Infants with late breast milk acquisition of HIV-1 generate interferon-gamma responses more rapidly than infants with early peripartum acquisition

Abstract

Infants infected with HIV-1 after the first month of life have a lower viral set-point and slower disease progression than infants infected before 1 month. We investigated the kinetics of HIV-1-specific CD8⁺ T lymphocyte secretion of interferon (IFN)-γ in infants infected before 1 month of life compared with those infected between months 1 and 12 (late infection). HIV-1 infection was assessed at birth and at months 1, 3, 6, 9 and 12 and timing of infection was determined by HIV-1 gag DNA from dried blood spots and verified by plasma HIV-1 RNA levels. HIV-1 peptide-specific IFN-γ responses were measured by enzyme-linked immunospot at months 1, 3, 6, 9 and 12. Timing of development of IFN-γ responses was compared using the log–rank test and Kaplan–Meier survival curves. Infants infected late developed HIV-1-specific CD8⁺ T cell responses 2·8 months sooner than infants infected peripartum: 2·3 versus 5·1 months after HIV-1 infection (n = 52, P = 0·04). Late-infected infants had more focused epitope recognition than early-infected infants (median 1 versus 2 peptides, P = 0·03); however, there were no differences in the strength of IFN-γ responses. In infants infected with HIV-1 after the first month of life, emergence of HIV-1-specific CD8⁺ IFN-γ responses is coincident with the decline in viral load, nearly identical to what is observed in adults and more rapid than in early-infected infants.

Introduction

Paediatric HIV-1 infection has several key features that distinguish it from adult HIV-1 pathogenesis. HIV-1-infected infants have a slower decline from peak viral loads, higher set-point viral loads, a more rapid disease course and, additionally, the age of the neonate at the time of infection influences viral replication kinetics and survival [1–3]. In a cohort of anti-retroviral naive infants, those infected <2 months of age had significantly higher mean HIV-1 viral set-points than infants infected =2 months of age [3]. Host and viral factors that may explain the differences in HIV-1 viral load between infants who acquired infection within the first month of life versus later include physiological age-related decline in CD4 T cell counts, differences in viral infectious dose and co-receptor usage of HIV-1 in cervico-vaginal secretions compared with breast milk [4–7], levels of immune stimulation in response to infant bacterial and viral co-infections and vaccinations [8] and the maturing immune system in the neonate [9]. We hypothesized that infants infected at an older age may mount a more effective immune response, as measured by interferon (IFN)-γ secretion, compared with infants infected within the first month of life. More rapid development of these responses may contribute to the delayed disease progression observed in these infants.

The perinatal transmission cohort of Nairobi, Kenya consists of infants born to HIV-1-infected women between 1999 and 2003 who received zidovudine for prevention of infant HIV-1 infection. This cohort enabled us to analyse HIV-1-specific CD8⁺ T cell responses and viral loads prospectively in a group of infants infected before the first month of life [10]. Here, we extend the analysis to include infants infected after the first month of life and compare the kinetics of HIV-1 replication and virus-specific CD8⁺ T cell responses in vertically infected infants.

Materials and methods

Study cohort

This cohort is a subset from a larger perinatal transmission cohort study conducted in Nairobi, Kenya from 1999 to 2004. Details of the cohort have been presented elsewhere [2,11]. This research complied with University of Washington, Fred Hutchinson Cancer Research Center and University of Nairobi human subjects guidelines and policies. Written informed consent was obtained from all mothers on behalf of themselves and their infants. Mothers were recruited during pregnancy from prenatal care clinics and zidovudine was initiated between 34 and 36 weeks' gestation for prevention of mother-to-child transmission of HIV-1, as was the standard of care at the time of this study. Infant peripheral blood (1–3 ml) was obtained at birth and months 1, 3, 6, 9 and 12 for HIV-1 immunity and testing. Additional blood was collected for viral load determination within 48 h of life and at months 15, 18, 21 and 24. All infants were human leucocyte antigen (HLA)-typed using the amplification refractory mutation system polymerase chain reaction (PCR) employing sequence-specific primers [12].

