HIV-1–Specific Cytotoxic T Lymphocytes and Breast Milk HIV-1 Transmission

Abstract

Background

Breast-feeding by infants exposed to human immunodeficiency virus type 1 (HIV-1) provides an opportunity to assess the role played by repeated HIV-1 exposure in eliciting HIV-1–specific immunity and in defining whether immune responses correlate with protection from infection.

Methods

Breast-feeding infants born to HIV-1–seropositive women were assessed for HLA-selected HIV-1 peptide–specific cytotoxic T lymphocyte interferon (IFN)–γ responses by means of enzyme-linked immunospot (ELISpot) assays at 1, 3, 6, 9, and 12 months of age. Responses were deemed to be positive when they reached ⩾50 HIV-1–specific sfu/1 × 10⁶ peripheral blood mononuclear cells (PBMCs) and were at least twice those of negative controls.

Results

A total of 807 ELISpot assays were performed for 217 infants who remained uninfected with HIV-1 at ∼12 months of age; 101 infants (47%) had at least 1 positive ELISpot result (median, 78–170 sfu/1 × 10⁶ PBMCs). The prevalence and magnitude of responses increased with age (P = .01 and P = .007, respectively); the median log₁₀ value for HIV-1–specific IFN-γ responses increased by 1.0 sfu/1 × 10⁶ PBMCs/month (P < .001) between 1 and 12 months of age. Of 141 HIV-1–uninfected infants with 1-month ELISpot results, 10 (7%) acquired HIV-1 infection (0/16 with positive vs. 10/125 [8%] with negative ELISpot results; P = .6). Higher values for log₁₀ HIV-1–specific spot-forming units at 1 month of age were associated with a decreased risk of HIV-1 infection, adjusted for maternal HIV-1 RNA level (adjusted hazard ratio, 0.09 [95% confidence interval, 0.01–0.72]).

Conclusions

Breast-feeding HIV-1–exposed uninfected infants frequently had HIV-1–specific IFN-γ responses. Greater early HIV-1–specific IFN-γ responses were associated with decreased HIV-1 acquisition.

An estimated 80% of breast-feeding infants born to HIV-1–seropositive women escape HIV-1 infection despite ingesting hundreds of liters of HIV-1–infected breast milk [1]. Thus, continual exposure to HIV-1 does not invariably lead to transmission. There are at least 2 models that may explain this outcome. The first is that infants escape infection because they are insufficiently exposed to HIV-1; the other is that they receive an immunizing, but not infective, dose of HIV-1 that protects them from subsequent infection. HIV-1–specific cytotoxic T lymphocyte (CTL) interferon (IFN)–γ secretion has been reported in several small studies of HIV-1–exposed uninfected infants [2–5]. Legrand et al. [3] demonstrated HIV-1 nef– and HIV-1 gag–specific dual IFN-γ and tumor necrosis factor–α secretion by CD8⁺ T cells in 16 exposed uninfected infants; this study provided important evidence confirming the presence of polyfunctional HIV-1–specific CD8⁺ T cell responses in exposed uninfected infants.

A key foundation for several HIV-1 vaccines is the concept that HIV-1–specific CTL responses provide protection against infection, but evidence to support this hypothesis has been limited and conflicting. HIV-1–specific CTL responses have been detected in exposed uninfected individuals [6]. However, such studies have not discerned whether HIV-1–specific immune responses are markers of exposure or correlates of protection. In animal models, exposed uninfected sex workers, and HIV-1–infected individuals, HIV-1 infection or superinfection occurs despite the presence of CTLs, suggesting that simply having a cellular immune response is insufficient for protection [7–9]. The recent disappointing results of the STEP trial, which noted no protection with a vaccine (Merck V520) designed to induce HIV-1–specific CTLs, confirms the need to define protective immune correlates [10].

A prospective study of exposed uninfected individuals is necessary to determine both the incidence of CTL responses and whether CTLs protect from infection. An ideal setting for this approach is breast-feeding infants of HIV-1–infected mothers; for these infants, repeated exposure is common, measurable, and consistent, which is not the case for sexually exposed adults. Infants repetitively exposed to HIV-1 may provide a natural instance of vaccination with live virus, which can inform vaccine design based on mimicking a natural model of effective immune response. We hypothesized that exposed uninfected infants experience sufficient viral exposure to induce systemically detectable HIV-1–specific immune responses and that these early immune responses could protect against subsequent infection. Thus, we designed a prospective cohort study of infants born to HIV-1–seropositive mothers to determine the prevalence, magnitude, and effect of HIV-1–specific cellular immune responses.

