Human Immunodeficiency Virus Type 1 RNA in Breast‐Milk Components

BRIEF REPORT

Human Immunodeficiency Virus Type 1 RNA in Breast-Milk Components Irving F. Hoffman,1,3 Francis E. A. Martinson,1,3 Paul W. Stewart,2 David A. Chilongozi,1,3 Szu-Yun Leu,2 Peter N. Kazembe,3,4 Topia Banda,3,4 Willard Dzinyemba,3 Priya Joshi,1 Myron S. Cohen,1,3 and Susan A. Fiscus1,3 1 Center for Infectious Diseases and 2School of Public Health, Department of Biostatistics, University of North Carolina, Chapel Hill; 3University of North Carolina Project and 4Lilongwe Central Hospital, Department of Pediatrics, Lilongwe, Malawi

We conducted the present study to determine which of the 4 components of breast milk (whole milk, skim milk, lipid layer, and breast-milk cells) had the highest sensitivity and concentration of human immunodeficiency virus (HIV) type 1 RNA burden and to determine biological correlates to these factors. The probability of detection of HIV (sensitivity) and the concentration of HIV-1 RNA were both associated with the choice of milk component, CD4+ cell count, concentration of blood serum HIV-1 RNA, and the presence of breast inflammation. Whole milk demonstrated higher sensitivity and mean concentration than any other single component. Sensitivity was enhanced by analyzing all 4 components of breast milk. The prevalence of human immunodeficiency virus (HIV) type 1 infection among women who attend antenatal clinics in Malawi is 15%–30%. Without therapeutic intervention, up to onethird of infants born to these HIV-infected mothers acquire the virus by age 1 year [1]. An inexpensive, single-dose nevirapine regimen administered to both mother and infant has been shown to reduce peripartum transmission [2], and an implementation program is being introduced worldwide [3]. However, 195% of mothers in sub-Saharan Africa breast-feed, and no cost-effective intervention has been shown to significantly prevent breast-milk transmission of HIV. Received 9 January 2003; accepted 1 May 2003; electronically published 1 October 2003. Presented in part: 7th Conference on Retroviruses and Opportunistic Infections, 30 January– 2 February 2001, San Francisco, CA (abstract 714). Financial support: AIDS International Training and Research Program Fogarty International; National Institutes of Health (RO-1DK49381); University of North Carolina Center for AIDS Research (P30 AI50410). Reprints or correspondence: Dr. Irving F. Hoffman, Center for Infectious Diseases, University of North Carolina, 547 Burnett-Womack, CB 7030, Chapel Hill, NC 27599-7030 (hoffmani@ med.unc.edu). The Journal of Infectious Diseases 2003; 188:1209–12  2003 by the Infectious Diseases Society of America. All rights reserved. 0022-1899/2003/18808-0017$15.00

