Dr Filteau Centre for International Child Health, Institute of Child Health, 30 Guilford Street, London WC1N 1EH, UK.
Previously, we have found that subclinical breast inflammation, as indicated by raised breastmilk concentrations of sodium and the inflammatory cytokine, interleukin-8 (IL-8), was highly prevalent in Bangladesh and associated with poor infant growth. In order to investigate further the prevalence of subclinical breast inflammation and to assess the impact of dietary intervention, we studied rural Tanzanian women taking part in a study of dietary sunflower or red palm oil supplementation during late pregnancy and lactation. We measured breastmilk concentrations of IL-8, the anti-inflammatory cytokine, transforming growth factor-β2 (TGF-β) and the ratio of sodium to potassium. We also estimated systemic inflammation by plasma concentrations of the acute phase proteins, α1-acid glycoprotein and C-reactive protein. There were highly significant intercorrelations among milk Na/K ratio and concentrations of IL-8 and TGF-β, the last only after treatment with bile salts which also improved TGF-β recovery in the enzyme-linked immunosorbent assay (ELISA). Plasma acute phase protein concentrations tended to correlate with milk Na/K ratio and IL-8, suggesting that subclinical breast inflammation was related to systemic inflammation. Dietary supplementation with vitamin E-rich sunflower oil but not provitamin A-containing red palm oil decreased milk Na/K, IL-8 and TGF-β at 3 months postpartum; however, the effect was significant only for Na/K ratio. The results suggest that milk Na/K ratio, IL-8, and TGF-β all measure the same phenomenon of subclinical breast inflammation but that Na/K ratio, having the lowest assay variability, is the most useful. Subclinical breast inflammation may result in part from systemic inflammation and may be improved by increased dietary intake of vitamin E-rich sunflower oil.
Breastmilk contains many immunologically active components which are usually considered to be present to promote infant health and immune development. However, some components, for example, certain cytokines, could equally be present to protect the mammary gland. Research on the mediators of mastitis has been conducted largely within the dairy field and there is evidence that tumour necrosis factor-α, granulocyte–macrophage colony-stimulating factor, complement fragment 5a and interleukin-8 (IL-8) are produced in the mammary gland primarily for maternal protection.1–5 Because breastmilk protects the infant in part by down-regulation of inflammation,6 inflammatory cytokines protecting the mammary gland might even have adverse effects on infant health.
Overt mammary inflammation (mastitis) causes the tight junctions in the mammary epithelium to become leaky such that plasma constituents can cross over into the milk.7, 8 Sodium is an easily measured and abundant plasma constituent which crosses over and elevations in milk sodium concentration, after the first 3 days postpartum and in the absence of weaning (that is, such that milk production is at least 400 ml/day), are considered indicative of mastitis.9 In previous work in rural Bangladesh, we found that in women taking part in a trial of vitamin A supplementation, 25% at 2 weeks postpartum and 12% at 3 months postpartum had milk sodium levels characteristic of mastitis.10 Because that study was not designed to look at breast health and function, we do not know why increased mammary permeability was so prevalent. However, we did find, as have others,11 that raised milk sodium was associated with elevated levels of the chemokine, IL-8, suggesting the presence of subclinical inflammation. Furthermore, we found that raised milk sodium was associated with poor infant growth between the ages of 2 weeks and 3 months in these predominantly breast-fed Bangladeshi infants. Elevated milk sodium has also been associated with poor weight gain in American infants.12 Therefore, we felt this phenomenon merited further investigation.
There is evidence from animal studies that increased dietary vitamin A13 or vitamin E and selenium14 can decrease mammary inflammation. There exist plausible mechanisms whereby both vitamins might decrease epithelial inflammation. Vitamin A has well-documented anti-inflammatory effects and also promotes epithelial development and integrity.15 Vitamin A, in its β-carotene precursor form, can also act as an antioxidant.16 Vitamin E is an important antioxidant and thus could both decrease production of inflammatory cytokines and decrease epithelial damage resulting from release of free radicals during the inflammatory response.17
Recent evidence18 shows that, at least in some parts of the world, vitamin A deficiency is a public health problem among women as well as among young children and vitamin A supplements can improve women’s health and survival. Interestingly, supplements of β-carotene were at least as effective as supplements of preformed retinol palmitate.18 We have conducted a trial of the feasibility and effectiveness of providing carotenoid-rich red palm oil (RPO) to pregnant women in rural Tanzania in order to improve their vitamin A status.19, 20 As RPO may improve vitamin A status, not only by providing carotenoids but also by providing oil, which promotes absorption of vitamin A from other foods, a second treatment group was given sunflower oil (SO). SO contains no detectable vitamin A precursors but, unlike RPO, is a rich source of α-tocopherol (vitamin E). The main outcome of interest was maternal vitamin A and carotenoid status and is described elsewhere.20 For the present work we hypothesized that supplementation with either RPO or SO would decrease breast inflammation. Because inflammatory cytokines might have adverse effects on the infant gut,5, 6 we also investigated milk levels of transforming growth factor-β (TGF-β), an anti-inflammatory cytokine in milk that is involved in wound healing21 and in the infant’s development of immune competence and oral tolerance.22, 23 TGF-β was also of interest because it has been shown to be influenced by the vitamin A metabolite, retinoic acid, in bronchial epithelial cells.24
MATERIALS AND METHODS
Study design and subjects
The study was conducted in rural Singida, a poor dry area where vitamin A-rich foods are relatively unavailable, and dietary fat intake is low.25 The prevalence among children 1–6 years old of plasma retinol below 0·7 μmol/l was 60%,25 which indicates a serious problem of vitamin A deficiency. Ethical approval for the work was obtained from the ethics committees at Oxford Brookes University and the Tanzanian Food and Nutrition Centre. Women were recruited when visibly pregnant and the study midwife had confirmed gestational age.