METHODS

The study was reviewed and approved by the Institutional Review Board of the University of Washington and the Ethical Review Committee of Kenyatta National Hospital.

Clinical procedures

HIV-1–seropositive pregnant women were enrolled after provision of written informed consent. Eligibility criteria included age ⩾18 years, gestation ⩽32 weeks, and willingness to adhere to scheduled infant blood samplings for at least 1 year. Mothers were counseled regarding breast milk HIV-1 transmission and were supported in their feeding choice. Mothers received short-course zidovudine (300 mg twice daily from 34 weeks of gestation and every 3 h during labor) to prevent HIV-1 infection in their infants [11]. Within 48 h of birth, blood was collected from infants for molecular HLA typing and HIV-1 detection. Blood was collected again from the infants at 1, 3, 6, 9, and 12 months of age for HIV-1 DNA, RNA, and IFN-γ enzyme-linked immunospot (ELISpot) assays.

HLA typing

Molecular class I HLA typing was performed using DNA from infant peripheral blood mononuclear cells (PBMCs). HLA alleles were determined using amplification-refractory mutation system polymerase chain reaction (PCR), with primers designed for East African alleles [12].

ELISpot assays

IFN-γ ELISpot assays were performed on freshly isolated PBMCs, as described elsewhere [13–15]. Briefly, PBMCs from infants were added to 96-well Millipore plates (Millipore) coated with 1-DIK monoclonal antibodies at 15 µg/mL (Mabtech) in duplicate at 2 × 10⁵ PBMCs/well and stimulated with either phytohemagglutinin (Murex Biotech) at 20 µg/mL (positive control), medium alone (negative control), or HLA-selected peptide at 20 µg/mL (1 peptide per well, except for clade variants, for which equivalent amounts were used for each variant). Spot-forming units were defined as the average number of spots in duplicate wells, and the number of HIV-1–specific spot-forming units was calculated as the total number minus the average number in negative control wells.

Peptides for ELISpot assays were selected using a predefined algorithm based on infant HLA (table 1). The median number of peptides tested per infant per visit was 11 (range, 1–29). Spots were counted by eye in plates until January 2001, after which an automated ELISpot assay reader was used (Autoimmun Diagnostika). In assays with both counting methods conducted concurrently, κ concordance was 1.0 (P < .001), and correlation was 0.94 (P < .001). Eye-counted results were used before machine counting was instituted, and machine results were used thereafter. Spot counts were entered into a database without links to HIV-1 status, and HLA-matched assays were computed as positive or negative on the basis of a predetermined computer algorithm using published criteria (⩾50 HIV-1–specific sfu/1 × 10⁶ PBMCs, with experimental values at least twice those of negative control wells) [16, 17]. Assays were conducted blinded to infant HIV-1 status.

Table 1.

Peptide epitopes used for stimulation in enzyme-linked immunospot assays, by HLA type.

