaxis, which may help overcome antibody production and cost barriers. Additionally, the potential for less frequent administration is advantageous compared to the daily dosing of ARVs [32], and may improve adherence. Investment in local capacity to manufacture novel biologics and immunotherapies will lend further support for the program. The advent of Nirsevimab, a long-acting monoclonal antibody to prevent infant severe disease caused by respiratory syncytial virus (RSV) infection, serves as a guiding example [33].

A barrier to advancing bnAb-based prevention of vertical transmission is the paucity of data on the phenotypic and genotypic characteristics of HIV viruses transmitted to infants. As discussed below, few panels of vertically transmitted viruses currently exist, and all were collected over a decade ago. Over the past decades, the genetic diversity of HIV has not only greatly expanded, with circulating recombinant forms (CRFs) alone globally increasing in prevalence by 8% between 2010 and 2021 [34], but have also become more resistant to autologous sera [35] and bnAbs [36, 37]. The AMP trials [20] demonstrated that bnAb susceptibility varies by clade and geography [36]. In the pediatric setting, where infants are exposed to both virus and antibodies from the pregnant or breastfeeding mother, vertically transmitted viruses may undergo additional selection pressures in addition to the genetic drift observed globally. Therefore, it is imperative to evaluate the sensitivity to bnAb neutralization of vertically transmitted viruses in the areas with the highest infection burden. A multi-disciplinary and global taskforce convened by the International AIDS Vaccine Initiative (IAVI) identified this as a key step necessary for advancing the infant bnAb program [30].

In this manuscript, we review the data available on vertically transmitted viruses, including risk factors for pediatric HIV infection, and identify gaps in knowledge that remain to be addressed. We also characterize known virus features as well as the diversity of published sequence data from transmitted lineages. We use the term “transmitted lineage” instead of the more common “transmitted founder” to acknowledge that these viruses are samples from larger virus populations, where each sequence set represents a genetic lineage or family that shares a unique common ancestor.

Sequence and Literature Selection

Because the focus of this review was characterizing HIV env sequences sampled from infants, one of the criteria we based our literature search was the availability of such sequences. Therefore, both literature and sequence searches were conducted together, using the sequence search interface available on the LANL sequence database (www.hiv.lanl.gov/components/sequence/HIV/search/search.html) as follows:

1. Under “Genomic region” select “Env CDS”.

2. Select “Patient Information -> Risk Factor -> Mother->Child”.

3. Submit query.

4. At the time of drafting the paper (June 2024), this search yielded 4,167 records.

5. We further selected “One record per patient” and obtained 271 records, sampled between 1985 and 2019.

6. Next, we manually reviewed all records and down selected based on the following criteria:

   1. Infant sequences had been sampled within one year from birth;
   2. most of gp120 was included in sequence (i.e. at least all five variable regions);
   3. sequence(s) were linked to a PMID, retrievable using the accession number(s).

7. Using the above criteria, from the sequence records we collected the PMIDs of 74 papers, of which only 30 were mother/infant studies.

8. Of the 30, 10 had full env sequences from infants, of which 6 were confirmed postnatal transmissions. These were from the following mother/infant cohorts: WITS, MPH, ZEBS, VTS, CH009, and BAN. From the references of some of these papers we later found two additional mother/infant cohorts: PACTG076 and NBT. See Table 1 for details and references.

9. Five out of the 30 papers had maternal sequences only.

Table 1.

Mother-infant study cohorts with published env sequences for infants acquiring HIV vertically. Note that only a small subset of the enrolled participants had env sequences generated.

	Participants (transmission mode)	Cohort Name	Subtype	Country	Years	Enrolled	Primary Reference (PMID)
Mother/Infant Pair Studies	Mother/Infant (in utero and intra-partum)	Pediatric AIDS Clinical Trials (PACTG 076)	CRF01_AE	Thailand	1991–1993	477 women, 415 infants	Connor 1994 (7935654)
	Mother/Infant (postnatal)	Nairobi Breastfeeding Trial (NBT)	A, C, A/D, C/D	Kenya	1992–1998	425 women, 410 pairs	Nduati 2000 (10703779)
	Mother/Infant (in utero and intra-partum)	WITS: Women and Infants Transmission Study	B	US	1988–2004	788 women, 657 infants	Landesman 1996 (8628356)
	Mother/Infant (in utero and intra-partum)	MPH: Malaria and HIV-1 in Pregnancy	C	Malawi	2001–2003	480 women	Mwapasa 2004 (15096809)
	Mother/Infant (postnatal)	Zambia Exclusive Breastfeeding Study (ZEBS)	C, G	Zambia	2001–2004	1,200 women	Thea 2004 (15296810)
	Mother/Infant (postnatal)	Vertical Transmission Study (VTS)	C	KwaZulu-Natal (South Africa)	2001–2006	2,722 women, 1,132 infants	Coovadia 2007 (17398310)
	Mother/Infant (postnatal)	CHAVI CH009	C	Malawi	2008–2009	41 women	Salazar-Gonzalez 2011 (21191008)
	Mother/Infant (postnatal)	Breastfeeding, Antiretrovirals, and Nutrition (BAN)	C	Malawi	2004–2010	2,369 women	Chasela 2010 (20554982)
	Mother/Infant (postnatal)	SAMBULELO (Phase II VRC07–523LS clinical trial)	C	South Africa	2024-	Ongoing	McFarland 2021 (34009371)
Infant/Child Only Studies	Perinatal infections only	Perinatal/LILAC (NISDI)	B, F, F1, F2	AR, BR, PE, MX, JM	2002–2011	922 infants	Hazra 2009 (19036797)
	Infants, children, and adolescents	Pediatric/PLACES (NISDI)	B, F, F1, F2	AR, BR, PE, MX, JM	2007–2011	1,629 children	Hazra 2009 (19036797)
	Infants only, perinatally infected, early ART	Children with HIV Early Antiretroviral (CHER)	C	South Africa	2005–2013	377 infants	Cotton 2013 (24209829)
	Infants only, perinatally infected/exposed, early ART	TARA (Toward AIDS Remission Approaches)	C	Maputo (Mozambique)	2017–2020	89 infants	Lain 2022 (36360495) Dinh 2023 (38045254)

