Trained immunity in newborn infants of HBV-infected mothers

The newborn immune system is characterized by an impaired Th1-associated immune response. Hepatitis B virus (HBV) transmitted from infected mothers to newborns is thought to exploit the newborns’ immune system immaturity by inducing a state of immune tolerance that facilitates HBV persistence. Contrary to this hypothesis, we demonstrate here that HBV exposure in utero triggers a state of trained immunity, characterized by innate immune cell maturation and Th1 development, which in turn enhances the ability of cord blood immune cells to respond to bacterial infection in vitro. These training effects are associated with an alteration of the cytokine environment characterized by low IL-10 and, in most cases, high IL-12p40 and IFN-α2. Our data uncover a potentially symbiotic relationship between HBV and its natural host, and highlight the plasticity of the fetal immune system following viral exposure in utero.

I nfants have higher susceptibility to severe infections than adults due to functional differences in their immune system 1,2 . Hepatitis B virus (HBV) infection is a serious global health problem that causes liver inflammation and cancer in chronically infected adults 3 . As a large part of HBV chronic infections are acquired at birth, HBV is viewed as the prototypical pathogen that is thought to hijack the immaturity of the neonatal immune system and to establish a persistent infection through the induction of an 'immunotolerant state' in the host. Data from experimental animal models (that is, HBV-transgenic mice) showing the presence of immunological defects that impair T-and B-cell priming 4-6 support this scenario.
However, the dogma of immune defects induced by HBV is at odds with the efficacy of HBV vaccination in infants born to HBV þ mothers 4 and with observations obtained in malaria-HBV co-infected young subjects in whom reduced parasitemia 7 and an increased incidence of cerebral malaria, a T-helper (Th)1mediated malaria complication 8,9 , were reported. Recent data have also shown that chronically HBV-infected adolescents labelled as 'immunotolerant' do not display any tolerogenic T-cell features 10 .
In addition to these inconsistencies between experimental data in HBV models and data obtained during natural infection, there is an increased recognition that the neonatal immune system is not defective 11 . Instead, it presents unique functional features [11][12][13] and the functional maturation of neonatal immunity can be modulated by external factors. For example, bacterial colonization and vaccinations with live vaccines can decrease infant mortality and protect them against unrelated pathogens by inducing an increased functional efficiency of their innate immune system that has been termed 'trained immunity' 14,15 .
To directly characterize the impact of HBV exposure on the newborn immune system, we performed a comprehensive immunological analysis of the cord blood (CB) cells from neonates of HBV chronically infected mothers. We report that HBV exposure in utero does not induce generic immunological defects but, on the contrary, is associated with a mature immunological profile that enhances the ability of the neonatal immune cells to respond to unrelated pathogens in vitro.
Importantly, we confirm this unique cytokine pattern (high IL-12p40 and low IL-10) in an independent cohort of CB samples from Caucasian HBV þ mothers (refer Supplementary Table 1 and Supplementary Fig. 2a,b). HBV-exposed CB plasma had increased IFN-a2 compared with healthy controls, although this difference was not statistically significant. The production of the pro-inflammatory cytokine IL-6 was minimal (o5 pg ml À 1 ) and no difference was observed in the level of TNF-a. Nevertheless, HBV-exposed neonates of Caucasian HBV þ mothers showed decreased production of IL-8 in their CB plasma than controls, even though it was not statistically significant ( Supplementary  Fig. 2a). As type-III IFN was recently reported as an innate antiviral factor produced by human primary hepatocytes in response to HBV infection 18 , we tested the production of IFN-l together with other type-I IFN (that is, IFN-b) in the CB plasma. Our data showed that although IFN-b was undetectable in all the CB plasma of both Asian and Caucasian cohorts, IFN-l was only detectable in 1 out of 18 HBV-exposed CB plasma ( Supplementary Figs 1 and 2a).
