Gut microbiome immaturity and childhood acute lymphoblastic leukaemia

Abstract

Acute lymphoblastic leukaemia (ALL) is the most common cancer of childhood. Here, we map emerging evidence suggesting that children with ALL at the time of diagnosis may have a delayed maturation of the gut microbiome compared with healthy children. This finding may be associated with early-life epidemiological factors previously identified as risk indicators for childhood ALL, including caesarean section birth, diminished breast feeding and paucity of social contacts. The consistently observed deficiency in short-chain fatty-acid-producing bacterial taxa in children with ALL has the potential to promote dysregulated immune responses and to, ultimately, increase the risk of transformation of preleukaemic clones in response to common infectious triggers. These data endorse the concept that a microbiome deficit in early life may contribute to the development of the major subtypes of childhood ALL and encourage the notion of risk-reducing microbiome-targeted intervention in the future.

In this Perspective article, Mel Greaves and co-workers outline emerging evidence that suggests that children with newly diagnosed acute lymphoblastic leukaemia may have a delayed maturation of the gut microbiome compared with healthy children, a deficit that might be associated with early-life epidemiological factors and could contribute to the risk of transformation of preleukaemic clones in response to common infectious triggers.

Introduction

Acute lymphoblastic leukaemia (ALL) accounts for one-third of paediatric cancer cases in developed societies¹. Although treatment of paediatric ALL is highly successful with cure rates of around 90%², the longer-term complications of therapy and impact on quality of life during treatment are substantial. Two-thirds of childhood ALL survivors face severe morbidity for decades after their disease is eradicated, as well as 20 times higher mortality compared with their healthy, age-matched counterparts³. This largely unaccounted for burden of ALL justifies the key importance of ongoing research into its aetiology and pursuit of a longer-term goal of primary prevention^4,5.

Research over recent decades has unravelled the natural history and clonal evolution of the main genetic subtypes of B cell precursor-ALL (BCP-ALL), which account for the majority of childhood ALL cases and have a peak incidence around the ages of 2–6 years. These observations have endorsed a two-stage model for this cancer originally termed the ‘delayed infection’ hypothesis^6,7, which has features in common with the so-called hygiene hypothesis proposed for allergies and type 1 diabetes⁸. The first stage is manifested by genomic lesions arising in progenitor cells in utero leading to the development of clinically covert preleukaemic clones with only modest, proliferative advantage⁹. Backtracking of BCP-ALL cases with neonatal Guthrie cards and umbilical cord blood samples as well as comparative genomics in monozygotic twins with ALL have confirmed the prenatal origin of the predominant initiating chromosomal aberrations ETS translocation variant 6 (ETV6)::runt-related transcription factor 1 (RUNX1)¹⁰ and  high hyperdiploidy^11,12. However, as shown in mouse models¹³ and human umbilical cord blood samples¹⁴, these initiating events are not sufficient for leukaemic transformation^15,16. ETV6::RUNX1 fusions (in-frame and in lymphocytes) are present in 1–5% of healthy newborn babies, but the overwhelming majority (~99%) will not develop leukaemia indicating low penetrance of the disease and the need for additional, postnatal mutational events^14,17.

The ‘delayed infection’ hypothesis predicted that persistent preleukaemic clones acquire the essential, postnatal, secondary mutations as a result of a dysregulated immune response to common infections or chronic inflammation⁷. The nature and diversity of these ‘triggering’ infections for ALL remain uncertain, although respiratory viruses are implicated by epidemiological studies^18,19. Experimental modelling data have highlighted possible mechanisms via which inflammatory cytokines might both expand preleukaemic clones and trigger the commonly observed secondary genetic changes²⁰. The highly recurrent secondary genetic changes are primarily copy-number alterations (deletions) in genes elicited by off-target immunoglobulin heavy chain V(D)J recombination-activating protein (RAG) activity²⁰. In this context, activation-induced cytidine deaminase (AID), which can be expressed in BCPs following repetitive strong inflammatory signals, has been shown to cooperate with RAG to result in genomic instability that can drive the evolution of pre-leukaemic clones¹³.

But, critically, the delayed infection model also predicted that the infection-driven dysregulated immune response triggering these crucial second hits was contingent upon a deficit of microbial exposure in infancy and a consequent failure of adequate immune network priming or maturation. Epidemiological evidence supports that contention via surrogate measures⁷. The risk of BCP-ALL is increased by caesarean section (C-section) birth^21,22, brief or absent breastfeeding^23,24 and paucity of social contacts during infancy^25–27. We note that these social risk factors are shared with type 1 diabetes and allergies, raising the possibility of a common underlying immune priming deficit²². More recently, such early-life exposures were shown to have a profound impact on the acquisition and robustness of the neonatal and infant gut microbiome^28–31, which, in turn, is recognized as fundamental to the maturation of the naive immune network of infants³². This led to the suggestion that the key risk factor of microbial underexposure (or ‘delay’) in BCP-ALL resides in the pivotal role of the microbiome and the prediction of a deficient gut microbiome in patients who develop BCP-ALL^4,7.