HLA	Protein	Sequence (5′→3′)
A1	p17	GSEELRSLY
A1	Rev	ISERILSTY
A2	RT	ILKEPVHGV/ILKDPVHGV
A2	Nef	ALKHRAYEL
A2	p17	SLFNTVATL/SLYNTVATL
A2	p24	TLNAWVKVI/TLNAWVKVV
A2	RT	VIYQYMDDL
A2	Nef	VLEWRFDSRL
A2	Nef	PLTFGWCYKL
A3	p17	KIRLRPGGK
A3	Nef	QVPLRPMTYK
A3	gp160	TVYYGVPVWK
A3/11/33	RT	AIFQSSMTK/SIFQSSMTK
A3/31	gp160	RLRDLLLIVTR
A3/11/31/33	RT	DLEIGQHRTK
A3	Nef	DLSHFLKEK
A24	Gp160	YLRDQQLL/YLKDQQLL
A24/B44	p24	RDYVDRFYKTL/RDYVDRFFKTL
A24	Nef	DSRLAFHHM
A25	p24	DTINEEAAEW/ETINEEAAEW
A26/B70/72	p24	YVDRFFKTL
A29	gp160	FNCGGEFFY
A30	RT	KLNWASQIY
A30	p17	RSLYNTVATLY
A30	gp160	IVNRVRQGY
A30	RT	KQNPDIVIYQY
A30	gp41	KYCWNLLQY
A6802	Protease	DVTLEDINL
A6802	Int	ETAYFILKL
A6802/74	RT	ETFYVDGAAN
A6802/74	Protease	ITLWQRPLV
A6802	Protease	DTVLEEMNL
B7/8101/42	Nef	TPGPGVRYPL/TPGPGIRYPL
B7	gp160	IPRRIRQGL
B7	p24	SPRTLNAWV
B7	p24	GPGHKARVL
B7/8101	p24	ATPQDLNTM
B7	Nef	FPVTPQVPLR
B8	Nef	FLKEKGGL
B8	p24	DIYKRWII/EIYKRWII
B8	RT	GPKVKQWPL
B8	gp160	YLKDQQLL/YLRDQQLL
B8	p17	GGKKKYRL/GGKKKYKL
B14/Cw8	p24	DRFFKTLRA/DRFYKTLRA
B14/Cw8	p24	DLNTMLNTV/DLNMMLNIV
B14/Cw8	p24	RAEQASQEV/RAEQATQEV
B14	gp160	ERYLKDQQL
B15/Cw4	gp160	SFNCGGEFF
B18/49	p24	FRDYVDRFYK/FRDYVDRFFK
B18/49	Nef	YPLTFGWCY/YPLTFGWCF
B27	p24	KRWIILGLNK/KRWIIMGLNK
B35	p24	PPIPVGDIY
B35	Nef	VPLRPMTY
B35	RT	HPDIVIYQY/NPDIVIYQY
B35	gp160	TAVPWNASW/TNVPWNSSW
B35	RT	EPIVGAETFY
B37/73	Nef	YFPDWQNYT
B42	p24	GPGHKARVL
B42	Nef	TPQVPLRPM
B45	RT	GAETFYVDGA
B51	Pol	EPIVGAETFY
B53	p24	ASQEVKNWM/ATQEVKNWM
B53	p24	TPQDLNMML/TPQDLNTML
B53	p24	DTINEEAAEW
B53	p24	QATQEVKNW
B53	p24	VKNWMTETLL
B57/5801	p24	TSTLQEQIGW/TSTLQEQIAW
B57/5801	p24	KAFSPEVIPMF
B57/5801	RT	IVLPEKDSW
B57	p24	ISPRTLNAW/LSPRTLNAW
B57	Pol	KITTESIVIW
B70/72	RT	DVKQLTEVV/DVKQLAEAV
Cw4	p24	QASQEVKNW
Cw4	p17	KYRLKHLVW
Cw8	RT	VTDSQYALGI
Cw8	gp160	NCSFNISTSI
Cw8	Nef	KAAVDLSMFL

NOTE. RT, reverse transcriptase.

HIV-1 DNA and RNA assays and determination of infant HIV-1 status

HIV-1 DNA PCR filter paper assays were conducted using methods with 98% specificity and 99% sensitivity [18]. HIV-1 RNA levels were quantified using the Gen-Probe transcription-mediated assay. Infants were deemed to be HIV-1 infected if 2 consecutive assays were HIV-1 DNA or RNA positive or if a single HIV-1 assay was positive and it was the last available assay.

Control study

The specificity of ELISpot assays was determined in a control study conducted among 20 infants who had been born to HIV-1–seronegative mothers who were determined to be uninfected by ELISA and HIV-1 RNA assay. HIV-1 DNA and ELISpot assays were performed in these infants.

Depletion of CD8⁺ cells

PBMCs were depleted of CD8⁺ T cells by means of anti-CD8 monoclonal antibody–coated magnetic beads (Dynal). In each case, 98% of CD8⁺ T cells were depleted from the population.