After informed consent, women were entered into the trial in their third trimester and allocated to one of three treatments: C – dietary advice to increase their intake of green leafy vegetables plus 4 kg rice/family/month (a small amount designed as an incentive rather than as a dietary supplement); RPO – dietary advice plus red palm oil; SO – dietary advice plus sunflower oil. Oils were provided monthly, throughout the third trimester and the first 3 months postpartum, in amounts based on family size such that there was enough for each woman to consume daily four plastic spoonfuls of oil (about 12 g, representing about 2 mg provitamin A carotenoids and 1 mg α-tocopherol per day in the RPO group and no provitamin A carotenoids plus 5·7 mg α-tocopherol per day in the SO group); this protocol ensured the woman herself received adequate amounts of oil even after she had shared it with her family. Compliance was monitored by checking remaining oil each month and, in this population where food was generally rather scarce, oils were almost always finished by the next monthly visit. The study was not blind in that it was impossible to disguise the dietary treatments, and women were allocated to treatments based on their village of residence in order to minimize cross-over between groups.
There were no significant differences between villages in a socioeconomic score derived from several variables; for example, all but one of the women were small-scale farmers, none of the villages had electricity, and all were at some distance from water supplies, especially in the dry season. Seasonal effects on diet and infection were minimal since women were recruited simultaneously in the three villages. There were no significant differences between villages in infant birth date (July–December 1997) or in the amount of time before delivery that women were given oils (75 days, 95% CI 68–83 days).
Venous blood samples were collected at recruitment and 1 and 3 months postpartum and placed on ice; plasma was separated at the base laboratory. Spot milk samples were collected at 1 and 3 months from each breast by manual expression, which most women could do readily, before and after her infant fed on each breast. Attempts were made to collect equal amounts from each breast which were then mixed giving a total of 40 ml. Plasma and milk samples were stored at −20° for a week at the base laboratory and then in Dar es Salaam where vitamin A analyses were conducted. Samples were then transported to London on dry ice where they were stored at −60° until further analysis within 6 months of collection.
Milk carotenoids and tocopherols were measured by high pressure liquid chromatography (HPLC) as described elsewhere;26 between-assay coefficients of variation for the different analytes ranged from 14 to 18%. Milk sodium was measured by flame photometry as described previously10 and expressed as a ratio to potassium in order to account for variations in fat content of spot milk samples. Interassay coefficients of variation were 2·5% for Na and 1·6% for K. Na/K ratios greater than 0·6 were considered indicative of subclinical inflammation.10 Milk IL-8 was measured by enzyme-linked immunosorbent assay (ELISA) using commercially available antibodies, as described previously.10 The interassay coefficient of variation for IL-8 assessment in our quality control sample, a transitional milk sample generously donated by a London woman, was 22%, n=10.
Milk TGF-β was measured by sandwich ELISA using a monoclonal anti-TGF-β1,β2,β3 antibody (Genzyme, Cambridge, MA) for capture, a biotinylated polyclonal anti-TGF-β2 for detection (R+D, Abingdon, UK), avidin–horseradish peroxidase as the enzyme conjugate (Sigma, Poole, UK) and recombinant human TGF-β2 as the standard (Genzyme). The assay thus detected TGF-β2, the major form in secretions,21 including milk.27 TGF-β has proved somewhat difficult to quantitate in milk and other groups have had to first centrifuge to remove fat and then acidify to activate TGF-β.27, 28 However, in preliminary work to improve the very low recovery of the TGF-β standard spiked into milk, we found, as others have found for IL-10,29 that incubation of milk with an equal volume of 12 mmol/l sodium taurocholate from bovine bile (Sigma) in distilled water for 30 min before use in the ELISA assay greatly increased the amount of TGF-β that could be detected in our quality control milk sample (0·66 ng/ml (SD=0·06, n=5), before treatment; 3·00 ng/ml (SD=1·21, n=5) after treatment or 457% of the initial value). This suggests that some TGF-β is bound in the fat portion of the milk and quantitation should be done without removing lipid. Bile salt treatment increased recovery of the standard spiked into either phosphate buffered saline (135%, n=3) or milk (469%, n=5).