HIV-1 has been detected in both the cellular and cell-free components of breast milk. Nduati et al. [4] detected HIV-1 DNA by polymerase chain reaction (PCR) in breast-milk cells (BCs) in 58% of HIV-infected women. Ruff et al. [5] found HIV1 DNA in the BCs of 70% of infected women 0–4 days after delivery and in ∼50% of infected women 6–12 months postpartum. Lewis et al. [6] found cell-free virus by quantitative competitive–reverse-transcription PCR in only 39% of the breast milk studied. Except in cases of local breast inflammation [7], the detection rate and concentration of HIV RNA found in the blood is generally higher than that found in breast milk [6]. Most previous studies have centrifuged colostrum or breast milk to obtain BCs or cell-free skim milk (SM) for analysis. However, HIV-1 is an enveloped virus, and the virions could preferentially concentrate in the lipid layer (LL) of breast milk, which has not been investigated. Accordingly, we evaluated breast-feeding women in Malawi, to determine which component of breast milk had the highest detection rate (sensitivity) and concentrations of HIV-1 RNA and to determine the biological correlates to these factors. The present study was a pilot for a larger cohort intervention trial. Patients and methods. The study enrolled lactating women at Lilongwe Central Hospital, Lilongwe, Malawi. After we obtained written informed consent and did HIV pretest counseling, blood and breast milk were obtained. Two weeks later, a follow-up evaluation took place with posttest counseling, HIV clinic referral, and repeat blood and breast-milk collection. Baseline blood samples were tested for antibodies to HIV (HIV-1/HIV-2 EIA; Genetic Systems, and HIV-1+2; Murex) and were assayed for CD4+ cell counts (FACSCOUNT; Becton Dickinson). Follow-up HIV-positive blood samples were assayed for HIV-1 RNA copies/mL (Roche Monitor, version 1.5; Roche Diagnostics). At both visits, breast milk was collected from the right breast by manual expression into a sterile container but was collected from each breast separately if any breast inflammation was observed. The bioMeriuex NucliSens HIV-1 QT assay was used to quantitate HIV-1 RNA from breast-milk samples from HIV-positive women, according to the manufacturer’s instructions. Women with breast inflammation were immediately treated on the basis of clinical presentation. Prior to the initiation of the study, institutional review board approval was obtained from the University of North Carolina Medical School and the Health Sciences Research Committee of the Ministry of Health, Malawi. A 1-mL aliquot of whole breast milk was removed, and the remaining breast milk (9 mL) was centrifuged at 1000 g for 10 BRIEF REPORT • JID 2003:188 (15 October) • 1209

Table 1. Mean breast-milk virus load and sensitivity of detection, by breast inflammation status.

All women (n p 33)

Breast inflammation (n p 4)

No breast inflammation (n p 29)

P

Skim milk

2.03  0.57

5.89  0.79

1.91  0.41

!.0001

Lipid layer

2.46  0.47

5.88  0.76

2.27  0.36

!.0001

Breast-milk cells

1.46  0.60

4.91  0.72

1.47  0.45

!.0001

Whole milk

2.52  0.39

5.25  0.73

2.29  0.33

.0001

Skim milk

39  7

87  6

31  8

!.0001

Lipid layer

43  7

93  7

36  8

.0009

Breast-milk cells

25  6

68  12

18  6

.0019

Whole milk

46  7

94  7

38  8

.0013

Parameter, breastmilk component a

Virus load

Sensitivityb

a

Entry is the estimated mean (log10 HIV RNA copies/mL) 1 SE. Entry is 100% ⫻ the probability of detecting HIV RNA in the milk component specimen 1 SE. b

min. Aliquots of the LL and SM were removed and stored. The BCs were washed once in PBS and then resuspended in 1 mL of PBS. All aliquots of all components (whole milk [WM], SM, LL, and BCs) were stored frozen at ⫺70C until tested. BCs were not counted. Sensitivity was defined as the probability of detecting HIV in an aliquot of breast milk and was dependent on the components assayed (WM, SM, LL, or BCs). To estimate the sensitivity for each component of milk, repeated-measures logistic regression (general estimating equation 1) [8] was applied. The lower quantitation limit was set at 200 HIV RNA copies/mL. For the analyses that estimated the sensitivity for each component of milk and associations with CD4+ cell count and blood virus load, the baseline and follow-up data were pooled. To obtain inferences about the mean levels of HIV RNA (log10 copies/mL) in each component of milk, linear model methods appropriate for left-censored virus load measurements were used. In the primary analysis, the dependent variable was the larger of the 2 virus loads (maximum of baseline and followup). Explanatory variables of interest were CD4+ cell count at baseline, blood plasma virus load at follow-up, and mastitis/ inflammation occurrence on either occasion. All computations for data analysis were done using SAS System software (version 8.2; SAS) and required the use of a custom-written algorithm for the supplemental expectation-maximization procedure of Meng and Rubin [9]. Results. Thirty-three (26.6%) of 124 women enrolled in the study were HIV seropositive. The seropositive women were similar to the HIV-uninfected women in age (26.4 vs. 25.5 years), education (6.0 vs. 6.3 years), age at first pregnancy (18.3 vs. 18.3 years), and number of pregnancies (3.3 vs. 3.3). The 1210 • JID 2003:188 (15 October) • BRIEF REPORT