Statistical methods

Analyses were restricted to infants whose mothers reported any breast-feeding. Categorical data were compared using χ² and Fisher’s exact tests, and continuous data were compared using the Mann-Whitney U test. For paired comparisons, the Wilcoxon signed-rank test was used for continuous outcomes, and McNemar’s test was used for categorical outcomes. Linear regression analysis was used to determine the change in magnitude of HIV-1–specific responses with age for each infant; the Wilcoxon signed-rank test was used to determine whether the median slope differed from 0. For Kaplan-Meier and Cox regression analyses among infants who were HIV-1 uninfected at 1 month of age, the following time intervals were used: the time to the midpoint between the last HIV-1–negative and the first HIV-1–positive result for infants who became HIV-1 infected between 1 and 12 months of age; the time to the last visit for uninfected infants who were lost to follow-up or died before 12 months of age; and 12 months for infants who remained uninfected at 12 months of age.

RESULTS

Recruitment and follow-up

From July 1999 through October 2002, 36,059 women were offered testing for HIV-1 at 8 clinics, of whom 88% accepted testing. Among HIV-tested women, 4512 (14%) were HIV-1 seropositive, 3190 (71%) of whom received results and were referred to the study clinic. Of 1539 women who came to the study clinic, 510 (33%) were eligible, interested, and enrolled. Delivery information was available for 476 (93%) of the infants, including 474 (99.6%) singleton or first-born infants who were followed up (7 second-born twins were excluded); 465 (98%) had HIV-1 testing at least once. By 1 month of age, 72 infants (15%) had acquired HIV-1 infection, 9 HIV-1–uninfected infants (2%) were lost to follow-up, and 10 uninfected infants (2%) had died, with 374 HIV-1–uninfected infants remaining in follow-up, of whom 284 (76%) were breastfed (figure 1).

Figure 1.

Participant flow from enrollment to follow-up, focusing on breast-feeding infants who were HIV-1 uninfected at 1 month of age and subsequently followed up with HIV-1 and HLA-selected HIV-1 enzyme-linked immunospot (ELISpot) assays.

Prevalence, magnitude, and longitudinal changes in HIV-1–specific CTL responses in breast-feeding HIV-1–uninfected infants who remained uninfected at 1 year of age

Among 217 exposed HIV-1–uninfected infants who remained uninfected at >11.5 months of age, filter paper HIV-1 DNA assays were serially negative for an average of 5.7 time points (range, 3–7). In addition, 195 (90%) of these infants had at least 1 confirmatory negative HIV-1 RNA assay result (mean, 1.8; range, 1–7). Of these breast-feeding uninfected infants, 101 (47%) had at least 1 time point with a positive ELISpot result.

The prevalence of positive ELISpot assays increased over time and ranged from 12% of 112 infants at age 1 month to 22% of 194 infants at age 12 months (table 2). The magnitude of HIV-1–specific responses ranged from 50 to 4535 HIV-1–specific sfu/1 × 10⁶ PBMCs, and the number of peptides that elicited a positive response ranged from 1 to 11 in infants with positive assays. The prevalence of positive ELISpot results increased with age (P = .01), and the number of HIV-1–specific spot-forming units (maximal response to a single epitope) also increased with age (P = .007) in paired comparisons between the ages of 1 and 12 months. The median change in magnitude per month was 1.02 HIV-1–specific sfu (interquartile range, −2.02 to 7.23; P < .001). Positive assays had a higher median number of peptides tested than negative assays (12 vs. 10 peptides; P = .009).

Table 2.

Prevalence, magnitude, breadth, and durability of HIV-1–specific cytotoxic T lymphocyte responses in HIV-1–exposed infants who never became HIV-1 infected during follow-up (up to ∼12 months of age).

Age	Prevalence of HIV-1–specific responses,a proportion (%)	Magnitude of HIV-1–specific responseb		Magnitude of the sum of HIV-1–specific responses among infants with positive ELISpot resultsc	Mean (range) no. of positive epitopes/mean (range) no. tested among infants with positive ELISpot results
		All infants tested	Infants with positive ELISpot results
1 month	13/112 (12)	13 (5–28) [n = 112]	105 (58–273) [n = 13]	125 (71–351)	1.3 (1–2)/10.4 (4–18)
3 months	21/146 (14)	16 (8–35) [n = 146]	78 (55–114) [n = 21]	80 (55–223)	2.1 (1–10)/12.2 (5–19)
6 months	25/169 (15)	20 (8–33) [n = 169]	85 (64–285) [n = 25]	125 (70–371)	2.1 (1–7)/10.6 (2–17)
9 months	40/186 (22)	25 (8–58) [n = 186]	103 (73–265) [n = 40]	185 (86–495)	2.6 (1–10)/12.7 (2–24)
12 months	43/194 (22)d	25 (10–58)e [n = 194]	170 (80–445) [n = 43]	310 (103–1003)	3.0 (1–11)/11.7 (2–28)

NOTE. All infants in this table were followed up for at least 11.5 months, and mothers reported breast-feeding during follow-up. ELISpot, enzyme-linked mmunospot; IQR, interquartile range; PBMCs, peripheral blood mononuclear cells.