IL-12p40 has been reported to act either as a single cytokine or as a component of IL-23 or IL-12p70 (ref. 19), but we detected modest level of IL-12p70 (mean ± s.e.m. in pg ml À 1 ; healthy, 6.8±1.4, HBV, 6.5±0.1) and the absence of IL-23 in the CB plasma of neonates from HBV þ mothers ( Supplementary Fig. 1), suggesting that IL-12p40 in the CB was present as a single agonistic cytokine 19 . Furthermore, we estimated the ratio of IL-12p70/IL-10 as an indication of a Th1/Th2 balance. Our data showed a significantly higher ratio of IL-12p70/IL-10 in HBV-exposed CB plasma than controls (mean±s.e.m. ratio; HBV, 4.3 ± 2.9, healthy, 0.9 ± 0.2) (Fig. 1b). This shift in Th1 cytokine balance was also observed in the independent cohort of CB samples from neonates of Caucasian HBV þ mothers ( Supplementary Fig. 2b).
HBV exposure in utero enhances innate immune activation. The detection of elevated levels of IL-12p40, combined with the detection of low IL-10 and Th2 cytokines, does not support the hypothesis that HBV induces a state of immune tolerance in newborns. Furthermore, elevated levels of IL-12p40 has been associated with sepsis control in newborns 20 , suggesting that this cytokine might be linked with increased immunological maturity. Therefore, we first analysed the frequency of different antigenpresenting cells (APCs) in HBV-exposed and healthy CB ( Supplementary Fig. 3). The frequency of total APCs (or HLA-DR þ cells) and of the various APC subsets was not affected by HBV exposure in utero.
In contrast, the functional profile of CD14 þ monocytes, the most abundant population of innate immune cells present in the CB, is significantly different in HBV-exposed CB as compared with healthy controls. CD14 þ monocytes were sorted directly ex vivo from the CB of healthy (n ¼ 4) or HBV þ Asian mothers (n ¼ 3; 2 HBeAg À , 1 HBeAg þ ; refer Supplementary Table 1), and were analysed for the expression of 511 immune genes with Nanostring technology 21 . There were no significant differences in immune gene profile between CB monocytes of HBeAg þ and HBeAg À mothers, but notably a total of 400 immune genes were differentially expressed between HBV-exposed and healthy CB monocytes (Fig. 2a). Non-supervised hierarchical clustering showed six gene clusters corresponding to genes uniquely upregulated in healthy (cluster I, n ¼ 104) and HBV-exposed (cluster III, IV and V, n ¼ 8, n ¼ 195 and n ¼ 47, respectively) CB monocytes, as well as genes that were not distinctly enriched in either populations (cluster II and VI, n ¼ 32 and n ¼ 14, respectively) (Fig. 2a).
The differentially expressed genes between healthy and HBVexposed CB monocytes can be broadly grouped into six different gene categories (Fig. 2b). Specifically, HBV-exposed CB monocytes expressed higher levels of messenger RNA associated with major histocompatibility complex class II processing and presentation, complement components, Th1-related cytokines (including IL-12p40 encoded by the mRNA IL12B, IFN-a2, IFN-g and IL-15) and signalling molecules ( Fig. 2b and Supplementary Table 2). Furthermore, the chemokine CXCL13, whose defect in HBV transgenic models has been recently suggested to predispose to HBV chronicity 6 , was significantly upregulated in HBV-exposed CB monocytes (Fig. 2c). On the other hand, the mRNA expression of pro-inflammatory cytokines (IL-1b, IL-6, IL-8 and TNF-a) was lower in HBV-exposed CB monocytes than healthy CB monocytes ( Fig. 2b and Supplementary Table 2), further confirming the plasma cytokine data (Fig. 1a). Interestingly, the immune gene profile of HBV-exposed CB monocytes was more similar to the immune profile of healthy adult peripheral blood monocytes than that of the control CB monocytes, suggesting an increased immune maturation state of HBV-exposed CB monocytes (Fig. 2a,b). In addition, analysis of IFN-stimulated gene (ISG) expression revealed significant increase in the expression of several ISGs in HBV-exposed CB monocytes than controls ( Supplementary  Fig. 4), in line with the enhanced production of IFN-a2 in this cohort of HBV-exposed CB plasma (Fig. 1a).

ARTICLE
The expression of some of these cytokines (that is, constitutively lower levels of pro-inflammatory cytokines IL-6, IL-8 and TNF-a, and the chemokines CCL3 and CCL4) was validated at the protein level in the supernatant of ex vivo-sorted monocytes ( Supplementary Fig. 5). Direct ex vivo production of IL-12p40 or IFN-a2 was not detectable ( Supplementary Fig. 5), but after activation with TLR8 agonist (ssRNA40) 13 the production of IL-12p40 was markedly upregulated and was significantly higher in HBV-exposed CB monocytes than in controls (Fig. 2d).