HIV-1 diagnosis and viral load

HIV-1 infection status was determined from dried blood using PCR amplification of HIV-1 gag DNA sequences [13] and verified by quantitative analysis of infant plasma HIV-1 RNA using a transcription-mediated amplification method sensitive for detection of multiple HIV-1 subtypes to a limit of 100 copies per ml (Gen-Probe HIV-1 Viral Load Assay, Gen-Probe Incorporated, San Diego, CA, USA) [14]. Early infection was defined as occurring before 1 month of age and categorized further as in utero (within 48 h of birth) and early peripartum (between 48 h after birth and 1 month of age). Late transmission was defined as infant HIV-1 infection after 1 month of age.

The IFN-γ enzyme-linked immunospot assay

Serial assays for HIV-1-specific IFN-γ secretion were conducted on infant peripheral blood mononuclear cells (PBMC) stimulated with individual 9- and 10-mer peptides from known HIV-1 CTL epitopes based on the infant's molecular HLA type, as described previously [10]. Briefly, IFN-γ secretion was detected from freshly isolated PBMC following overnight incubation with peptides tested in duplicate wells. HIV-1 peptide-specific spot-forming units (SFU) per 10⁶ PBMC were calculated from the average spots in peptide-stimulated wells per 10⁶ PBMC minus the average background spots from wells containing medium alone per 10⁶ PBMC. The following criteria were used to determine a positive assay: (i) a response to phytohaemagglutinin of =100 SFU after subtraction of background; (ii) HIV-specific SFU/10⁶ = 50; and (iii) SFU of peptide-stimulated wells = (2 × background control SFU) [15].

The complete peptide panel (provided by S. L. Rowland-Jones, Oxford University) contained 78 peptides that bind 29 common HLA types in Kenya, as described previously [10]. The HLA diversity in the late transmission cohort described here was narrower than in the overall cohort and the total number of peptides, and their distribution within the HIV-1 genome are: 21 (40%) peptides from gag, 14 (26%) from pol, eight (15%) from env, 9 (17%) from nef and one from (2%) rev.

Statistical analysis

Timing of infection for categorization of in utero, peripartum and late infection was defined as the mid-point between the time of the last negative and first positive tests for HIV-1. Timing of HIV-1 infection was taken as the mid-point between the last negative and first positive HIV-1 RNA or DNA result. Peak viral load was defined as the highest viral load measured within 6 months post-infection. Differences between early-infected infants and late-infected infants in time to peak viral load were analysed using an independent-samples t-test. Analyses of viral load include all HIV-1-infected infants (n = 85), while analyses of HIV-1-specific IFN-γ enzyme-linked immunospot (ELISPOT) responses is limited to those infants who had assays conducted during the first year of life (n = 74). Time to detection of IFN-γ responses was examined using the log–rank test and Kaplan–Meier survival curves. To reduce potential for bias because of less precisely defined timing of HIV-1 acquisition, we excluded infants infected in utero who had a poorly defined and potentially long interval between infection and the first IFN-γ assay at 1 month of life.

Results

Cohort and timing of HIV-1 infection

Of 510 women enrolled, 85 (17%) women had infants who became HIV-1-infected during 12 months of study follow-up. Of 85 HIV-1-infected infants, 72 (84%) were classified as early infection, with HIV-1 DNA/RNA detected during the first month of life. Thirteen (15%) infants were classified as late post-natal transmission, defined as first detection of HIV-1 DNA/RNA occurring between 1 month and 1 year of life. Timing of infection was narrowed to a 3-month window in 12 of 13 infants infected late: six (50%) were infected by 3 months of age, two (17%) were infected between 3 months and 6 months of life, three (25%) between 6 months and 9 months of life and one (8%) between 9 and 12 months of life. The remaining infant was first detected as HIV-1-infected at 3 months of age and was negative for HIV-1 DNA at birth and month 1, but plasma specimens were not available to test for HIV-1 RNA as a confirmation of the negative filter paper DNA assays.