Sequence Data

Analyses in this study used 139 full env nucleotide sequences from 13 infants and 226 from 13 mothers from Zambia [38], GenBank accession numbers: GU939124–GU939142, HM036739–HM037037, KY229265–KY229682; 195 V1–V5 env region sequences from 9 infants from Malawi [39], GenBank accession numbers: JN108036–JN108067, JN108203–JN108257, JN108284–JN108383, JN108481–JN108520, JN108549–JN108649, and JN108670–JN108760; and 119 V1–V4 env region sequences from 7 infants from Kenya [40], GenBank accession numbers: AY174897–AY175103. All infants tested negative at birth, and all were breastfed. The infants from Zambia and Malawi tested positive at 6 weeks of age, with no other testing done between birth and week 6, and were considered intra-partum infections, although transmission via breastmilk could not be excluded. In addition to testing negative at birth and positive at week 6, the 13 infants from Zambia also tested negative at 1 month of age and were therefore considered breastmilk infections. While for some infants sequencing from later time points were also available, these analyses only considered sequences sampled at 6 weeks of age. All studies were conducted prior to the availability of ART, however, women and their newborn infants in the Malawi cohort received single dose Nevirapine. In the Zambia cohort, all women received Nevirapine and all infants were given cotrimoxazole prophylaxis from 6 weeks to 12 months of age.

Env Diversity and Mutation Rate Analysis

Sequences were aligned using the LANL tool Gene Cutter (www.hiv.lanl.gov/content/sequence/GENE_CUTTER/cutter.html) and then manually revised to correct for software artefacts. Phylogenetic trees were obtained using IQ Tree software with ultrafast bootstrap branch support [41] and Highlighter plots (left) were obtained using the Highlighter tool from the LANL database (www.hiv.lanl.gov/content/sequence/HIGHLIGHT/highlighter_top.html). Viral diversity was measured via mean pairwise Hamming distances, defined as the number of mutations between every sequence pair and averaged over the total number of sequence pairs within each sample, per nucleotide site. Highlighter plots for each infant sample were used to identify the presence of recombinants and multiple transmitted lineages. When present, recombinants were excluded from the diversity analysis. When multiple lineages were detected, Hamming distances were calculated within lineages.

Fitting a Poisson model

In order to estimate the early viral mutation rate in infants, we used the LANL tool Poisson Fitter (www.hiv.lanl.gov/content/sequence/POISSON_FITTER/pfitter.html) using the inferred consensus as reference sequence and an estimated time of infection of 42 days for the infants from the Kenya and Malawi cohorts, and 14 days for the infants from the Zambia cohort. The model and methods for the Poisson Fitter tool are described elsewhere [42, 43].

Differences between Adult and Pediatric HIV Infections

There are immune features unique to the vertical transmission context that need to be taken into consideration when evaluating the phenotypic and genotypic characteristics of the transmitted virus lineages. HIV acquisition by infants occurs in the presence of antibodies derived from the pregnant or lactating mother, and infants are exposed to viruses that have evolved and adapted to the maternal immune system and HLA repertoire. The latter overlaps with the infant’s HLA repertoire and hence, at least in part, shapes the infant’s immune responses [13, 44]—an important difference from adult transmissions, where the likelihood of sharing HLA alleles is much lower. Indeed, mother-infant HLA concordance has been found to increase the risk of vertical transmission [45], and, additionally, certain maternal HLA types and/or mutations have been also associated with an increase in vertical transmission risk [46, 47]. In adults, transmitted viruses tend to be more infectious compared to viruses sampled from chronic infections [48] but are generally more sensitive to autologous plasma from the transmitting partner [49]. On the other hand, in vertical transmissions several studies have found that viruses transmitted either perinatally or via breastfeeding are generally resistant to parental autologous plasma [50–53] compared to non-transmitted parental variants, although other studies have also found no difference [39].

The immune landscape in infancy is different than in adults, likely affecting the viral genetic bottleneck as well as post-infection virus evolution and disease prognosis. Compared to adults living with HIV, young children living with HIV tend to have higher CD4 T cell counts, higher viral load peaks and set points, and higher rates of viral replication [54–56] [57]. There are also notable differences in the nature and dynamics of autologous antibodies in pediatric HIV: a higher proportion of infants and children living with HIV generate bnAbs compared to adults [58, 59], and infants with HIV generate bnAbs faster than adults [58]. Infant bnAbs may therefore be induced more rapidly compared to adults [60] [61] and indeed they tend to be less mutated than those in adults [60].

The three possible routes of vertical transmission —in utero, intra-partum, and postnatal—involve biologically distinct mechanisms. In utero acquisition of HIV is believed to occur primarily through the placenta [62], whereas intra-partum infection likely occurs either through transplacental microtransfusion or viral exposure during passage through the birth canal and subsequent absorption of the virus through the infant’s digestive tract [63]. On the other hand, postnatal transmissions via breastmilk are unique in that infants are exposed to both cell-free and cell-associated virus in the oral and gastroenteric tract [62, 64]. Multiple studies have shown that, while there is viral exchange between plasma and breast tissue [65–67], limited compartmentalization in the breast occurs, leading to local replication and clonal amplification of nearly identical viral quasispecies in the mammary gland [66–70]. Therefore, it is possible that multiple yet very similar transmitted lineages initiate infection in the infant.

In summary, compared to sexual HIV transmission, vertical HIV transmission is unique: viruses replicate in the breast milk compartment, which comprises cell-free and cell-associated HIV particles; the infant immune landscape is shaped by exposure to maternal antibodies and virus and potentially shared HLA alleles, and infants produce antibodies faster and with fewer mutations. These differences may mean that a bnAb combination selected to prevent sexual HIV transmission may have different efficacy at preventing vertical transmission.

Passive Immunization via Pregnancy and/or Lactation.

Maternal immunization has been used as an effective means to protect the mother, the fetus, and the infant during the first few months of life against several pathogens [71]. Immunization of pregnant women living with HIV could potentially provide benefits to both the mother and the baby, as maternal antibodies are transferred across the placenta and to the gastrointestinal tract during breastfeeding, which could contribute to prevention of vertical HIV transmission. This strategy for blocking transmission to the infant is potentially easier than immunizing the infant against the entire diversity of circulating HIV strains, as the enhanced maternal immunity would only need to effectively target the autologous maternal viruses to block transmission. In fact, autologous virus neutralizing antibodies have been achieved by current HIV vaccine candidates [72] and could leverage the phenomenon of immunologic imprinting by raising humoral immune responses against the original infecting HIV strain using a targeted immunogen [73].