Ex vivo phenotypic analysis confirmed the maturation and activation status of HBV-exposed CB monocytes. The levels of HLA-DR (HLA-class II presentation) and costimulation markers (CD40, CD80 and CD86) were significantly higher in HBVexposed CB monocytes than in controls (Fig. 2e). Functionally, HBV-exposed CB monocytes induced a higher level of proliferation of allogeneic peripheral blood mononuclear cells than healthy CB monocytes (Fig. 2f).
HBV exposure in utero induces robust Th1-polarized response. Newborn T cells produce IL-8 but are defective in Th1 cytokine production 11 . As IL-12p40 has been shown to increase IFN-g production in adult T cells, we analysed the ability of CB T cells to produce Th1 and other important T-cell cytokines (that is, IL-17, IL-21 and IL-22). Figure 3a shows the frequency of CB CD3 þ T cells producing the indicated cytokines after polyclonal stimulation, in comparison with CD3 þ T cells present in healthy or HBV-infected young adults (12-30 years). As expected, both HBV-exposed and  CIITA  C1S  C3  C4B  C1QA  C1QB  C5  CFB   CCL15  CCL3  CCL4  CCR2  CCR5  CXCL1  CXCL13  CXCL2  CXCR4   IL6  IL8  IL1A  IL1B  IL1R2  IL1RAP  IL1RN  TNF   IRAK1 IFNG  IL10  IL12A  IL12B  IL12RB1  IL15  LTA  TNFSF11  TNFSF8  TNFRSF17   IRAK4  IRF3  IRF5  IKBAP  IKBKB  IKBKG  MAP4K1  MAP4K2  MAP4K4  STAT2   TRAF6   STAT5A   HLA-DQA1  HLA-DQB1  HLA-DRA  HLA-DRB1  HLA-DRB3  HLA-DPA1  HLA-DPB1 10,000  (a) Immune gene profiling on sorted CD14 þ monocytes performed using Nanostring technology. Non-supervised hierarchical clustering of the expression of 400 immune-related genes differentially expressed between CD14 þ monocytes from healthy ('Healthy', n ¼ 4) and HBV-exposed ('HBV', n ¼ 3) CB. CD14 þ monocytes from healthy adult peripheral blood mononuclear cells (PBMCs; 'Adult', n ¼ 3) were included for comparison. Table shows the comparative mRNA levels, number of genes and representative gene categories in each gene cluster. (b) The differentially expressed genes between healthy and HBV-exposed CB monocytes can be broadly grouped into six gene categories. Graphs show the mean mRNA expression (in Nanostring counts) for representative immune genes in monocytes of healthy CB, HBV-exposed CB and adult peripheral blood within these six different gene categories. P-values between healthy and HBV-exposed CB were calculated using two-way analysis of variance with Bonferroni's post test (refer to Supplementary Table 2). (c) The mRNA expression (in Nanostring counts) of the chemokine CXCL13 in monocytes measured with Nanostring technology. (d) Sorted CD14 þ monocytes from healthy (n ¼ 4) and HBV-exposed (n ¼ 3) CB were incubated with 1 mg ml À 1 ssRNA40 (TLR8 agonist) for 18 h and IL-12p40 in the supernatant measured using luminex. Data show mean ± s.e.m. of each group. (e) The median fluorescence intensity (MFI) expression of HLA-DR, CD40 and CD80/CD86 on CD14 þ monocytes from healthy (n ¼ 5) and HBV-exposed (n ¼ 10) CB. (f) Sorted CD14 þ monocytes from healthy (n ¼ 7) or HBV-exposed (n ¼ 8) CB were incubated with allogeneic CFSE-labelled CD3 þ T cells (E:T ratio 1) for 7 days and CFSE staining analysed using flow cytometry. T-cell proliferation index was calculated using Flowjo. Horizontal line in dot plots represents the median. P-values in c-f were calculated using Mann-Whitney test, one-tailed. *Po0.05, ****Po0.0001.
The increased Th1 maturation in HBV-exposed CB was confirmed by direct ex vivo analysis of T cells expressing T-bet, the transcriptional regulator of Th1 differentiation (Fig. 3d). No differences were found between CB T cells of HBV-exposed or healthy controls in their ability to produce IL-17, IL-22 and IL-21 (Fig. 3a), even though decreased IL-21 production has been implicated in HBV vertical infection and chronicity 22 .