Time–course of HIV-1-specific IFN-γ response

The IFN-γ ELISPOT assay was used as a measure of HIV-1-specific cellular immunity. The panel of peptides and frequency of positive responses are shown in Table 1. Assays were conducted on 61 of 72 (85%) infants infected before 1 month of life and analysis of the IFN-γ responses in these infants has been published [10]. Assays were conducted on 12 of 13 infected after 1 month of life (Fig. 1). The majority of late-infected infants (11, 92%) were infected with HIV-1 at or before the emergence of the virus-specific IFN-γ responses (92%), while one (8%) infant had a transient HIV-1 gag response detectable in a blood sample with undetectable HIV-1 RNA at 6 months of age (B1-264), indicating a response prior to detection of infection. Four of 12 (33%) of the infants mounted responses directed against two or more epitopes with evidence of an evolving response, while eight (66%) mounted IFN-γ responses directed against a single epitope. The magnitude of the responses ranged from 50 HIV SFU to 2520 HIV SFU, the mean magnitude of all positive responses was 439 HIV SFU and the mode response was 210 HIV SFU.

Table 1.

Panel of HIV-1 peptides, humamn leucocyte antigen (HLA) restriction and frequency of positive individuals.

GAG No. of positive individuals/ no. of individuals tested, %*		POL No. of positive individuals/ no. of individuals tested, %		ENV No. of positive individuals/ no. of individuals tested, %		REV No. of positive individuals/ no. of individuals tested, %		NEF No. of positive individuals/ no. of individuals tested, %
B14/Cw8-DRFF/YKTLRA†	2/2, 100%	B35-EPIVGAETFY	1/1, 100%	B14-ERYLKDQQL	2/2, 100%	A1-ISTERILSTY	1/2, 50%	Cw8-KAAVDLSMFL	1/2, 50%
A24-RDYVDRFY/FKTL	1/2, 50%	A2-ILKE/DPVHGV	1/4, 25%	A29-FNCGGEFFY	1/1, 100%			B42-TPQVPLRPM	2/5, 40%
A1-GSEELRSLY	1/2, 50%	A6802/A74-ETFYVDGAAN	0/4	A24-YLR/KDQQLL	1/2, 50%			B42-TPGPGI/VRYPL	2/5, 40%
B42-GPGHKARVL	2/5, 40%	A6802-ETAYFILKL	0/1	A30-IVNRVRQGY	1/3, 33%			B49-YPLTFGWCY/F	1/5, 20%
A30-RSLYNTVATLY	1/3, 33%	A2-VIYQYMDDL	0/4	A30-KYCWNLLQY	1/3, 33%			A2-PLTFGWCYKL	0/4
Cw4-KYRLKHLVW	1/3, 33%	A6802-DTVLEEMNL	0/1	B15/Cw4-SFNCGGEFF	0/3			A2-VLEWRFDSRL	0/4
B53-AS/TQEVKNWM	1/5, 20%	A30-KLNWASQIY	0/3	B35-TA/NPWNA/SSW	0/1			A2-ALKHRAYEL	0/4
A2-SLF/YNTVATL	1/4, 25%	B5801-IVLPEKDSW	0/2	Cw8-NCSFNISTSI	0/2			A24-DSRLAFHHM	0/1
Cw4-QASQEVKNW	0/3	A30-KQNPDIVIYQY	0/3					B35-VPLRPMTY	0/1
B49-FRDYVDRFY/FK	0/4	A6802-DVTLEDINL	0/1
B53-QATQEVKNW	0/5	B35-H/NPDIVIYQY	0/1
A2-TLNAWVKVI/V	0/4	Cw8-VTDSQYALGI	0/2
B53-TPQDLNM/TML	0/5	A6802/A74-ITLWQRPLV	0/4
B57/B5801-TSTLQEQIG/AW	0/2	B45-GAETFYVDGA	0/2
B53-VKNWMTETLL	0/5
B14/Cw8-DLNM/TMLNI/TV	0/2
B53-D/ETINEEAAEW	0/3
B35-PPIPVGDIY	0/1
B57-I/LSPRTLNAWL	0/2
B57/B5801-KAFSPEVIPMF	0/2
B14/Cw8-RAEQAS/TQEV	0/2