Two phase I clinical trials conducted by the AIDS Vaccine Evaluation Group (AVEG), AVEG104 and 102, evaluated the safety and immunogenicity of recombinant HIV Env subunit vaccines administered to pregnant women living with HIV [74]. This gp120-based vaccine strategy did not enhance the ability of the plasma of pregnant women to neutralize autologous viruses compared to those that received placebo immunization [75]. However, in a non-human primate model of maternal immunization of Simian-Human Immunodeficiency Virus (SHIV) infected dams, heterologous gp120 immunization led to enhancement of binding antibodies that dominantly recognized the Env of the originally infecting SHIV, signaling that antigenic seniority was leveraged by this vaccination of previously infected animals [73]. Thus, future perinatal HIV vaccine studies could further explore a strategy to boost maternal immune responses specifically against the virus that the infant is exposed to and prevent vertical transmission. While this is promising as a form of infant passive immunization, maternal vaccines have major challenges. There is emerging evidence that antibodies are not transferred equally across the placenta [76–78], and the extent and type of antibody responses transferred can differ by antigen [79]. Additionally, and partly related to this antibody “placental sieve”, viral escape from autologous virus neutralization needs to be considered to ensure complete transmission prevention and avoid selection of pre-existing resistant strains [51]. Recent studies indicate that antibody glycosylation may impact IgG transplacental transfer [77–79]. Therefore, a maternal vaccine would need to induce antibody specificity and glycoforms that are not only able to fully block the maternal viral population, but also can be efficiently transferred across the placenta.

HIV Env Gene Sequences from Mother-Infant Cohorts

Vertical transmissions happen over a known window of exposure. In theory, a prospective study of infants at risk of vertical transmission that applied frequent HIV testing during this exposure window would afford the unique opportunity to: (i) identify characteristics of vertically transmitted lineage(s), and (ii) design and evaluate targeted interventions, such as passive immunization with bnAbs to which most transmitted variants are susceptible. However, historically such studies have been difficult to carry out, and, to date, few studies have sequenced and characterized vertically transmitted viral lineages. Current GenBank entries of full env sequences in particular, where bnAb escape mutations are found, have been obtained from mother-infant cohorts that date from 2010 or earlier, and most include only clade C viruses (Table 1). While subtype C is still the most prevalent subtype globally [80], other clades such as A, D, G and many CRF forms are found at a frequency of 9% or higher in countries like Kenya, Nigeria, and the DRC, where vertical transmission rates are still unacceptably high [81–85].

Two historical mother-infant cohorts contributed insights into the biology and diversification of vertically transmitted HIV viruses in the pre-ART era: the Nairobi Breastfeeding Trial [8], conducted in Kenya from 1992 to 1998, and the Women and Infants Transmission Study (WITS) [86], conducted in the US from 1988 to 2004. The former cohort led to the publication of subtype A, C, and recombinant A/D and A/C env sequences from in utero, intra-partum, and confirmed breastmilk transmission pairs [52, 53] (Table 2), whereas env sequences collected from mother-infant intra-partum transmission pairs enrolled in the WITS cohort were all subtype B (51] [87]. Additional clade B full env sequences sampled in the early 1990s (1990–1995) were published by Kishko et al. [88] from 5 mother-infant pairs, all infected at birth (Table 2). Two of these mothers had been treated with ZDV at the time of delivery.

Table 2.

Subtype, cohort, publication year and reference, transmission route, number of transmission pairs, median number of env sequences and subtype for the studies listed in section 2.

Subtype	Reference (PMID)	Year	Transmission Route	No. Mother/ Infant Pairs	Median No. Sequences per Infant (Range)	Cohort (Country)
A, C, D	Verhofstede (12584330)	2003	PP or BM	14	20 (4–36)	Kenya
A, C, A/D, C/D	Wu (16378985)	2006	PP or BM	8	3 (1–5)	Nairobi BF (Kenya)
A, C, A/D, C/D	Rainwater (17346133)	2007	PP, IP, BM	4	3 (1–5)	Nairobi BF (Kenya)
C	Kwiek (18427205)	2008	IU, PP	2	1, 9	MHP (Malawi)
C	Samleerat (21593171)	2011	IU, PP	19	27 (11–43)	MHP (Malawi)
CRF01_AE	Russell (18700833)	2008	IU, PP	17	11 (6–15)	PACGT 076 (Thailand)
A, C	Baan (21916748)	2012	IU, PP	7	12 (10–17)	Rwanda
C	Russell (23075434)	2013	BM	3	4 (2–6)	MHP (Malawi)
A, C, A/D, C/D	Mabuka (23856624)	2013	BM	12	2 (1–5)	Nairobi BF (Kenya)
C	Danaviah (25793402)	2015	BM	11	12 (3–16)	VTS (South Africa)
C, G	Nakamura (28122636)	2017	IU, BM	22	12 (4–23)	ZEBS (Zambia)
B	Kumar (29672607)	2018	PP	16	27 (15–44)	WITS (US)
B, C	Martinez (32156815)	2020	IU, PP	4	25 (20–47)	WITS/CH009 (US/Malawi)
B, C	Marichannegowda (34337555)	2021	IU	12	33 (7–53)	WITS/CH009 (US/Malawi)

Transmission route abbreviations: PP = peripartum, IU = in utero, and BM = breastmilk.

Via GenBank search, we found two additional sets of non-subtype C env sequences obtained from vertical transmission pairs, both collected from clinical trials conducted in the 1990s. Samleerat et al. [89] published gp120 CRF01_AE sequences from 17 mother-infant pairs infected perinatally (either in utero or at birth). All mothers had been enrolled in PACGT 076, a phase III randomized clinical trial part of the Pediatric AIDS Clinical Trials Group (PACTG) clinical trial network conducted in Thailand from 1991 to 1993 [90]. Verhofstede et al. [40] published subtype A1, C, and D partial env sequences (regions V1-V5) from 14 mother-infant pairs enrolled in the nonintervention arm of a prospective study conducted in Mombasa, Kenya, from 1996 to 1999 [91].

The Zambia Exclusive Breast-Feeding Study (ZEBS) [92] was conducted in Zambia between May 2001 and September 2004 and enrolled close to 1,500 pregnant women living with HIV. Mother-infant pairs received single-dose nevirapine (NVP). Nakamura et al. [38] sequenced full-length envs (Table 2) from 22 mother-infant transmission pairs enrolled in the ZEBS cohort, of which 13 were confirmed breast milk transmissions, six were in utero transmissions, and three were undetermined.