^aPositive responses were defined as ⩾50 HIV-1–specific sfu/1 × 10⁶ PBMCs, with experimental values at least twice those of negative control wells (background)

^bData are the median (IQR) of maximum responses to a single epitope, given in HIV-1–specific sfu/1 × 10⁶ PBMCs

^cData are the median (IQR) of summed positive responses, given in HIV-1–specific sfu/1 × 10⁶ PBMCs.

^dSignificantly higher prevalence at month 12 than at month 1 (P = .01)

^eSignificantly higher magnitude at month 12 than at month 1 (P = .007)

Of 62 infants with ELISpot assays conducted at all 5 time points, 26 (42%) had 1 positive assay, and 13 (21%) had >1 (2 positive assays for 6 infants, 3 for 5 infants, and 4 for 2 infants). Of 9 HIV-1–uninfected infants with positive ELISpot results at month 1 of age and assays results at all 4 subsequent time points, 3 infants had positive ELISpot results once, some had repeated responses to consistent epitopes, and some had diversified epitope recognition (figure 2).

Figure 2.

Results of serial HIV-1 enzyme-linked immunospot (ELISpot) assays for 13 infants who had a positive HIV-1–specific immune response at 1 month of age. A, Results for 9 infants who had ELISpot assays at all 5 visits. B, Results for 4 infants with incomplete follow-up ELISpot assays. Each box represents results for 1 infant, subdivided by visits. Each color bar represents a peptide epitope, specified at the bottom of each box. The Y-axes show log₁₀-transformed HIV-1–specific spot-forming units per 1 × 10⁶ peripheral blood mononuclear cells (PBMCs), and the X-axes show visits at 1, 3, 6, 9, and/or 12 months of age.

Immune response: comparison with HIV-1–infected infants, specificity, and phenotype

In a previous study, we determined HIV-1–specific immune responses at the same time points in 61 infants derived from the same cohort who acquired HIV-1 infection before 1 month of age [14]. These infected infants were excluded from the current analysis, which focuses on HIV-1–uninfected infants and later HIV-1 acquisition. HIV-1–specific responses in age-matched infants with early HIV-1 infection in the previous study were 2.5–9-fold higher in prevalence and 2–5-fold higher in magnitude than those the uninfected exposed infants analyzed here.

To determine the specificity of responses, exposed uninfected infants were compared with 20 control infants born to HIV-1–uninfected mothers. None of the 20 HIV-1–unexposed control infants (median age, 3 months; age range, 1–9 months) had a positive ELISpot result, suggesting high specificity (100% [95% confidence interval {CI}, 84%–100%]). Because exposed uninfected and infected infants underwent sampling at multiple time points and control infants were assessed only once, each time point was compared with the entire group of control infants. In exposed uninfected infants at months 1, 3, 6, 9, and 12 of age, the prevalence of ELISpot responses was higher than that among the 20 control infants (P = .2, P = .08, P = .08, P = .02, and P = .02, respectively). Among infants with early HIV-1 infection from the previous study, the prevalence of positive results was significantly higher than that in unexposed control infants at all time points (P < .001).

To confirm the phenotype of lymphocytes secreting IFN-γ in ELISpot assays, responses of CD8⁺ T cell–depleted and undepleted PBMCs were assessed in 2 separate assays of HIV-1–infected infants. In each case, after depletion of 98% of CD8⁺ T cells, positive responses were abrogated, with levels falling to that of background stimulation or lower (data not shown).

To address the potential role of maternal microchimerism, maternal HLA type was determined for a subset of 23 HIV-1–exposed uninfected infants with HIV-1–specific immune responses. Twelve (52%) of the 23 HIV-1–exposed uninfected infants had responses to epitopes selected by nonmaternal (i.e., paternal) HLA alleles, suggesting that these responses originated in the infants.