^*Expressed as the number of babies with a positive peptide response over the number of infants tested with that peptide, followed by the percentage. Assays were conducted at multiple ages per infant, thus responses to a given peptide were counted once per infant if there was ever detection of a response with a particular peptide. Peptides that stimulated positive responses in ≥50% of infants are indicated in bold type.

^†HLA restriction of each peptide precedes the sequence. Clade variants are indicated by a backslash in the amino acid sequence.

Fig. 1.

Spectrum of interferon (IFN)-γ responses in infants infected with HIV-1 after the first month of life. Twelve infants were infected with HIV-1 after the first month of life and before 1 year of life. The magnitude of the IFN-γ responses to individual peptides tested are represented by coloured bars on the left vertical axis, stacked to indicate the cumulative positive response. Note that three scales are used to represent low [500 HIV spot-forming units (SFU)], medium (1200 HIV SFU) and high (3000 HIV SFU) level responses. The HLA restriction, viral region and sequence of the peptides are indicated in the legend. The number of peptides tested at each time-point is shown above the axis or bar; ‘NT’ indicates that the individual was not tested at that time-point. HIV-1 plasma viral load was log₁₀-transformed and plotted on the right vertical axis.

To evaluate the quality of the HIV-1-specific response in infants with a similar duration of infection, we compared IFN-γ ELISPOT results from the time-point closest to 3 months post-infection in infants infected early and late. Infants infected early in life (n = 41) were tested at 3 months of age, corresponding to 3 months post-infection, while 3 months post-infection occurred at 6–12 months of age in infants infected later in life (n = 13). There was no significant difference in the strength of the IFN-γ response as measured by the median of mean and peak positive responses (P > 0·5). Three months post-infection, the median of mean (25–75th quartiles) HIV SFU was 275 (137–436) and 297 (94–638) for infants infected early compared with those infected late respectively. The median of the peak HIV SFU response was 382 (137–832) and 297 (94–768) for infants infected early compared with those infected late respectively. Although the breadth of the response was narrow in both groups, the early-infection group recognized a median of two peptides compared with a median of one peptide in the late-infection group (P = 0·03).

The interval between viral infection and stimulation of detectable immune responses may affect the ability of that immune response to contain viral replication. We modelled time to development of HIV-1-specific IFN-γ responses post-infection using survival analysis (Fig. 2). Infants infected with HIV-1 after birth but before the first month of life (n = 40) mounted their first detectable IFN-γ response a mean of 5·1 months after infection [95% confidence interval (CI), 3·9-6·3 months]. In comparison, infants infected with HIV-1 after the first month of life (n = 13) first generated a detectable IFN-γ response at a mean of 2·3 months after infection (95% CI, 1·5–3·1 months), significantly sooner than infants infected early, P = 0·04. To address if this observation was an artefact of the timing of IFN-γ assays, we compared the time between HIV-1 infection and the first IFN-γ assay, whether negative or positive, in the infants infected early compared with those infected late and found no significant difference (P = 0·4).

Fig. 2.

The proportion of infants with a positive human immunodeficiency virus (HIV)-1-specific interferon (IFN)-γ response was modelled using Kaplan–Meier curves. The solid line indicates infants who were infected after birth but before 1 month of age; the dashed line indicates infants who were infected after 1 month of age. The time since HIV-1 infection was taken as the mid-point between the last negative and first positive HIV-1 result. Infants infected with HIV-1 early in life had detectable IFN-γ responses a mean of 5·1 months after infection, 95% confidence interval (CI) (3·9, 6·3), while infants infected later in life had detectable IFN-γ responses a mean of 2·3 months after infection, 95% CI (1·5, 3·1), P = 0·04.