These results suggest that bile salt treatment affects both the binding of TGF-β by milk fat and the detection of the recombinant standard in the ELISA assay. Assay variability, which is frequently large for ELISA assays of trace, and even of more concentrated components of milk,10, 30 became even larger for TGF-β after bile salt treatment (CV=80% for bile salt-treated; 53% for untreated; n=26). Some of this variability was because the quality control milk used happened to have little TGF-β and, before bile salt treatment, was close to the assay detection limit. We present TGF-β data both with and without bile salt treatment although, as discussed below, we believe the concentrations determined after bile salt treatment more closely represent the in vivo situation.
Plasma concentrations of the acute phase proteins, C-reactive protein (CRP) and α1-acid glycoprotein (AGP) were measured by ELISA. The AGP method has been described previously.31 CRP was measured by a sandwich ELISA using both capture and horseradish peroxidase-conjugated antibodies to CRP from Dako (Cambridge, UK) and a CRP standard from Behring Diagnostics (Milton Keynes, UK).
Data was analysed using SPSS 6·0. Cytokine concentrations and Na/K ratios were log-normally distributed so analyses were conducted on log-transformed data and geometric means are presented. In addition to standard techniques such as correlation and anova, repeated measures anova, using data from 1 and 3 months postpartum, was used to determine dietary treatment effects on milk cytokines and Na/K. This analysis permitted some of the interindividual variability (usually large for milk immunological components) to be removed. Acute phase protein concentrations and days postpartum that the sample was collected were included as covariates in these analyses.
To estimate the prevalence of subclinical breast inflammation, Na/K ratio was categorized into normal (<0·6), slightly raised (0·6–1·0) or very high (>1·0) as in our previous work.10 There were 72, 7 and 4 women in each category, respectively, at 1 month and 78, 8 and 1, respectively, at 3 months. Thus, the percentages in the combined raised groups were 13% at 1 month and 11% at 3 months. Na/K ratio and concentrations of IL-8 and TGF-β after bile salt treatment were highly intercorrelated (usually P<0·001, data not shown) at both time points. However, for untreated TGF-β, correlations with Na/K ratio and IL-8 at 1 month were not significant (P=0·86 and P=0·34, respectively) and significance was comparatively low at 3 months (P=0·01 and P=0·06, respectively). TGF-β concentrations measured after bile salt treatment correlated with pretreatment values (r=0·49, n=83, P<0·001 at 1 month; r=0·47, n=79, P<0·001 at 3 months). TGF-β concentration after bile salt treatment was a mean (95% CI) of 3·24 (2·71–3·88) times that before treatment at 1 month postpartum and 1·99 (1·60–2·48) times at 3 months postpartum. These ratios are significantly different (paired t-test, n=77, P=0·01) and result from the lower amounts of TGF-β detectable after bile salt treatment in the 3 month samples (different from 1 month, paired t-test, P=0·001), as there was no difference in TGF-β in untreated samples between the two time points.
Milk Na/K ratio was significantly correlated with plasma AGP (r=0·31, P=0·005, n=81) and CRP (r=0·23, P=0·037, n=81) at 1 month but not 3 months, and IL-8 was significantly correlated with AGP (r=0·28, P=0·013, n=77) and CRP (r=0·36, P=0·001, n=77) at 3 months but not 1 month. TGF-β, before or after bile salt treatment, did not correlate with plasma AGP or CRP at either time postpartum. Raised AGP was much more common postpartum than during pregnancy and there were no treatment group differences in the means or in the proportion with raised acute phase proteins ( Table 1). The same women tended to have raised values at both 1 and 3 months, as indicated by χ2 analysis of those with raised and normal AGP at the two times (P=0·003).
Table 1. Geometric mean plasma acute phase proteins and proportion of women with concentrations raised above normal during pregnancy and postpartum
Milk cytokines and Na/K ratios in each treatment group at both sampling time points are shown in Fig. 1. The same basic pattern is seen in each case, reflecting the intercorrelations among the variables, although a significant treatment group difference was seen only for Na/K. Women in the SO group had lower mean Na/K ratio at 3 months than women in the other groups (Duncan’s multiple range test, P<0·05). In the repeated measures anova, including as covariates plasma AGP and CRP and the number of days postpartum each sample was collected, this group difference was reflected as a significant interaction (P=0·01) between treatment group and the sampling time point, that is, 1 or 3 months postpartum. Neither cytokines nor Na/K ratios correlated significantly with milk concentrations of retinol, any of the carotenoids measured, or α-tocopherol at either 1 or 3 months (data not shown).