HLA-selected peptide-stimulated ELISpot assays in breast-feeding infants who were HIV-1 uninfected at month 1 of age

At 1 month of age, 141 HIV-1–uninfected infants underwent ELISpot testing and follow-up. During the subsequent follow-up, 10 acquired HIV-1 infection, 13 were lost to follow-up, 6 died, and 112 never acquired HIV-1 infection. Characteristics of these 141 infants are outlined in table 3. The median duration of zidovudine exposure was 27.0 days, and the median duration of labor was 9.0 h. The median maternal plasma HIV-1 RNA level was 4.6 log₁₀ copies/mL; this decreased to 4.0 log₁₀ copies/mL at delivery (after zidovudine), and returned to 4.7 log₁₀ copies/mL at 1 month post partum.

Table 3.

Maternal and infant characteristics for 141 breast-feeding infants with HIV-1–negative enzyme-linked immunospot results at 1 month of age and at follow-up for subsequent HIV-1 risk.

Characteristic	Subset with HLA-selected peptide-stimulated assay at 1 month of age
Delivery
Cesarean section, no. (%)	21/140 (15)
Duration of labor, h	9.0 (6–12)
Duration of ruptured    membranes, h	0.5 (0.1–2.3)
Duration of zidovudine    treatment, days	27.0 (17–37)
Episiotomy, no. (%)	14/139 (10)
Maternal characteristics
Maternal age, years	24 (21–28)
Maternal plasma HIV-1 RNA level,    log10 copies/mL
At 32 weeks of gestation	4.6 (4.2–5.1)
At delivery	4.0 (3.4–4.7)
At 1 month after delivery	4.7 (4.2–5.2)
Breast milk HIV-1 RNA level at    month 1, log10 copies/mL	2.3 (1.9–3.1)
Potential cofactors for HIV-1 exposure    after delivery
Mastitis in first month, no. (%)	10/111 (9)
Infant thrush in first month, no. (%)	12/141 (9)

NOTE. Data are median (interquartile range) values, unless otherwise indicated.

Cofactors and protective correlates for HIV-1 transmission via breast milk

In univariate analyses, maternal plasma and breast milk HIV-1 RNA levels at month 1 were associated with transmission (P = .004 and P = .03, respectively). None of the 16 infants who had a positive ELISpot result at 1 month of age acquired HIV-1 infection, versus 10 (8%) of 125 with negative ELISpot results at month 1(P = .6). In multivariate analyses, there was no association between positive ELISpot responses and protection when results were adjusted for maternal plasma HIV-1 RNA level (P = .9).

In analyses evaluating HIV-1–specific IFN-γ spot-forming units as a continuous variable, this variable showed a trend toward a protective effect in univariate analyses (P = .11) (table 4). In multivariate analyses, maternal plasma HIV-1 RNA levels and infant HIV-1–specific spot-forming units were each associated with significant independent effects on transmission. Maternal plasma HIV-1 RNA level was associated with increased transmission risk (adjusted hazard ratio [aHR], 9.1 [95% CI, 1.7–47.7]). Each log₁₀ increase in infant HIV-1–specific spot-forming units was associated with a decreased risk of transmission (aHR, 0.09 [95% CI, 0.01–0.72]). In an alternate regression model with the follow-up time shortened for women who discontinued breast-feeding early, results were similar (aHR, 0.08 [95% CI, 0.01–0.66]).

Table 4.

Cofactors for postnatal breast milk HIV-1 transmission at 1 month post partum.

Cofactora	HR (95% CI) for infant HIV-1 infection
	Univariate	Multivariateb
Maternal plasma HIV-1 RNA level	5.4 (1.7–17.2) [P= .004]	9.1 (1.7–47.7) [P= .009]
Breast milk HIV-1 RNA level	2.5 (1.1–5.6) [P = .03]	1.7 (0.7–4.4) [P = .24]
Infant HIV-1–specific response in ELISpot  assay at 1 month of age	0.4 (0.2–1.2) [P = .11]	0.09 (0.01–0.7) [P = .02]

NOTE. Ten of 141 infants acquired HIV-1 between 1 and 12 months of age. CI, confidence interval; ELISpot, enzyme-linked immunospot; HR, hazard ratio.

