Since 1985, the ANRS French Perinatal Cohort (CO 01-ANRS-EPF, Agence Nationale de Recherche sur le SIDA-Enquête Périnatale Française) has prospectively collected data on HIV-infected pregnant women and their children in 90 centers throughout France. Informed consent is obtained from the mothers during pregnancy or at the time of delivery. The children receive standard care, including clinical and biological examinations at birth and 1, 3, 6, 12 and 18-24 months, as previously reported 15. The cohort study was approved by the Cochin Hospital Institutional Review Board and by the French computer database watchdog commission (CNIL). Mother and infant plasma and cells were collected between 1990 and 2005 and stored in Necker hospital virology laboratory.

HIV-1 infection was diagnosed in the newborn when at least two separate samples were positive by HIV-1 RNA/DNA detection or by a viral culture. A positive test at birth or before 7 days of age indicates intrauterine transmission, while a negative test at birth and a positive test more than 7 days later indicate intrapartum transmission. An infant is considered uninfected when two tests performed one month after discontinuation of antiretroviral prophylaxis are negative.

Newborns were included in this analysis if: (1) they were born and enrolled in metropolitan France in centers participating in the EPF cohort between 1997 and 2004; (2) they were HIV-1-infected; and (3) if frozen samples were available for resistance testing.

For each mother-child pair, we analyzed the first available HIV-1-positive sample(s) from the infant's delivery sample and the mother's. If drug resistance was detected in the newborn diagnostic sample, available follow-up samples from the infant were tested for genotypic resistance.

Other data, including the mothers' viral load values and the mothers' and infants' antiretroviral treatment histories, were obtained from the ANRS-EPF database.

Plasma HIV-1 RNA was quantified with the Cobas Amplicor HIV-1 Monitor 1.5 assay kit (Roche Diagnostics, Meylan, France; detection limit 400 or 40 copies/mL).

The ANRS consensus method was used for population-based nucleotide sequence analysis of the whole protease gene (codons 1 to 99) and codons 1 to 305 of the reverse transcriptase gene on HIV-1 RNA in plasma and HIV-1 DNA in PBMC 16. Drug resistance mutations were identified by following the International AIDS Society-USA 2007 Drug Resistance Group guidelines 17http://www.iasusa.org. Specifically, we considered the following mutations (relative to the reference wild-type (WT) strain HXB2): protease inhibitors (PI): D30N, L33F/I, M46I/L, G48V, I50L/V, V82A/F/L/S/T, I84A/C/V, and L90M; nucleoside reverse transcriptase inhibitors (NRTI): M41L, A62V, K65R, D67N, K70R, L74V, V75I, F77L, Y115F, F116Y, Q151M, M184V, L210W, T215Y/F/C/D/E/S/I/V/A/G/H/L/N and K219E/Q/R; and non nucleoside reverse transcriptase inhibitors (NNRTI): L100I, K103N, V106A/M, V108I, Y181C/I, Y188C/H/L, G190A/S, P225H, M230L, and P236L. Mixtures of WT and mutant sequences were considered drug-resistant. Interpretation of genotypic drug susceptibility was done according to the 2007 French ANRS algorithm http://www.hivfrenchresistance.org.

In order to characterize the plasma and cellular viral quasispecies, clonal analyses were performed on samples from three mother-child pairs. The maternal samples were obtained at delivery and the children's samples were obtained both at birth and subsequently. These three pairs were chosen as being representative of three different situations, and because suitable plasma/cell samples for them were available. The RT or protease gene was amplified. Purified PCR products were cloned into the pCR Topo 2-1 plasmid (TOPO TA Cloning kits, Invitrogen BV, the Netherlands) as recommended by the manufacturer. DNA was purified with the Mini-Prep kit (Qiagen) and clones were analyzed by dye terminator sequencing on an ABI Prism 3100 genetic analyzer.

Mother-child clustering of pol sequences was confirmed by phylogenetic analysis. All sequences of HIV-1 RNA and DNA clones from each mother-child pair were aligned with Clustal W 1.7 software. Pairwise evolutionary distances were estimated with DNADist using Kimura's two-parameter method. The phylogenetic trees were then constructed with a neighbor joining method (Neighbor program implemented in the Phylip package) 18. The reliability of each tree topology was estimated from 100 bootstrap replicates 18.

From January 1997 to December 2004, 6170 mother-child pairs were enrolled in the ANRS-EPF cohort, representing approximately 70% of births to HIV-1-infected mothers in France. 92 newborns were infected during this period despite prophylaxis. It is important to note that the newborn samples were used to diagnose HIV infection and that the remaining stored samples were usually very limited.

HIV-1-positive plasma and/or PBMC samples from 60 children (33 boys and 27 girls) were available for drug resistance studies. Samples were also available from 32 of these children's mothers. The children's samples were obtained at a median age of 29 days (1 to 313 days), and 72% of plasma samples were collected less than 60 days after birth. The children's median HIV-1 RNA viral load at diagnosis was 4.5 log₁₀ copies/ml (2.1 to 7.3 log₁₀).

Twelve (20%) of the 60 newborns had resistant variants at diagnosis of HIV-1 infection, according to the 2007 IAS (International AIDS Society) list (Table 1). Six of these children were infected in utero and four intrapartum; the timing of infection could not be determined in the remaining two children as no birth sample was available. The mutations were associated with resistance to NRTI in 10 cases [thymidine analog mutations (TAMs) in six cases, T69N in one case, M184V in one case, and both mutations in two cases], NNRTI in two cases, and PI in one case.