Three mother-infant cohort studies were conducted in Malawi throughout the first decade of the 2000s, all of which provided env sequences from in utero and peripartum transmissions, and two of which sequenced infant viruses acquired via breastmilk. The first cohort was the Malaria and HIV in Pregnancy study (MPH), a cross-sectional study of pregnant women living with HIV with and without placental malaria, conducted from 2001 to 2003 [93]. All sequences collected from other-infant pairs enrolled in the MPH cohort were subtype C [39, 94].

Also from Malawi, the CHAVI 009 cohort study was conducted between 2008 and 2009 and enrolled 41 lactating women chronically infected with subtype C HIV [67]. Full-length env sequence analysis from 12 women allowed the identification of the transmitted lineages from seven infants who acquired HIV postnatally [95]. This study also highlighted the importance of sampling mother-infant viral populations as close as possible to the time of transmission in order to make the most accurate inference of the vertically transmitted viruses. For one infant, the identification of the transmitted lineage was unambiguous as both mother and infant were sampled at the same time point, which was also the first time the infant tested positive, and one of the maternal sequences was identical to the transmitted lineage found in the infant. However, this was not the case for a second infant for whom sequencing was not done until 12 months of age, despite testing positive at 6 months of age. In six months the infant and maternal viral populations had significantly diverged, and the best inference of the transmitted lineage was done by selecting the closest maternal sequence via phylogenetic analysis [95].

The third and most recent mother-infant cohort from Malawi was the Breastfeeding, Antiretrovirals, and Nutrition (BAN) cohort, conducted between 2004 and 2010 [96, 97]. A final relevant study is the Vertical Transmission Study (VTS) [98], which enrolled pregnant women with and without HIV in KwaZulu Natal, South Africa, between 2001 and 2006. C2–V5 env region sequences were obtained from the infants’ dried blood spot samples (DBS), all of whom had acquired HIV via breastfeeding [68].

In summary, HIV env sequences for infants who acquired HIV vertically are few in number and inadequately representative: they date from more than 10 years prior, do not represent all circulating subtypes, have uncertain attribution in terms of acquisition route, and, across all studies that published env sequences, fewer than forty sequences from transmitted lineages were from confirmed breast milk transmissions (Table 2).

Env Mutation Rates and Viral Diversity in Early Infant Infection

The immune selection pressure in the infant and the subsequent adaptation from the mother to the infant’s immune environment play a role in the evolution of vertically transmitted viruses. In particular, vertically transmitted viruses replicate in an immune environment for which they have already been, at least partially, pre-selected: this can happen by escaping maternal autologous antibodies that are also passed on to the infant, and by having adapted to immune responses shaped by maternal HLA genes that are also, at least in part, shared by the infant. Antibodies passively acquired from the mother may further shape viral evolution in the infant. Additionally, compared to adults, infants and children who acquire HIV tend to have higher viral load peak and set point, even after ARV initiation [54–56] [57]. In non-human primate (NHP) models of neonatal infection, higher viral loads correlated with higher viral diversity, suggesting higher mutation rate compared to adults [99] [100]. On the other hand, mathematical modeling suggests that replication rate in children is comparable to that observed in adults [101].

Over the past decades, several studies have measured the evolutionary rate of HIV in infants. Among studies that focused on the rapidly evolving env gene, early reports that sequenced the env C2-V5 region in infants from the US indicated the presence of highly homogeneous viral populations within the first two months of infection [102]. Subsequent studies that studied longitudinally sampled env sequences from vertically infected infants did not always report consistent findings. However, between-study comparisons are challenging due to differences in measures of viral evolutionary rates and/or divergence across studies. To fairly compare the estimated evolutionary rates in env found in the literature, when available, we downloaded the infant sequences and performed the same analysis across all studies. All together, we considered 139 full-length env sequences from 13 infants from Zambia [38]; 195 V1-V5 env region sequences from nine infants from Malawi [39]; and 119 V1-V4 env region sequences from seven infants from Kenya [40]. All sequences were from the infants who tested negative at birth. Subtype, country, and number of sequences per infant are outlined in Table 3.

Table 3.

Env characteristics and estimated mutation rates from studies of infants acquiring HIV vertically. Sequences were obtained from three published studies (38–40) according to the following criteria: all infants tested negative at birth and positive at one month of age or later; a region of roughly 1,000 or more nucleotides of env was sequenced from these infants.

Study	Infant ID	Subtype	Country	Env Region	No. Seq.	Last HIV Negative (All Positive at 6 Weeks)	Mean %Diversity**	Estimated Mutation Rate***
Verhofstede et al.	IP1	A	Kenya	V1-V4	11	Birth	0.05	5.3×10−6
	IP2	A/D	Kenya	V1-V4	17	Birth	NA	NA
	IP3	A	Kenya	V1-V4	21	Birth	0.09	8.5×10−6
	IP4	A	Kenya	V1-V4	16	Birth	0.05	4.5×10−6
	IP5	A	Kenya	V1-V4	21	Birth	0.61	NA
	IP6*	A	Kenya	V1-V4	16	Birth	0.14	1.3×10−5
	IP7	A	Kenya	V1-V4	20	Birth	0.12	1.1×10−5
Russell et al.	312*	C	Malawi	V1-V5	32	Birth	0.37	NA
	819*	C	Malawi	V1-V5	40	Birth	0.06	NA
	874	C	Malawi	V1-V5	31	Birth	0.2	2.0×10−5
	1100	C	Malawi	V1-V5	16	Birth	0.01	1.2×10−6
	1846	C	Malawi	V1-V5	36	Birth	0.19	1.8×10−5
	1945	C	Malawi	V1-V5	36	Birth	0.06	5.9×10−6
	2038*	C	Malawi	V1-V5	43	Birth	0.09	8.7 ×10−6
	2684*	C	Malawi	V1-V5	26	Birth	0.06	7.8×10−6
	2909	C	Malawi	V1-V5	32	Birth	0.09	1.3 ×10−5
Nakamura et al.	1B	C	Zambia	gp160	15	1 month	0.38	3.7 ×10−5
	2B*	C	Zambia	gp160	10	1 month	0.73	NA
	3B	C	Zambia	gp160	16	1 month	0.15	1.4×10−5
	4B	C	Zambia	gp160	6	1 month	0.36	3.4×10−5
	5B	C	Zambia	gp160	16	1 month	0.42	4.1×10−5
	6B	C	Zambia	gp160	12	1 month	0.25	2.4×10−5
	7B	C	Zambia	gp160	15	1 month	0.33	3.1×10−5
	8B	G	Zambia	gp160	14	1 month	0.25	2.4×10−5
	9B	C	Zambia	gp160	12	1 month	0.24	2.3×10−5
	10B	C	Zambia	gp160	6	1 month	0.26	2.5×10−5
	11B*	C	Zambia	gp160	6	1 month	1.00	NA
	12B	C	Zambia	gp160	7	1 month	0.09	8.5×10−6
	22B	C	Zambia	gp160	4	1 month	0.13	1.3×10−5