Despite having a more Th1-polarized response, HBV-exposed neonates do not seem to harbour any HBV-specific T cells.
Various attempts to detect HBV-specific T cells in HBV-exposed CB were unsuccessful. CB cells were analysed directly ex vivo with HBV-specific HLA tetramers or after in vitro expansion with peptides covering the whole HBV proteome. However, we were not able to detect any clear population of HBV-specific T cells in the CB of HBV-exposed neonates ( Supplementary Fig. 8).
HBV exposure in utero triggers a state of trained immunity. We next analysed whether the enhanced immune maturation detected in HBV-exposed CB could result in a better ability of the neonatal immune cells to respond to unrelated pathogens.
We tested CB mononuclear cells against Pseudomonas aeruginosa, a bacteria that can cause severe infections in underweight neonates 23 , as well as other bacteria known to be involved in neonatal sepsis in the clinics, such as uropathogenic Escherichia coli (UPEC), Salmonella typhimurium, Acinetobacter baumanii and Listeria monocytogenes. Cytokine production in the supernatant was measured after 18 h of bacteria stimulation and we detected a strong Th1 cytokine signature (IFN-g, IL-12p40 and TNF-a) in bacterial-stimulated HBV-exposed CB compared with healthy controls (Fig. 4). Specifically, the production of IFN-g was increased significantly when HBV-exposed CB cells    were challenged with UPEC, IL-12p40 production was significantly higher after exposure to P. aeruginosa, UPEC and L. monocytogenes, and TNF-a production was significantly elevated on exposure to UPEC, A. baumanii and L. monocytogenes, compared with controls. Similar trend for higher production of Th1 cytokines was observed after exposure to S. typhimurium. Therefore, our data demonstrates that HBV exposure in utero increased the nonspecific production of Th1 cytokines towards unrelated pathogen challenge in vitro.

HBV-induced immunological changes are neonatal in origin.
The increased immune maturation state detected in HBVexposed CB cells could be due to either functional changes of neonatal monocytes/T cells or increased frequency of maternal immune cells in HBV-exposed CB. Immune cells of the mother are known to cross the placenta 24 , but whether HBV infection might have an effect on the degree of CB micro-chimerism is not known. We therefore quantified the frequency of maternal cells in HBV-exposed CB using two alternative methods. We first used fluorescence in situ hybridization (FISH) to quantify the number of maternal cells (expressing XX chromosomes) among bulk cells or sorted cells (CD14 þ monocytes or CD3 þ T cells) from healthy and HBV-exposed CB of male neonates (n ¼ 2 per group). No significant differences in the frequencies of maternal cells between HBV-exposed and healthy CB were detected (Fig. 5a,b). The mean frequencies of maternal cells in healthy and HBV-exposed CB, respectively, were 0.52% versus 0.75% (bulk), 0.64% versus 1% (CD3 þ T cells) and 0.49% versus 0.5% (CD14 þ monocytes; Fig. 5a). We then used quantitative PCR on single cells to measure the expression of genes selectively expressed only by maternal (female) cells (XIST long noncoding RNA) 25 or male neonates (XKRY and TTY1 genes) 26 as an alternative method to confirm that HBV-exposed CB was not preferentially enriched with maternal cells. A total of 136 CD14 þ cells from one HBVexposed CB male neonate were analysed and we did not detect any maternal cells (0/136; Fig. 5c). This gave us a frequency of maternal cell in CD14 þ monocytes of HBV-exposed CB to be o0.7%, which is in line with the data obtained with FISH analysis (Fig. 5a) and with the reported frequency of maternal cells in healthy CB 27 . These results demonstrate that the immunological changes observed in HBV-exposed CB is unlikely to be due to an increase in maternal cell contamination, but probably due to genuine maturation of the neonatal immune cells.

HBV-induced immune maturation is associated with HBsAg.
Conventionally, prenatal HBV infection is thought to occur in a minority of cases, as HBV-DNA, a sign of active HBV replication, is only detected in the CB of a few HBeAg þ mothers 28 . On the other hand, HBV can translocate efficiently through intact trophoblastic barrier at early gestation 29 , but as the liver develops only after 12 weeks of gestation 30 HBV infection and replication in hepatocytes might not occur.