According to the 2007 ANRS algorithm, 6 of the 12 children had variants with resistance to at least one antiretroviral drug [overall frequency 10% (6/60)]. Resistance to NRTI, NNRTI and PI was observed in four children, two children and one child, respectively. One child had variants resistant to both NRTI and NNRTI (child #10, Table 1).

In all but one case, the neonates' drug resistance profiles were related to the antiretroviral drugs received by the mother and/or by the child (Table 1). Infant #10 harbored viruses with mutations associated with NNRTI resistance, without being exposed perinatally to this drug class. His mother had never received NNRTI, but she had probably been infected with NNRTI-resistant virus transmitted by her husband, who was treated with a regimen containing nevirapine, stavudine and lamivudine.

The viral subtypes were determined in 53 children, and were subtype CRF02_AG in 23 cases (43%), B in 19 cases (36%), A in 5 cases (9%) and another subtype in 6 cases (11%). Among the 10 subtyped resistant viruses, 5 (50%) belonged to subtype B, three (30%) to CRF02_AG, one to A and one to F.

DNA-based resistance results were available for 8 of the 12 children with resistant viruses in plasma. In 6 cases HIV-1 RNA and DNA harbored the same resistance mutations (Table 1), while no mutation was detected in HIV-1 DNA in the other two cases.

Samples from 32 mother-child pairs were available, including 10 of the 12 children with resistant virus in the plasma (Table 1). The resistance pattern was the same in six mother-child pairs. In the remaining four cases the mothers harbored different mutations or no mutation. Interestingly, child #9, whose mother harbored PI resistance mutations L10I, L63P and L90M and RT resistance mutations Y181C, L210W and T215D, only harbored the PI resistance mutations. The mother was receiving didanosine, saquinavir and lopinavir/ritonavir, probably leading to the selection of a dominant PI-resistant quasispecies. Among the 22 remaining mother-child pairs, 20 mothers had wild-type viruses (in plasma), while the other two mothers harbored resistant viruses that were not transmitted to the child.

Longitudinal resistance studies were performed in 8 of the 12 cases in which serial samples were available (median 4 samples per child), over a median period of 52 months (Table 2). The same resistance mutations persisted in the plasma and PBMC for 6 months to 5 years, regardless of the antiretrovirals used in six children. Additional mutations had accumulated in the RNA and the DNA during failing regimens. In two children (#6, #12), no zidovudine resistance mutations were detected when zidovudine prophylaxis was discontinued. Interestingly, no resistance mutations were detected in mother samples and in birth children cells (Table 1 and 2).

To better understand how drug-resistant HIV-1 strains detected in newborns are acquired, we conducted clonal analyses of plasma and PBMC viral populations in three mother-child pairs. The maternal samples were taken at delivery, and the children's samples were taken both at birth and at a later time.

In mother-child pair #9, 110 protease gene clones were sequenced (Figure 1). In the mother, all 21 plasma clones harbored the L90M major mutation and other minor mutations. Her PBMC harbored heterogeneous variants (12/21 wild-type, 8/21 L90M and 1/21 I84V), according to the temporal archiving of resistant variants in lymphocytes during therapeutic regimens that contrasted with the homogeneity reported in the plasma under selective therapeutic pressure. In her child, who was infected in utero, all plasma and cellular variants harbored the L90M mutation (40/40 at birth and 28/28 at month 30), even during the period without PI selective pressure. Phylogenetic analysis confirmed the homogeneity of the child's specimens at birth with a genetic intravariability of protease gene that increased over time (from 0.003% to 0.01%). This case suggests the perinatal transmission of L90M variants with early archiving in the child's lymphocytes and persistence over time.

In mother-child pair #11, 70 RT gene clones were sequenced (Figure 2). The mother acquired HIV-1 infection during pregnancy and was rapidly treated with zidovudine, lamivudine and indinavir/ritonavir. The child was infected in utero, despite elective Cesarean section and the intensification of postnatal zidovudine prophylaxis by the addition of lamivudine. All plasma and cellular quasispecies detected in the newborn (35/35 at month 3 and 26/26 at month 7) harbored the M184V lamivudine resistance mutation. However, this mutation was not detected in the mother's delivery plasma sample (9/9 wild-type). Phylogenetic analysis confirmed low genetic intravariability (mean 0.006%) of the RT gene in the mother and her child, in keeping with the high homogeneity due to the primary infection in the child and his mother. M184V variants may have arisen during lamivudine treatment of the mother and prophylaxis of the infant, leading to the massive early lymphocyte infection and persistence of lamivudine resistance. However, we cannot exclude an abacavir-selective pressure on the M184V resistance-associated mutation or a minor maternal M184V variant transmission.

In mother-child pair #12, 61 RT gene clones were sequenced (Figure 3). The mother had advanced HIV-1 disease and poor adherence to treatment as reflected by high viral load (4.4 log₁₀ copies/mL). Resistance was undetectable even by clonal analysis (28/28 wild-type). Zidovudine prophylaxis was initiated at birth and continued for 6 weeks despite the diagnosis of HIV-1 in utero infection in the newborn. In the child, the K70R mutation was detected in 42% of clones (10/24) at month 1 and in 0% at month 12. Genetic intravariability was low (0.005%) in the child, as expected, during primary infection. In this case, wild-type viruses were detected concomitantly in the RNA from the mother and in the DNA from the child (only 1/10 resistant clones), suggesting that most archived viruses in the child were WT viruses transmitted by the mother. Zidovudine resistance, present at the time of diagnosis, occurred during suboptimal zidovudine pressure. Zidovudine discontinuation led to the re-emergence of wild-type variants in the plasma at month 12, confirming that the reservoir consisted mainly of wild-type viruses.