^*Multiple lineages detected (analyzed separately when over 3 sequences, otherwise minor lineages excluded from analysis)

^**Defined as the mean percent number of mutations across all sequence pairs within an individual. Calculated on the most abundant lineage when multiple lineages were detected. NA when it was not possible to isolate recombinants.

^***Defined as the number of new mutations from the founder lineage, per nucleotide, per day of infection. Calculated when diversity followed a Poisson distribution, assuming an infected time of 42 days (since birth) for the infants who last tested negative at birth and 14 days for those who last tested negative at 1 month. NA when sample did not fit a Poisson distribution.

The infant cohort from Zambia was the group that overall had been infected for the least amount of time, since all infants tested negative at one month of age. Yet this cohort exhibited the highest within-lineage median diversity, 0.26% (range 0.09%−1%], compared to the other two cohorts whose median diversity was 0.1% for both (range 0.01%−0.37% for the Malawian infants, and 0.05% and 0.7% for the Kenyan ones)—less than half that of the Zambia cohort (Table 3). Additionally, only four of 13 infants in this cohort had at least one sequence that was identical to the inferred consensus sequence. In all other infants, all sequence pairs differed at one or more sites.

We measured the HIV evolutionary rate in these infants by estimating the env mutation rate under a Poisson model of random accumulation of mutations [42, 43] and assuming an infection time of 6 weeks for the infants who last tested negative at birth, and 2 weeks for those who last tested negative at 1 month of age (the time from the last negative test). The assumptions of such models were usually met in early infection, within the time window when these infants had acquired HIV. However, there were two infants in each cohort whose env diversity either did not conform to a Poisson distribution or there were too many recombinants to reliably subset the sequences into distinct lineages. For the remaining infants, the estimated evolutionary rate was between 1.2×10⁻⁶ and 1.2×10⁻⁴ mutations per site per generation cycle, with the highest values from the Zambian cohort. The median rate for the Malawian and Kenyan cohorts was 8.5 ×10⁻⁶ and 8.7 ×10⁻⁶ respectively, and 7.6 ×10⁻⁵ for the Zambian cohort (range 2.5 ×10⁻⁵ – 1.2×10⁻⁴ mutations per site per generation cycle; Table 3).

In adult acute infections, the estimated mutation rate is of the order of 10⁻⁵ [103–106], while a mutation rate of the order of 10⁻⁴ is on the high end of the plausible spectrum [107]. Under a model of random accumulation of mutations in the very early phase of infection, Keele et al. [42] estimate it at 2.16×10⁻⁵, which was derived from Mansky and Temin [104] after excluding APOBEC mutations and recombinants. The estimated mutation rates from the Kenyan and Malawian infants are at most half of this value (Table 3). Notably, these rates would be underestimated if the infants had been truly infected post-birth, rather than at birth, which is possible given that all infants had been breastfed. However, looking at the infants from Zambia, all of whom tested negative at birth and at 1 month of age and therefore were most likely to have acquired the virus via breast milk, only one had an estimated mutation rate about the same as the adult estimate, while all others yielded from double to one order of magnitude higher estimates (Table 3). Among six infants, five of whom had a higher than adult mutation rate, no single sequence was identical to the overall consensus, suggesting that the common ancestor of the viral lineage occurred in the maternal population.

While the mutation rate estimates from the Zambian cohort trend higher compared to similar estimates from acute adult infections sampled within 2–4 weeks from infection, they are compatible with a scenario where multiple low diversity viral variants, originated from a common ancestor in the mammary gland, are then transmitted to the infant via breast feeding. In such a scenario the most recent common ancestor originated in the maternal population, not the infant, and the increased diversity from the mutations accumulated during replication in the mammary gland, rather than post-transmission, artificially inflating the estimates of the evolutionary rate.

Taken together, these results seem to indicate that intra-partum infections have lower evolutionary rates compared to adult acute infections. On the other hand, the viral population of early breast milk transmission can be more complex, with multiple low diversity variants that seed the infection. This causes the Poisson model to yield over-estimated mutation rates since the high similarity between variants makes it difficult to separate the distinct transmitted lineages. Finally, we should note that while variation across individuals is expected, additional factors such as disease progression and severity of immunosuppression have been found to affect evolutionary rates in infants [13, 102, 108, 109].

Genetic Characterization of Vertically Transmitted Viruses

We have mentioned the importance of characterizing the phenotypic features of the transmitted lineages that initiate HIV infection across the different transmission routes. In infancy, the window of exposure is known, and under ideal conditions frequent sampling from birth throughout lactation would allow for early sampling of the virus and higher precision in identifying the actual transmitted lineages. With less frequent sampling, the chance of sampling the virus early in infection is lower, and the longer the time from infection, the higher the viral divergence from those initial founders. Vertically transmitted viruses undergo a genetic bottleneck that results in a more homogeneous viral population compared to the virus circulating in the mother, and, in subtype C infections, vertically transmitted viruses have less glycosylation sites in env and shorter variable loops compared to maternal variants [53, 95, 110, 111]. While this genetic bottleneck has also been observed in sexual transmissions [42], the different biological pathways involved in vertical transmissions suggest that different strategies need to be employed when identifying transmitted lineages. For example, in the 13 established breast milk transmissions published by Nakamura et al. [38] the env mean percent diversity ranged from 0.1% to 1%. This is significantly lower than the diversity measured in the paired maternal viruses (1%−4.5% range, p=0.0002 by paired Wilcoxon test; Fig. 1). In sexual transmissions sampled within a similar exposure window, and with similarly low viral diversity and comparable number of sampled sequences, the majority of infections were found to have been initiated by a single transmitted lineage [42, 112]. This was demonstrated by the fact that early viral diversity, within the first two months from viral exposure, increased over time following a Poisson distribution [42, 43]. In contrast, when looking at the breastmilk transmission pairs from Nakamura et al. [38], while all had low viral diversity, eight out of 13 infants had one or more subgroups of sequences that either shared one or more mutations at a few distinct sites from the rest of the sequences, or a higher number of mutations than expected by random variation alone. We chose mother-infant pairs 3 and 5 to illustrate these two different viral patterns in Fig. 2: both infants clearly show much lower diversity compared to the maternal sequences, however, in infant 3 we see a bulk of identical sequences and randomly scattered mutations across the remainder of the sequences (Fig. 2A), whereas no two sequences are identical in Infant 5. In fact, if we were to calculate a consensus sequence out of this sample, it would not match any of the actual env sequences found in this infant (Fig. 2B).