To better understand the mechanisms responsible for the functional maturation of the immune cells present in the CB of neonates born to HBV þ mothers, we tested whether HBV or HBV products can be traced in their CB plasma or mononuclear cells. HBV-DNA was detected in only two of the four CB plasma of neonates born to HBeAg þ mothers (Supplementary Table 1) and was undetected in all the other CB plasma of HBeAg À mothers. In addition, we were unable to detect the presence of HBV-DNA in the CB mononuclear cells from healthy (n ¼ 2), HBeAg À (n ¼ 2) and HBeAg þ (n ¼ 1) mothers (Supplementary  Table 1), despite the latter being tested positive for HBV-DNA in the CB plasma.
However, as in woodchuck HBV model the woodchuck HBV can persist selectively in the blood of offspring of animals with very low level of viral replication 31 , we used immunofluorescence 32 to test whether HBV or HBV products can be found in populations of purified CB immune cells. We purified CB CD2 À cells (enriched for APCs), stained them for hepatitis B surface antigen (HBsAg) and the number of HBsAg þ cells was then quantified in ten random fields at Â 20 magnification (Fig. 6a). Despite most of the neonatal plasma of HBV þ mothers were negative for HBV-DNA, HBsAg þ cells were detectable in CD2 À cells (APC-enriched cells) but not in CD2 þ (T and NK) cells at a mean frequency of 0.6 ± 0.2% (Fig. 6b). No positive immunostaining for HBsAg was detected in healthy CB (Fig. 6a,b). Thus, despite their low quantity, the presence of HBsAg þ cells in the CB of neonates of HBV þ mothers indicates that their immune system has been in contact with the virus or viral products before birth.
IL-12p40 and IFN-a2 induce CB immune cell maturation. The low frequency of HBsAg þ immune cells detected in the CB of HBV-exposed neonates suggests that it is unlikely that HBV antigens can directly cause the maturation of monocytes in HBV-exposed CB. We hypothesize that the altered cytokine environment detected in the CB plasma of HBV-exposed neonates could be responsible for the induction of monocyte/ T-cell maturation.

Discussion
The concept of trained innate immune responses has been documented in plants 33,34 , invertebrates [35][36][37] , mice [38][39][40] and, more recently, in vaccinated humans 41 . However, no such evidence has been demonstrated so far in newborns during the course of a natural viral infection. In this work, we demonstrated that HBV exposure in utero induces a state of trained immunity characterized by enhanced innate immune cell maturation and increased Th1 development. Importantly, this immune system maturation results in a better ability of the neonatal immune cells to respond to unrelated pathogen exposure. Additional immunological changes observed in HBV-exposed neonates were the higher production of IL-12p40 and lower production of IL-10 and pro-inflammatory cytokines (IL-6, IL-8 and TNF-a) in the CB plasma than controls. This immunological pattern (high Th1-related/low IL-10 and pro-inflammatory cytokines) was also observed in HBV-exposed CB monocytes. Therefore, HBV exposure in utero induced complex changes in the newborn's immune system that were not exclusively stimulatory in nature but were generally compatible with an advanced immune maturation state. Indeed, during the first year of life 12 , the infant's immune system does not only acquire a more pronounced Th1 T-cell profile but also decrease its ability to produce IL-10 and pro-inflammatory cytokines.
A further alteration induced by HBV exposure was the detection of higher levels of IFN-a2 that were only statistically significant in the CB plasma of Asian but not Caucasian HBV þ mothers. Whether such differences could be explained by different HBV genotypes infecting the two cohorts (HBV genotypes B/C in Asian patients versus HBV genotype D in Caucasian patients) will require further analysis.
Epidemiological, clinical and experimental evidences have already raised doubts about the concept of immunological tolerance during HBV infection 42 . Indeed, our data show that HBV, a virus thought to exploit the immaturity of the neonatal immune system to establish chronic infection, was unexpectedly inducing a state of 'trained immunity' with a more pronounced Th1 profile. However, in contrast to the evidences of global immune system maturation, we were not able to detect any HBVspecific T-cell response in the CB of HBV-exposed neonates. This is in contrast to the data obtained in human cytomegalovirus 43 and human immunodeficiency virus infections 44 , where virusspecific T cells can be detected in neonates. A possible scenario is that HBV has evolved a special relationship with its human host: although the defective priming of HBV-specific T cells can predispose to HBV chronicity, the induction of a trained immunity profile with a skewed Th1 response and suppression of pro-inflammatory events might have the advantage of decreasing mortality from exposure to unrelated pathogens.