mean CD4+ cell count among the HIV-positive women was 437 cells/mL. At baseline, the infants born to the HIV-positive and -negative women were similar in age (12.2 vs. 12.6 months), age at first mixed feeding (4.2 vs. 4.5 months), and prevalence of any formula feeding (9% vs. 9%). Of the 33 HIV-positive women at baseline, 30 (90.9%) returned for follow-up (median follow-up period, 14 days). HIV1 RNA was detected in the blood of all 30 women at followup (median 151,187 copies/mL; range, 1990–707,190 copies/ mL). Among these 30 women, HIV-1 RNA was detected in the breast milk of 21 women (70.0%) at baseline and in 16 women (53.3%) at follow-up. Fifteen (50.0%) of 30 women had virus detected in their breast milk at both visits, 8 (26.7%) of 30 had detectable levels at neither visit, and 7 (23.3%) of 30 had detectable levels at just one visit. HIV-1 RNA assays were done for each component (WM, LL, SM, and BCs) of the 72 breast-milk specimens obtained during the 2 visits, including a separate sample from each breast when inflammation was observed. The probability of detecting HIV-1 RNA in breast-milk components (sensitivity) indicated that WM and LL provide the greatest sensitivity (table 1). These results were invariant whether the analysis was visit specific, was based on data pooled from both visits, or included only those milk specimens for which HIV-1 RNA was detected in at least 1 component. Sensitivity was associated with clinical signs of breast inflammation (table 1). Four (12.1%) of 33 subjects had clinical signs of breast inflammation: an abscess on the areola, nipple discharge, malformed nipples, or dermatitis. Accounting for local inflammation did not change the relative ranking of the components with respect to sensitivity. An association of sensitivity with a

higher blood virus burden at follow-up (P p .0001 ) and with a lower baseline CD4+ cell count (P p .0034) was also detected. The amounts of HIV-1 RNA detected in breast milk varied widely (range, 160–5,000,000 copies/mL). Mean concentrations of HIV-1 RNA in each component of breast milk are presented in table 1 and indicate that WM and LL had the greatest concentrations. The magnitude of the breast-milk virus burden was associated with clinical signs of breast inflammation (table 1). The mean virus load in WM among the 4 subjects with breast inflammation was 5.25 log10 copies/mL, compared with a mean of 2.29 log10 copies/mL for the 29 subjects with no breast inflammation. In linear regression analyses, the blood plasma virus burden and CD4+ cell count were predictive of the virus load in each component of breast milk (table 2). For each log10-unit change in blood virus load, a 2.1-log10 unit chance in WM virus load is expected—for example, the predicted WM virus load was 1.5 for subjects with a blood virus load of 4.5, but the predicted WM virus load was 3.6 for subjects with a blood virus load of 5.5. The predicted virus burden in breast milk decreased as CD4+ cell counts increased. A 0.0024-log10 HIV-1 RNA difference was predicted for each 1-unit increase in CD4+ cell count. Discussion. A realistic treatment strategy exists to dramatically reduce peripartum HIV mother-to-child transmission. However, finding a strategy that will provide a similar reduction in mother-to-child transmission due to breast-feeding will be difficult. These difficulties are a result of cultural, educational, economic, biological, scientific, and ethical problems. An increased risk of HIV breast-milk transmission is associated with the level of breast-milk virus burden and the length, intensity, and characteristics of exposure to the infant [10]. Strategies that have been suggested to reduce infant risk include antiretroviral therapy (ART) for the mother, ART prophylaxis for the infant, exclusive breast-feeding, early cessation of breastfeeding, and, as is practiced in most developed countries, substituting formula for breast-feeding. The results of randomized clinical trials have indicated that there is an urgent need to determine the best and most realistic cost-effective strategy(s) that can be implemented in populations where breast-feeding cannot be curtailed. During these investigations, it will be essential that laboratory procedures are standardized to measure HIV in breast milk so that there is comparability, the breast-milk component(s) measured have the greatest likelihood of detecting virus (sensitivity), and that the component(s) most directly reflect the risk of transmission. In the present pilot study, breast-milk HIV-1 RNA was detected in 21 (70.0%) of 30 women at baseline and in 16 (53.3%) of 30 women at follow-up. This level of detection in breast milk is notably higher than has been found in past trials, especially considering the 12-month median age of the infants in our cohort [5], and highlights the significant gain in detection power when