Fig. 1: Mean Percent Viral Diversity in Breastmilk Vertical Transmissions from 13 Mother-infant Pairs in the ZEBS Cohort.

Mean pairwise Hamming distances (defined as the number of mutations between every pair of env sequences within an individual and averaged over the total number of pairs of sequence for an individual) per nucleotide are shown for 13 mother-infant transmission pairs described in Nakamura et al. (38). All infants tested negative at birth and at 1 month of age and positive at 6 weeks of age. Infant viral diversity is shown in red triangles on the left and maternal viral diversity on the right in red filled circles. P value was obtained via paired Wilcoxon test.

Fig. 2: Highlighter and Phylogenetic Tree of env Sequences from Two Breastmilk Transmission Mother-infant Pairs in the ZEBS cohort.

Phylogenetic trees (right) were obtained using IQ Tree software with ultrafast bootstrap branch support (41) and depict infant sequences as blue filled circles and maternal sequences as filled red squares. Highlighter plots (left) were obtained using the Highlighter tool from the LANL database (www.hiv.lanl.gov) and depict each sequence as a line in the same order from top to bottom as shown in the phylogenetic tree on the right. Mutations from the consensus sequence (shown at the top) are shown as color-coded vertical tic marks: green for A, aquamarine for C, orange for G, red for T, and gray for gaps. Green and magenta filled squares at the bottom show positions of the receptor and co-receptor binding domains respectively. Both pairs exhibit far less mutations and shorter branch lengths in the infant compared to the mother. However, infant 3 (A) shows a more homogeneous viral population, with most sequences identical to the consensus, whereas in infant 5 (B) no two sequences are identical and all pairs share high similarity and 3–10 mutations from one another.

Fitting a Poisson model to the sequence diversity distribution allows us to infer an estimated time since the first viral replication in the new host, as previously described [42, 43]. Of the 11 infant sequence sets from these breastmilk transmissions that did follow a Poisson distribution, only one (infant 12B, with the lowest mean diversity; Table 3) yielded an estimated time of infection within the time window between the sequence sampling time and the last negative time point. The other infant env sequences yielded estimated times of infection that were 2–10 weeks longer, suggesting that the true common ancestor of these lineages is to be found in the maternal viruses, not the infant’s. Taken together, these findings suggest that the majority of these breastmilk infections were initiated by multiple transmitted lineages with genetically similar common ancestors, consistent with the infant being exposed and infected multiple times.

Nakamura et al. also compared the sequences from the 13 postnatal transmissions to six additional in utero transmission pairs [38]. They found that in utero transmitted clade C gp160 sequences were shorter and less glycosylated than their corresponding maternal variants (p=0.008 and 0.001 respectively), yet this was no longer true when comparing gp160 sequences transmitted via breast milk. A different study, which only looked at the V1-V5 region of env, found that intra-partum transmitted clade C viruses had shorter env V1-V5 regions with fewer putative N-linked glycosylation sites compared to matched maternal sequences, but the same finding was not true for in utero transmitted viruses [39]. While these are not necessarily contradictory findings, given that Nakamura et al. compared in utero transmission with breast milk transmission, whereas Russell et al. compared in utero and peripartum transmitted viruses, the different results emphasize the need to obtain more env viral sequence data from vertical transmission pairs.

The issue of genetic diversity is particularly relevant in light of the latest deep sequencing technology, the Pacific Biosciences single molecule real-time platform, or PacBio [113]. Until now vertically transmitted virus sequences have only been studied using bulk PCR and cloning, or single genome amplification (SGA), which, on average, yield 10–50 sequences per individual [42]. Used with unique molecular identifiers (SMRT-UMI) to sequence study participants enrolled in the AMP trials [20], the PacBio technology generated an average of 270 sequences per sample. This sequencing depth led to the discovery of a higher rate of rare “minority” variants and an increased power to detect multiple transmitted lineages [Mullins J., et al., manuscript in preparation]. Studies that will employ PacBio sequencing technology to generate vertical transmission data are underway [30] and, if consistent with the findings from the AMP trials, they will significantly increase our ability to detect minority variants that would otherwise have remained undetected. It remains to be seen whether there are significantly more minority variants among vertically transmitted viruses and thus a wider range of viral phenotypes and features than previously described.

Role of Antibodies in Vertical Transmission

In the absence of ART, less than half of infants born to mothers living with HIV acquire infection [8–10], suggesting that maternal IgG responses may play a protective role. In fact, vertical transmissions are the only HIV transmissions that occur in the presence of antibodies that co-evolved with the transmitted virus in the recipient. HIV Env-specific IgG antibodies are present not only in the maternal serum, but also in breastmilk, though at lower concentrations than plasma [114]. Characterizing maternal antibody responses and their role in vertical transmission is key in designing immunization strategies to be used during pregnancy or lactation, and the impact of population level immunization on vertical HIV transmission. Passive immunization strategies to prevent vertical transmission must not only block viruses from initiating infection, but also quell any viral escape route to avoid selecting for more resistant, hard-to-neutralize viruses that can potentially evade immunity.

Several studies have reported that viruses transmitted either perinatally or via breastfeeding are generally resistant to maternal autologous serum neutralization, suggesting that neutralization escape variants are preferentially transmitted [50, 52, 53, 75]. Shorter variable loops and fewer glycosylation sites have been generally observed in transmitted viruses when comparing to later stage viruses [115] [116] [117, 118], and in studies of subtype C adult transmission pairs, these characteristics appear to also make viruses more sensitive to antibodies from the transmitting partner [49]. This paradigm is reversed in some but not all studies of vertical transmissions, where several studies have found that viruses transmitted either perinatally or via breastfeeding are generally resistant to maternal autologous plasma [50–53] compared to non-transmitted maternal variants (although other studies have also found no difference [39]).