The study confirms and extends our previous work with lactating Bangladeshi women.10 Raised milk Na/K ratio was equally common in Bangladesh (12%) and Tanzania (11%) at 3 months postpartum. Earlier time points are not comparable because samples in Bangladesh were taken at 2 weeks postpartum (25% raised Na/K) and in Tanzania at 1 month postpartum (13% raised Na/K). We have now also measured Na/K ratio in women bringing their infants to vaccination clinics in South Africa and find an even higher prevalence of raised Na/K ratio (J. Willumsen et al. submitted). Most studies of sodium in milk of Western women have recruited small numbers of middle-class, highly selected breastfeeders and we are unaware of prevalence data from developed countries which is comparable to the community-based data from our studies.
The causes and consequences of raised milk Na/K are still under investigation. In all three cohorts – the present one in Tanzania as well as those in Bangladesh10 and South Africa (J. Willumsen et al. submitted) – in which we have measured Na/K, this ratio was highly significantly and positively correlated with concentrations of the cytokines, IL-8 and TGF-β. Thus we believe that raised milk Na/K results from an inflammation-induced permeability of the mammary epithelium, during which cytokines that both promote and down-regulate inflammation, are produced. This inflammation could have either an infectious or a sterile cause, for example, by aggravation of the breast by poor infant positioning or by tight wrapping in cloths supporting the infant, a common practice in the two African cohorts. Infections inducing inflammation could be either local in the mammary gland or systemic, as indicated by the correlations with plasma acute phase proteins. Systemic inflammation can, through cytokines, increase epithelial permeability in the gut32 and thus possibly in the breast also.
Nutritional status may influence the degree of inflammation, as suggested by the decreased Na/K ratios in women supplemented with vitamin E-rich sunflower oil. Although the intervention was not individually randomized or double-blind, because of the impossibility of disguising dietary treatments, the lack of group differences at recruitment in maternal plasma retinol or α-tocopherol20 and plasma acute-phase proteins indicates no major differences in relevant micronutrient status or inflammation initially. The decreased Na/K ratio in the SO group at only 3 months may represent a type of dose–response as nutritional interventions often take time to have measurable effects. Dietary SO supplementation increased milk α-tocopherol concentration significantly by 3 months postpartum but milk retinol did not differ among groups at either 1 or 3 months.20 Thus, the significant effect of SO but not RPO supplementation on milk Na/K ratio at 3 months may have reflected the comparative abilities of the oils, given in the fairly small but logistically feasible amounts used, to improve vitamin A or E status.
The consequences of raised Na/K ratio also require further study. In Bangladesh10 and the United States12 raised milk sodium was associated with decreased infant growth. We were not able to investigate infant growth in the present study. Several mechanisms might link raised milk sodium with poor infant growth. Sodium could be raised because weak suckling or poor positioning reduced the infant’s intake and resulted in milk stasis. Inflammatory cytokines could damage the infant gut6 and, through decreased nutrient absorption or the systemic response to infection, result in poor growth. Direct inflammatory damage may be unlikely, however, because our evidence indicates that the anti-inflammatory cytokine, TGF-β, is raised together with the inflammatory one, IL-8.
The correlations among Na/K ratio, IL-8 and TGF-β which we have seen in work in three cohorts of women suggest they are all measuring the same biological phenomenon. We believe Na/K ratio is the ‘best’ measure of this phenomenon because it has the lowest assay variability. The reasons for this are likely twofold: first, sodium and potassium occur in much larger quantities in milk than do cytokines, and second, milk fat may bind cytokines as well as increase ELISA assay variability but is destroyed by flame photometry. The fact that TGF-β correlated with Na/K ratio, as well as with IL-8, much better after than before bile salt treatment suggests that the treated TGF-β concentrations corresponded more closely than the untreated to biological reality. Furthermore, the treated concentrations will represent those to which the infant’s gut is exposed after milk ingested has met the infant’s own bile. In summary, although we would recommend bile salt treatment for estimation of milk TGF-β concentration, we would recommend Na/K ratio for estimation of subclinical breast inflammation.
Raised acute phase proteins were more common postpartum than during pregnancy and reflect the high maternal mortality often found postpartum, especially in less-developed countries where women are at risk of infections as a consequence of delivery practices.33 Subclinical breast inflammation may be one component of this postpartum morbidity, possibly resulting from systemic infection, and may also have consequences for infant health. Thus, it is promising to find that improving maternal diet may help to decrease this inflammation.