Time to peak viral load

As an indication of the in vivo effectiveness of the HIV-1-specific IFN-γ response, we measured the time to peak levels of HIV-1 viral replication in infants infected early compared with those infected late. There was no difference in mean time to peak viral load between the two groups of infants: the peak plasma HIV-1 viral load occurred 2·4 ± 2·0 months post-infection in infants infected early and 2·2 ± 1·4 months post-infection in infants infected late (mean ± standard deviation, P = 0·7). Interestingly, the time to detectable IFN-γ responses paralleled the time to peak viral load in infants infected late, 2·3 versus 2·2 months, while in early-infected infants the time to detectable responses occurred nearly 3 months after the peak in viral load, 5·1 versus 2·4 months.

Discussion

In this study, we demonstrate that infants infected with HIV-1 after the first month of life generate HIV-1 peptide-specific IFN-γ responses significantly sooner than infants infected peripartum. These responses are more focused in late- than peripartum-infected infants; however, the strength of responses was not significantly different from infants infected peripartum. Our data are consistent with the view that infants infected with HIV-1 after birth but before the first month of life experience a slower emergence of an HIV-1-specific CD8⁺ T cell response and high levels of viral replication, the combination of which may mute the effectiveness of the nascent CD8 response. Although infants infected after the first month of life mount HIV-1-specific responses more rapidly, we found no direct evidence that these responses modulated viral load other than the coincident emergence of IFN-γ responses with viral load, a pattern similar to HIV-1 infection in adults [17], and that the set-point viral load was significantly lower in late-infected than early-infected infants in this cohort [18]. A limitation of this study is the evaluation of a single cytokine response which has been shown to under-represent the true magnitude and capacity of the CD8⁺ T cell response to HIV-1 infection; however, our comparison group was evaluated using the same technique.

The mechanism of delayed emergence of HIV-1-specific CD8⁺ T cell responses in neonates infected after birth but before the first month of life could be attributed to differences in adaptive and/or innate immunity in the neonate [9]. Neonatal CD8⁺ T cell responses may be less efficacious because of physiological deficiencies in CD4⁺ T helper and dendritic cell function or suppressed because of high levels of circulating T regulatory cells [16,19–21]. Alternatively, a period of HIV-1-free life may allow for the more rapid development of effector memory T cells with appropriate phenotype and function that are better able to respond to HIV-1 infection, consistent with the observation of improved survival in late-infected infants [22].

Cytotoxic T lymphocytes have been associated with control of HIV-1 replication in both acute and chronic infection of adults and children [23–26]. In a recent evaluation of the HIV-2 proteome, HIV-2-specific immune responses were shown to be correlated inversely with viraemia [27]. We have demonstrated that HIV-1-specific CD8⁺ T cell responses of neonates are not associated with control of peak viral replication in acute infection of infants born to women who received azidothymidine for prevention of mother-to-child transmission [10]. Here, we observed HIV-1-specific IFN-γ responses in the infants infected after the first month of life directed towards conserved epitopes in the major homology region of gag (A24-RDYVDRFY/FKTL and B14/Cw8-DRFF/YKTLRA), responses that were rare or absent in the early-infection cohort. Interestingly, this region was targeted most frequently in HIV-2-infected individuals who were able to control viraemia [27].

CD8⁺ T cell responses have been associated with both delayed clinical progression and correlated significantly with lower plasma viral load in infants receiving anti-retrovirals [28,29]. Infant CD8⁺ IFN-γ responses have been documented to select for emergence of viral escape variants, indicative of initial immune pressure [30]. As infants infected with HIV-1 in utero or peripartum are capable of responses in the first month of life, early intervention with anti-retrovirals may provide sufficient reduction in viral load to allow for the evolution and maintenance of a more effective CD8⁺ T cell response.

In conclusion, our findings of more rapid and focused CD8⁺ T cell responses later in infancy provide a potential mechanism for the observation that infants infected months after birth have a significantly reduced set-point viral load, reduced mortality and slower disease progression than infants infected in the first month of life.