^aMaternal plasma and breast milk HIV-1 RNA levels were measured in log₁₀ copies per milliliter, and the infant HIV-1–specific ELISpot assay response was measured in log₁₀ HIV-1–specific spot-forming units/1 × 10⁶ peripheral blood mononuclear cells.

^bIn the multivariate analysis, the HR for each cofactor was adjusted for the other 2 cofactors.

DISCUSSION

The present prospective analysis of this HIV-1–exposed infant cohort has provided several important observations relevant to HIV-1 transmission and the mechanisms of immune protection. First, of >200 breast-feeding HIV-1–exposed uninfected infants who had no detectable HIV-1 infection throughout the first year of life and were serially assessed for immune responses, almost half had HIV-1–specific CTL IFN-γ responses despite the absence of HIV-1 infection. Second, in breast-feeding HIV-1–exposed uninfected infants who remained uninfected at ∼1 year of life, age was associated with significant increases in the prevalence and magnitude of IFN-γ responses during infancy. Finally, although detection of HIV-1–specific immune responses per se was not associated with protection from HIV-1 transmission, the magnitude of early HIV-1–specific IFN-γ responses was associated with decreased breast milk HIV-1 transmission.

Detection of HIV-1–specific CTLs in an HIV-1–exposed uninfected infant was first documented in 1992, using conventional chromium release assays with viral stimulation [2]. Subsequently, HIV-1–specific CTLs have been detected in several small cohorts of HIV-1–exposed uninfected infants (most <30 infants), most recently by investigators using intracellular cytokine staining assays [3, 4, 19]. Here, we provide the first population-based estimate of the prevalence of HIV-1–specific immune responses after exposure to HIV-1 in breast milk. We observed that 47% of HIV-1–uninfected infants had HIV-1–specific IFN-γ responses during serial evaluation over the course of 1 year. At any single time point, only 12%–22% of infants had HIV-1–specific IFN-γ responses. Serial assessment enhanced our ability to detect responses in uninfected infants, who appeared to have intermittently detectable HIV-1–specific IFN-γ responses. Thus, smaller studies with assessment at a single time point maybe limited in their ability to detect responses.

Our data suggest that, rather than completely escaping viral exposure, many HIV-1–uninfected infants born to HIV-1–infected mothers are exposed to cell-associated HIV-1 and elicit immune recognition of HIV-1–infected cells. HIV-1–infected individuals harbor large populations of replication-incompetent virus [20]. Thus, such viruses may elicit cellular responses but be limited in their ability to cause productive infection. Sacha et al. [21] have demonstrated rapid induction of CTL responses (within 2 h) that are capable of eliminating HIV-1–infected cells before protein synthesis or productive infection ensue. An alternative possibility for our observation is that infants had restricted viral replication in oropharyngeal or esophageal lymphoid tissue without systemic viral detection, similar to what has been observed in primate models after low-dose lentiviral exposure [22]. HIV-1–exposed uninfected infants had levels as high as 4535 HIV-1–specificsfu/1 × 10⁶ PBMCs in response to a single peptide epitope and responded to as many as 11 epitopes in a single assay, levels comparable to those in adults after HIV-1 vaccination [23, 24]. These responses were lower than those in infected infants [14]. Thus, the observed grading of responses—none in unexposed infants, intermediate in exposed uninfected infants, and the highest responses in infected infants—fits with a model of exposure-mediated induction of HIV-1–specific responses. In this model, mucosal exposure in uninfected exposed infants elicits transient moderate responses, whereas infected infants exposed to systemically circulating and replicating virus develop more-robust and sustained immune responses.

We observed that HIV-1–exposed uninfected infants were able to mount HIV-1–specific immune responses within the first month of life, but these responses were less frequent and of lower magnitude than those in older infants. This difference could have been due to ongoing breast-feeding exposure to HIV-1. In addition, our observations are consistent with those of studies showing age-related enhancement of immune responses in HIV-1–infected infants [14, 25, 26]. In HIV-1–infected children with intrahost circulation of HIV-1, it is challenging to discriminate between the effects of age and the effects of time since exposure to HIV-1. In a study of adults, HIV-1–exposed uninfected individuals were noted to have CTL responses as late as 34 months after exposure [27]. Infants who were potentially exposed at delivery and again via breast-feeding may be analogous to recipients of a prime-boost vaccination, which results in sustained immune responses long after exposure. The combination of enhanced immune maturation and increased exposure may have contributed to our observation of age-mediated increased responses [8].