Table 2. Predicted difference in breast-milk virus load (log10 scale) per unit of difference in blood virus load (log10 scale) or per unit of decrease in CD4+ cell count. Parameter, breast-milk component

a

Estimate (SE)

b

P

Blood plasma virus load Skim milk

2.99 (0.93)

.0006

Lipid layer

2.62 (0.68)

!.0001

Cells

1.41 (0.65)

.0157

Whole milk

2.09 (0.54)

!.0001

CD4+ cell count Skim milk

⫺0.0075 (0.0028)

.0033

Lipid layer

⫺0.0068 (0.0022)

.0011

Cells

⫺0.0029 (0.0018)

.0490

Whole milk

⫺0.0024 (0.0012)

.0263

a

Dependent variable is the maximum virus load from baseline and followup visits. b Blood plasma virus load and CD4 cell count are not associated and have no predictive value.

all 4 breast-milk components are analyzed separately. It should be noted, however, that it was very difficult to work with the LL and that the SM may have gotten contaminated. The detection rate when only assaying the WM component was 46%. This suggests that the labor-intensive and technically difficult task of dividing and storing each breast-milk sample into its 4 components, plus the costly and time-consuming task of testing each component to glean the highest sensitivity from each sample, is advantageous, when it is feasible. Multiple assays of WM samples may also be advantageous, simpler, and less expensive, but that strategy was not addressed by our study. Previous data [11] suggested that the NucliSens assay would work best with breast-milk samples. Because the lowest detectable level (LDL) of HIV-RNA with the assay we used (NucliSens) was 200 copies/mL when using 200-mL samples of breast milk, 2-mL aliquots of WM could be used, which would reduce the LDL to 40 copies/mL and, thus, enhance sensitivity. However, the transmission probability at this low level of breastmilk RNA is unknown. At the least, as a minimum standard for detection of HIV-1 RNA in breast milk, a single assay of WM should be used, because WM represents all the breastmilk components, and, in our investigation, WM demonstrated the highest sensitivity among the individual components. Epidemiological data have suggested that transmission is greatest during the first weeks of life, but whether this is due to increased cellularity [11], an increased virus load [12] in colostrum and early milk, or the immaturity of the newborn gut [13] is unknown. In animal studies, oral transmission of HIV to newborn macaques was achieved with cell-free virus [14], whereas, in other lentivirus diseases transmitted by breast BRIEF REPORT • JID 2003:188 (15 October) • 1211