Importantly, an inverse association between vertical transmission risk and the overall breadth or magnitude of maternal IgG responses has not always been confirmed [119, 120]. Some studies have reported an inverse association between maternal antibody responses and risk of transmission and, additionally, showed evidence that the specificity of maternal responses may play a role in protection. Using plasma samples from mothers whose infants acquired HIV and matched mothers whose infants did not acquire HIV from the WITS study [86], Permar et al. found that the magnitude of maternal IgG responses specific for the V3 variable loop was predictive of a reduced risk of vertical transmission [121]. Subsequent studies from the same group found that the protective V3-specific IgG responses were directed at the C-terminal flank of the V3 crown in particular [122, 123], which could be a response generated via vaccination. However, the same association was not observed in Malawian mothers living with HIV who received ARV at delivery [124]. A subset of mothers from the same study (BAN) whose infants acquired HIV was found to have significantly higher heterologous neutralization activity compared to mothers whose infants did not acquire HIV [125]. Interestingly, when looking specifically at pregnant women living with HIV with broad and potent neutralization activity, multi-specific bnAb responses or bnAb responses that target uncommon epitopes are more common among mothers whose infants did not acquire HIV [97].

The role of maternal HIV specific IgA in vertical transmission is also unclear. Kuhn et al. [126] analyzed IgA concentrations in breast milk from 26 transmitting and 64 non-transmitting mothers living with HIV and detected HIV-specific IgA more often in transmitting mothers compared to non-transmitting ones. However, Pollara et al. [127] found no significant difference in the magnitude of total plasma or breastmilk IgA binding against a multiclade panel of HIV-1 Env antigens. Analysis of 19 transmitting and 57 non-transmitting mothers from the International Maternal-Pediatric-Adolescent AIDS Clinical Trials Network Promoting Maternal-Infant Survival Everywhere (PROMISE) trial, Hompe et al. [128] found no association between breast milk transmission risk and antibody responses, including breast milk Env-specific secretory IgA and plasma Env-specific IgA. Overall, the impact of maternal HIV-specific IgA responses on vertical transmission deserves further investigation.

Several studies have looked at the neutralization sensitivity of vertically transmitted viruses against first and second generation HIV bnAbs. Mabuka et al. tested 107 envs from mothers participating in the Nairobi Breastfeeding Trial and 10 infants (all of whom tested negative at birth, with nine testing positive at 6 weeks and one at 6 months of age) against a panel of 7 bnAbs: NIH45–46W, VRC01, b12, PGT128, PGT121, PG9 and PGT145 [129]. While they found no differences in bnAb sensitivity when comparing to envs from mothers or matched non-transmitting mothers, they did find that the vertically transmitted envs were significantly more sensitive to neutralization by V3 bnAbs PGT128 and PGT121 compared to heterosexually transmitted viruses sampled from the same region (Table 4). No differences were found for the other bNAbs. They also found significant differences in neutralization sensitivity among vertically transmitted viruses when comparing clade A envs to clade C envs, consistent with previous observations that bnAb sensitivity varies across different HIV subtypes [130].

Table 4.

Summary of studies that compared sensitivity of vertically transmitted viruses vs. heterosexually transmitted viruses or non-transmitted maternal viruses to individual HIV bnAbs. Studies are: Wu et al. (53), Mabuka et al. (129), Rainwater et al. (52), Nakamura et al. (38), Russell et al. (39), Kumar et al. (51). Abbreviations: BM = breastmilk transmitted viruses, IP = intrapartum transmitted viruses. IU = in utero transmitted viruses.

bnAb	Epitope	Vertically transmitted viruses vs. heterosexually transmitted viruses	Vertically transmitted viruses vs. non-transmitted maternal viruses
NIH45–46W	CD4bs	no difference (Mabuka, BM)	more sensitive (Kumar, IP) no difference (Mabuka, BM)
VRC01	CD4bs	no difference (Mabuka, BM)	sensitive (Russell, BM; Kumar, IP) no difference (Mabuka, BM)
b12	CD4bs		more resistant (Wu*; Kumar, IP) no difference (Rainwater, BM)
PG9	V2	no difference (Mabuka, BM)	more sensitive (Nakamura, BM) no difference (Nakamura, IU; Russell, BM)
PGT145	V2	no difference (Mabuka, BM)	no difference (Mabuka, BM)
PG16	V2		more sensitive (Nakamura, BM) no difference (Nakamura, IU; Russell, BM)
PGT128	V3	more sensitive (Mabuka, BM)	no difference (Mabuka, BM)
PGT121	V3	more sensitive (Mabuka, BM)	sensitive (Kumar, IP) no difference (Mabuka, BM)
2G12	Glycan		more resistant (Wu*, Russell, BM) no difference (Rainwater, BM)
2F5	MPER		more resistant (Wu*, Russell, BM; Kumar, IP)
4E10	MPER		more resistant (Wu*; Kumar, IP)
10E8	MPER		sensitive (Kumar, IP)

When comparing bnAb sensitivity of transmitted vs. non-transmitted viral variants isolated from the same mother, findings vary across studies (Table 4). Rainwater et al. looked at four breastmilk transmission pairs, all infected with subtype A, and found no difference in sensitivity against autologous plasma or monoclonal bnAbs 2G12 and b12 between transmitted viruses and maternal non-transmitted viruses [52]. This negative result was likely due to the small sample size as later studies with a larger number of transmission pairs found significant differences: Wu et al. studied 12 mother-infant pairs infected with subtype A or C viruses, and a few with either A/D or C/D recombinant viruses, and found breastmilk transmitted envs to be among the most resistant variants against b12, 2G12, 2F5 and 4E10 compared to non-transmitted maternal variants [53]. Kumar et al. replicated these results for b12 and 2F5 (Table 4), examining intra-partum transmission pairs infected with subtype B, and, additionally found that intra-partum transmitted envs were more sensitive to V3 bnAb PGT121 and CD4bs bnAbs VRC01 and NIH45–46W [51].