No infants with positive ELISpot results at 1 month of age acquired HIV-1 infection, in contrast to 8% of those with negative ELISpot results. Although the difference in transmission risk was large, power for the analysis was limited by the small number of infants with HIV-1–specific immune responses at 1 month of age. Because the empiric cutoffs used to dichotomize spot-forming unit values are arbitrary (although widely used in vaccine trials) and do not have a biological basis, we also examined the HIV-1–specific immune response as a continuous parameter. This approach is similar to analyses of HIV-1 load and takes into account the fact that the assays currently do not have a defined threshold linked to the quantity of HIV-1–specific CTLs required for preventive efficacy. Indeed, when we analyzed HIV-1–specific IFN-γ spot-forming units as a log₁₀-transformed continuous variable rather than empirically creating a cutoff, statistical power was increased, and there was evidence of a protective effect. Evaluation of log₁₀ HIV-1 spot-forming units as a continuous variable rather than with arbitrary cutoffs may be useful in HIV-1 vaccine trials, both to increase power and to discern biologically relevant cutoffs for the future.

We chose the ELISpot assay because it is the most sensitive assay for detection of HIV-1–specific CTLs, requires minimal quantities of blood, and, most importantly, is the main method used to detect CTLs in vaccine studies [28, 29]. HIV-1 CTL ELISpot assays have high sensitivity and specificity compared with conventional lysis assays, and their results specifically correlate with HIV-1–specific CD8⁺ T cell recognition [28]. ELISpot results were interpreted blinded to infant HIV-1 status, using a predefined computerized algorithm from published criteria. These systems eliminated potential observer bias in determining a positive assay. Lack of responses in HIV-1–unexposed infants suggests assay specificity, in addition to controls built into assays. Depletion of CD8⁺ T cells eliminated responses almost completely, consistent with CD8⁺ T cells being the origin of responses. In addition, the peptide size (8–9mer) used for HLA-selected ELISpot assay stimulation makes it unlikely that these were T helper responses. Finally, the absence of HIV-1 infection was confirmed in infants by conducting serial assays for both HIV-1 DNA and RNA.

Limitations of the present study include the fact that it was conducted before the availability of comprehensive peptide epitopes spanning HIV-1. Because we used peptides derived from HIV-1–infected individuals, we may have underestimated the prevalence of immune responses. It has been suggested that uninfected individuals respond preferentially to peptides distinct from those eliciting responses in infected individuals [30]. Conversely, stimulation with a single peptide per pool in our study may have yielded higher sensitivity for relevant peptides than the pooled peptide assays necessary for comprehensive epitope stimulation [31]. Our control cohort (20 infants), although small, was larger than or comparable to those in other studies. In addition, all assays included internal controls with standard cut-offs [32]. Imperfect specificity may lead to an overestimated prevalence but would bias age effects or transmission analyses to the null. We did not use intracellular cytokine staining or tetramer assays. However, polyfunctional CD8⁺ T cell responses to HIV-1 in intracellular cytokine-staining assays has been demonstrated in HIV-1–exposed uninfected infants by others [3].

HIV-1–specific CTLs are thought to play a pivotal role in containing primary HIV-1 infection and may defer or attenuate infection in animal models [33–35]. However, observations in animals, in highly exposed but persistently seronegative individuals, and in HIV-1 vaccine recipients demonstrate incomplete protection by CTLs, and HIV-1 superinfection occurs in HIV-1–infected individuals despite robust CTL responses, as measured by ELISpot assay [7–9]. Thus, the utility of inducing HIV-1–specific CTL responses by protective vaccines has been debated, particularly given the results of the recent STEP trial, which may reflect the general inadequacy of HIV-1 vaccines that rely on the inducement of HIV-1–specific cellular immune responses [11]. Alternatively, the findings of the STEP trial may reflect the failure of a specific vaccine product, in which case it is possible that prime-boost vaccine approaches that induce higher magnitudes or different specificities of cellular immune responses may be at least partially protective. The association we observed between levels of HIV-1–specific IFN-γ responses and protection from breast milk HIV-1 transmission lends some support for vaccine strategies that include as a component inducing responses to HIV-1 to prevent breast milk HIV-1 transmission and possibly to prevent transmission in other settings.