milk, infected macrophages have been shown to be the source of infection [15]. Of importance, our study not only measured cell-free RNA in WM, LL, and SM but also measured cellassociated HIV RNA in productively infected BCs, which may prove to be the most predictive measure of breast-milk transmission. Future studies will address this important question. Although the present pilot study had a sample size of only 33 HIV-infected mothers, several previously reported findings were corroborated. The concentration of (and the detection of) HIV-1 RNA in breast milk was associated with severity of immunosuppression and with blood virus load: both mean virus load and sensitivity in breast milk increased with increasing blood HIV-1 RNA concentration and with a decreasing CD4+ cell count. If high blood HIV RNA and/or low CD4+ cell count predict breast milk transmission, strategies that provide women with HIV clinical care, including a determination of their immunological and virological status, could be used to prioritize and intensify interventions, thus focusing on the women who are at the highest risk of transmission. Without the presence of local inflammation, breast milk generally has low levels of HIV-1 RNA. In our study, the difference in virus load between women with and without signs of breast inflammation was extreme (5.25 vs. 2.29 log10 copies/mL). Immediate steps should be taken in sub-Saharan African perinatal clinics to provide breast care counseling that includes information about manifestations of breast inflammation, breastfeeding practices during breast inflammation, and proper careseeking behaviors.

Acknowledgments

We thank Rakhi Kilaru, who assisted in expectation-maximization algorithm computations, and the clinical and laboratory staff of the University of North Carolina Project in Lilongwe, who did a superb job of implementing the protocol.

1212 • JID 2003:188 (15 October) • BRIEF REPORT

References 1. Miotti PG, Taha TE, Kumwenda NI, et al. HIV transmission through breastfeeding: a study in Malawi. JAMA 1999; 282:744–9. 2. Guay LA, Musoke P, Fleming T, et al. Intrapartum and neonatal singledose nevirapine compared with zidovudine for prevention of motherto-child transmission of HIV-1 in Kampala, Uganda: HIVNET 012 randomised trial. Lancet 1999; 354:795–802. 3. Buyse DS, Nuwaha FN, Karlin T, Wilfert C. Prevention of mother to child transmission of HIV: from research to action [abstract TuPeF5413]. In: Proceedings of the XIVth International AIDS Conference (Barcelona). Stockholm: International AIDS Society, 2002. 4. Nduati RW, John GC, Richardson BA, et al. Human immunodeficiency virus type 1–infected cells in breast milk: association with immunosuppression and vitamin A deficiency. J Infect Dis 1995; 172:1461–8. 5. Ruff AJ, Coberly J, Halsey NA, et al. Prevalence of HIV-1 DNA and p24 antigen in breast milk and correlation with maternal factors. J Acquir Immune Defic Syndr 1994; 7:68–73. 6. Lewis P, Nduati R, Kreiss JK, et al. Cell-free human immunodeficiency virus type 1 in breast milk. J Infect Dis 1998; 177:34–9. 7. Semba RD, Kumwenda N, Hoover DR, et al. Human immunodeficiency virus load in breast milk, mastitis, and mother-to-child transmission of human immunodeficiency virus type 1. J Infect Dis 1999; 180:93–8. 8. McCulloch CE, Searle SR. Generalized, linear, and mixed models. New York: Wiley, 2001. 9. Little RJA, Rubin DB. Statistical analysis with missing data, New York: Wiley, 2002. 10. Shepard RN, Schock J, Robertson K, et al. Quantitation of human immunodeficiency virus type 1 RNA in different biological compartments. J Clin Microbiol 2000; 38:1414–8. 11. Fowler MG. Update on perinatal HIV-1 transmission. Pediatr Clin North Am 2000; 47:21–38. 12. Rousseau C, Richardson B, Steele M, et al. Breastmilk HIV-1 viral load is associated with viral load in other compartments, host genotype, and perinatal transmission [abstract 792-W]. In: Program and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections (Seattle). Alexandria, VA: Foundation for Retrovirology and Human Health, 2002:340. 13. Walker WA. Antigen absorption from the small intestine and gastrointestinal disease. Pediatr Clin North Am 1975; 22:731–46. 14. Baba TW, Jeong YS, Penninck, Bronson R, Greene MF, Ruprecht RM. Pathogenicity of live, attenuated SIV after mucosal infection of neonatal macaques. Science 1995; 267:1820–5. 15. Oxtoby MJ. Human immunodeficiency virus and other viruses in human milk: placing the issue in broader perspective. Pediatr Infect Dis J 1988; 7:825–35.