Nakamura et al. [38] found no significant difference in sensitivity to V2 bnAbs PG9 and PG16 between transmitted and non-transmitted variants for in utero transmission pairs, although postnatally transmitted envs (via breastmilk) were more sensitive to both bnAbs when compared to non-transmitted maternal variants. This latter result was consistent with Russell et al., who also found breastmilk transmitted variants to be generally sensitive to VRC01 and resistant to 2F5 and 2G12 [94], in line with previous reports [51, 53]. All viruses tested in both studies were subtype C, with the exception of one infant in Nakamura et al., who was infected with subtype G.

The sensitivity of vertically transmitted variants to non-neutralization functions such as antibody dependent cell mediated cytotoxicity (ADCC) may be another important determinant of the success of passive immunization strategies to prevent vertical transmission. Mabuka et al. [131] examined ADCC in the breastmilk of a small subset of Kenyan women living with HIV whose babies acquired (n=9) or did not acquire (n=10) the virus. ADCC mediating antibodies were frequently detected in breastmilk, and high breastmilk ADCC capacity was associated with lower risk of infant virus acquisition [131]. However, in a larger cohort of 72 mother-infant pairs, no association was observed between ADCC activity in maternal plasma and infant acquisition risk, although ADCC activity in plasma of infants who became infected was associated with lower risk of infant morbidity [132]. In the BAN study, no association was observed between breastmilk ADCC activity against a heterologous virus strain and infant transmission [127], but Thomas et al. reported that infants with high ADCC activity against their maternal virus strains had a lower risk of virus acquisition [133]. Subsequently, the latter group reported that infants with a combination of high ADCC activity and exposure to more ADCC sensitive strains were less likely to acquire HIV [134]. Thus, it will be important to assess the sensitivity of vertically transmitted viruses to bnAb-mediated ADCC when developing bnAb combinations to prevent HIV vertical transmission.

Taken together, these results raise the hypothesis that vertically transmitted viruses tend to be more sensitive to V2 and V3 bnAbs and potent second generation CD4bs bnAbs such as VRC01 and NIH45–46W, while more resistant to membrane proximal external region (MPER) bnAbs such as 2F5 and 4E10, compared to maternal non-transmitted variants. However, as pointed out earlier, all viruses from these studies lack adequate representation and were obtained from cohorts sampled at least a decade ago; more recent studies will be needed to confirm the findings.

Conclusions

With infants still disproportionally affected by new HIV infections despite the advent of ARV therapy, new interventions are needed to eradicate vertical transmission of HIV. Following the AMP trials [19], which established that passive infusion of the CD4 binding site bnAb VRC01 can prevent sexual acquisition of VRC01-sensitive viruses [20], dual and triple combination bnAb regimens are under early phase study in adults [21–26] and infants (i.e., IMPAACT 2037) to establish their safety and pharmacokinetics. Combination bnAbs targeting distinct Env epitopes have been advanced for testing based on their potent and broad coverage of multi-clade viral panels. However, these panels are comprised of viruses exclusively sampled from adult transmissions.

HIV susceptibility to bnAbs varies by clade [130], geography, and over time. In fact, over the past decades, HIV viruses have greatly diversified, with CRF forms alone globally increasing in prevalence [34], becoming more resistant to autologous sera [35] and bnAbs [36, 37]. Because vertically transmitted viruses are derived from adults, it is reasonable to hypothesize that they have undergone the same genetic drift. However, many factors differentiate the context of vertical vs. sexual transmission: (i) each vertical transmission route, whether the placenta, the birth canal, or the mammary gland, is biologically different than that of the genital mucosa; (ii) transmission via all three routes occurs in the presence of autologous antibodies that the infant passively acquires from the mother; and (iii) fetal and infant immune systems are distinct from the mature adult immune system. Therefore, vertically transmitted viruses undergo distinct transmission bottlenecks and early evolutionary pressure compared to viruses acquired sexually. Little is known regarding how these unique features affect virus phenotypes and in particular their neutralization sensitivity to candidate bnAbs and bnAb combinations. The vertically transmitted viruses that have been characterized were sampled and isolated more than a decade ago. Some of the viruses were sampled prior to the advent of maternal ART, they are not geographically representative, and they only represent a subset of the clades responsible for vertical transmission. Many phenotypic features, such as variable loop lengths and number of glycosylation sites, have only been evaluated in clade C transmission pairs and may not necessarily apply to other subtypes.

The AMP trials have highlighted the need to update viral panels with more recent env sequences to better represent the diversity of HIV circulating globally [36]. Additionally, sequences obtained using PacBio SMRT-UMI technology [113] far surpassed the sampling depth of previous sequencing strategies such as SGA and revealed a more diversified viral population in adults than previously observed in early infection. This sequencing depth, applied to infant samples, is likely to reveal minority variants that would have otherwise been undetected. Identifying and testing for in vitro neutralization of such variants is crucial in preparation for bnAb combination studies, as missing low frequency viruses that carry pre-existing bnAb resistance could lead to selection of more resistant strains. Deep sequencing will also better inform the differences between maternal transmitted and non-transmitted variants and across the three routes of vertical transmission — in utero, intra-partum, and via breastfeeding. Here, based on env diversity of sequences sampled between 1996 and 2008, we uncovered striking differences in breastmilk transmitted viruses compared to intra-partum transmitted viruses: breastmilk transmissions were characterized by multiple low-diversity founder lineages. However, it is possible that low diversity minor variants are present in all infants at frequencies so low that sampling sizes of 10–20 sequences per infant is insufficient. It is also possible that PacBio SMRT-UMI [113] technology will reveal even higher diversity in breastmilk infection and deepen the complexity of such infections.

In conclusion, we highlight the urgent need to collect more recent panels of vertically transmitted HIV viruses, across different subtypes, sampling these as early as possible in the course of infant infection and using accurate deep sequencing technology to characterize their genetic diversity, phenotypic traits, and in particular their neutralization sensitivities to HIV bnAbs that are candidates for future clinical testing.

List of Abbreviations

ADCC

Antibody Dependent Cell-mediated Cytotoxicity

AMP

Antibody Mediated Trials

ART

Antiretroviral Therapy

ARV

Antiretroviral

bnAb

Broadly Neutralizing Antibody

CRF

Circulating Recombinant Form

DBS

Dried Blood Spot

HIV

Human Immunodeficiency Virus

PLWH

People Living with HIV

SHIV

Simian- Human Immunodeficiency Virus

VL

Viral Load