Nonetheless, more data need to be gathered to fully understand the impact of vertical HBV infection on its host. A limitation of our study is that as HBV-exposed neonates must be vaccinated and treated within 24 h of birth, we were unable to investigate the consequences of the establishment of a persistent HBV infection after birth. It could be possible that the establishment of chronic HBV infection in neonates may be associated with a more robust and persistent Th1 response and a better ability to control unrelated pathogens, or it may be associated with defects in the priming of adaptive immunity, as shown in HBV-transgenic mice 6 . We posit that the evidences of trained immunity shown here are more in line with our recent demonstration that young CHB-infected patients present a fully normal Th1 T-cell profile and do not show any increased defects in HBV-specific T-cell repertoire compared with HBV-infected adults 10 , but certainly a more precise evaluation of the immunological events that are occurring in the early phases of HBV infection is needed.
The question of how HBV exposure is inducing trained immunity is still open. The low frequency of HBsAg þ monocytes detected in the CB of HBV-exposed neonates does not support the scenario of a direct HBV infection of the neonates triggering trained immunity. It is perhaps more likely to be that the cause of induction of trained immunity lies with the cytokine environment detected in the HBV-exposed newborns, characterized by an increase production of IL-12p40 and, at least in some cases, IFN-a2. Both IL-12p40 and IFN-a2 have been shown to skew T-cell development towards Th1 maturation 19,[45][46][47] , and incubation of CB cells with these recombinant cytokines in our in vitro study supports such a possibility. Epigenetic-mediated functional reprogramming of immune cells has been reported as one of the mechanisms mediating trained immunity 41 , but whether such epigenetic events are occurring in HBV-exposed neonates requires further investigations. Certainly, our in vitro assays are a profound approximation of the events that are occurring during the natural development of the newborn immune system and a clear answer to this question will probably require the use of the woodchuck hepatitis B animal model wherein viral transmission from mother to offspring can occur even in the presence of extremely low quantity of virus 31 .
The source of IL-12p40, the cytokine that was found to be consistently elevated in the CB plasma of HBV-exposed neonates, is also at the moment unknown. Ex vivo production of IL-12p40 in sorted monocytes from HBV-exposed CB was not detectable, suggesting that despite the natural contact with HBV products, the circulating monocytes were not directly responsible for this cytokine production. A possibility is that the high levels of IL-12p40 detected in the HBV-exposed CB may be produced by other types of haematopoietic phagocytic cells, such as myeloid DCs or neutrophils. Alternatively, the placenta, which consists of multiple layers of cell barriers with precise immune function and has been shown to harbour HBV and HBV products 48 , might actually be the source of IL-12p40 production in response to HBV. Pregnancy is known to modulate the natural history of HBV infection, but whether placental cells can actually play a direct role in the modulation of maternal or neonatal infections remains unknown.
Despite these limitations, our data clearly show a novel interaction between HBV and its human host. The evidences of immune system maturation in the newborns as a result of HBV exposure in utero suggests the presence of a symbiotic relationship between HBV and humans, similar to that demonstrated in mice with persistent herpes simplex virus infection 49 . This symbiotic host-virus interaction could be the explanation as to why HBV has been so efficient in co-existing in a large part of the human population from its dawn 50 .

Methods
Patients and blood samples. Umbilical CB was obtained from two independent cohort of patients: the first cohort consists of 20 neonates born to HBV-seropositive women (HBV-exposed group) and 7 neonates born to HBV-seronegative women (non-exposed control group). The deliveries occurred at the National University Hospital, Singapore, and all women were of Asian ethnicity (Chinese or Malay). The second cohort consists of eight neonates born to HBV þ women and four neonates born to HBV À women. The deliveries occurred at the Dipartimento Materno Infantile, Azienda Ospedaliero Universitaria di Parma, Italy, and the women were of Caucasian ethnicity. None of the HBV-infected mothers in both cohorts received antiviral treatment before delivery. Basic clinical and demographic data were collected at the time of delivery (Supplementary Table 1). Maternal serum was tested for HBsAg, HBeAg and HBV DNA level (a few patients). All mothers in the HBV group were positive for HBsAg and negative for human immunodeficiency virus. At delivery, CB was collected from the umbilical vein using a direct dripping method into tubes containing heparin. Subsequently, plasma was separated from whole blood and stored at À 20°C and CB mononuclear cells were isolated by density-gradient centrifugation on Ficoll-Hypaque. The study in Singapore was approved by the Domain Specific Review Board at National University Hospital, which was in accordance with the guidelines of the Singapore National Healthcare Group. The study in Italy was approved by the Comitato Etico Azienda Ospedaliero Parma (Protocol 6274) and was in accordance with the guidelines of the Italian Minister of Health. Blood samples from 10 pediatric and young adult CHB patients (12-30 years old) and 33 age-matched healthy controls used for the Th1 T-cell analysis were obtained from a viral hepatitis clinic at The Royal London Hospital, UK. Ethics approval was obtained from Barts and The London NHS Trust Ethics Review Board. All donors gave written informed consent.
Nuclei preparation and FISH analysis. Nuclei were prepared for FISH analysis by resuspending the cells in 7 ml of 0.075 mol l À 1 KCl and incubating them at 37°C water bath for 15 min. Two milliliters of 3:1 methanol:acetic acid was added to the cells, centrifuged and the pellet was resuspended and washed twice with 7 ml of methanol:acetic acid solution. Samples were stored at least overnight at À 20°C until slides were prepared. Nuclei were dropped onto methanol-cleaned slides and air dried overnight on a 56°C hot plate.
Slides pretreatment was performed in the following order: 1 Â PBS at room temperature for 5 min, pepsin/HCl solution at 37°C for 5 min, 1 Â PBS at room temperature for 5 min, 1% formaldehyde at room temperature for 10 min, 1 Â PBS at room temperature for 5 min and dehydrated in successive washes of 70%, 80% and 100% ethanol at room temperature for 2 min each and allowed to air dry. Poseidon Chromosome X and Y Satellite Enumeration Probes were obtained from Kreatech (the Netherlands) and used according to manufacturer's protocol. Nuclei were counterstained with DAPI solution. Post-hybridization washes were performed as per manufacturer's instructions. Images were visualized and captured using the Isis Fluorescence Imaging System with the Nikon Eclipse 80i microscope. The number of maternal cells were manually counted in 20 random high-power fields and expressed as a percentage of total nuclei.
High-throughput single-cell quantitative PCR. Cells were first sorted into 5 ml lysis buffer in 96-well plates (CellsDirect Resuspension and Lysis Buffer, Invitrogen) and snap-frozen with dry ice. Right before reverse transcription, samples were heated at 65°C for 90 s and immediately snap chilled on ice for 5 min. Reaction buffer (1.4 ml) and 0.7 ml enzyme (Maxima First Strand cDNA Synthesis Kit, Thermo Scientific) were added to each sample and reverse transcription was performed with the following protocol: 10 min 25°C, 15 min 50°C, 5 min 85°C. Sequence-specific pre-amplification was performed using TaqMan PreAmp Master Mix (Invitrogen, PN 4391128) by activating the enzyme at 95°C for 10 min, denaturing at 96°C for 5 s, then annealing and amplification at 60°C for 4 min for 20 cycles. Unincorporated primers were inactivated by Exonuclease I by digesting at 37°C for 30 min and inactivation at 80°C for 15 min. The resulting complementary DNA was diluted fivefold in DNA Suspension Buffer (10 mM Tris, pH 8.0, 0.1 mM EDTA; TEKnova, PN T0221) before analysis with 2 Â Sso Fast EvaGreen Supermix With Low ROX (Bio-Rad Laboratories, PN 172-5211) with nested primers in 48:48 Dynamic Arrays on a Biomark System (Fluidigm). Ct values were calculated from the system's software (Biomark Real-time PCR analysis, Fluidigm). The list of primers used is shown in Supplementary Methods.
Raw data treatment and visualization. All Raw Ct values were normalized to the assumed detection Ct level of 30, following recommendation from Fluidigm technical support. For visualization purposes, heatmaps were produced using custom scripts and ggplot package in R.