Recent longitudinal studies have now revealed that a delayed maturation of the gut microbiome by the age of 12 months is associated with an increased risk of asthma diagnosis by 5 years of age^33–35. Although the emerging role of the gut microbiome in the pathogenesis of childhood ALL has been discussed in recent review articles^5,7,36–41, the rarity of the disease has thus far precluded any prospective longitudinal studies. Here, we comprehensively analyse the findings of existing case–control gut microbiome studies of childhood ALL at the time of diagnosis in the context of newly discovered maturation patterns of the gut microbiome. We discuss how early-life exposures associated with an increased risk of childhood ALL can induce gut microbiome instability and perturb its maturation, which in turn can jeopardize the integrity of the immune network. Finally, we propose methods to further delineate the role of the gut microbiome in BCP-ALL pathogenesis in future clinical trials and mouse models in the era of rapidly evolving gut microbiome research.

Conservation of gut microbiome maturation

Longitudinal studies support the existence of distinct phases of gut microbiome development during early childhood^30,42,43 (Fig. 1a). The neonatal gut microbiome, which is primarily shaped by the maternal gut microbiome⁴⁴, is characterized by a high relative abundance of the phyla Proteobacteria (for example, Enterobacteriaceae) and Actinobacteria (mainly Bifidobacterium spp.)^28,30, with the latter playing a catalytic role in the foraging of glycans from human breast milk⁴⁵. Cessation of breastfeeding, introduction of a solid diet and increased social exposure are associated with maturation of the gut microbiome towards an adult-like state⁴⁶. This is characterized by a rapid expansion of Firmicutes, which become the dominant phylum beyond the first year of life (>50% relative abundance) and the primary source of short-chain fatty acids (SCFAs, especially butyrate)^47,48 (Fig. 1b). More recently, the evolutionary-conserved development of the gut microbiome in healthy children was further elucidated through the identification of multiple successive trajectories involving specific bacterial genera⁴⁹. Such strong changes in the early-life gut microbiome community are reflected by a rapid increase in species richness and evenness within individual samples (α-diversity), followed by gross stabilization by the age of 5 years^30,42,43. The gut microbiome of children also exhibits progressively fewer differences in composition compared with adults (β-diversity) over time, correlating with increased exposure to a shared environment^49,50.

Fig. 1. Early development of the gut microbiome in healthy children.

The development of the gut microbiome during early childhood consists of four different stages (acquisition, developmental, transitional and stable) that are characterized by stereotypic changes in diversity and composition³⁰. a, During the first 3 years of life, the gut microbiome exhibits increasing richness and evenness within individual samples (α-diversity)^50,56 and decreasing compositional differences compared with adults (β-diversity)⁶¹, followed by gross stabilization. Different early-life exposures can either advance or delay the maturation trajectories of the gut microbiome. b, Recent longitudinal studies of healthy children have also revealed evolutionary-conserved changes in the relative abundance of major bacterial phyla during the first 3 years of life^30,50. Depicted trends in different gut microbiome metrics are estimates on the basis of the results of three recently published studies (Supplementary Table 1).

Early-life exposures that disrupt stereotypic waves of gut colonization and induce gut microbiome instability are associated with an altered composition of the gut microbiome during infancy, while also affecting the dynamic succession and functional capacity of taxa associated with later stages of development^51–54 (Fig. 1). The mode of birth is the predominant factor shaping gut microbiota composition during the neonatal period²⁸. The primary source of the neonatal microbiome is the maternal gut, as maternal skin and vaginal microbiota colonize the newborn baby only transiently⁴⁴. At birth, neonates are colonized by strains of Bacteroides matching those of the maternal gut, irrespective of the mode of delivery⁵⁵. However, children born by C-section demonstrate reduced colonization stability of Bacteroides by day 14 (ref. ⁵⁵) and reduced abundance throughout infancy^28,30,56,57 and up to 5 years of life⁴⁹. The low-Bacteroides profile of neonates delivered by C-section is accompanied by an increased relative abundance of opportunistic pathogens (for example, Enterococcus spp. and Klebsiella spp.), which is also observed in a small proportion of vaginally delivered neonates with a similar low-Bacteroides profile²⁸. Furthermore, children born by C-section exhibit an overall reduced gut microbiome stability²⁹ and reduced stool levels of SCFAs (especially acetate)⁵⁸ during the first months of life as well as a delayed maturation of the gut microbiome by the second year of life⁴⁶.

Breastfeeding becomes the predominant factor shaping gut microbiota composition after the neonatal period and until weaning³⁰. Lack of exclusive breastfeeding and early cessation is associated with reduced relative abundance of Bifidobacterium spp.⁵⁹ and lower stool levels of acetate⁶⁰. Cessation of breastfeeding initiates the transition towards an adult-like gut microbiome composition, which is characterized by expansion of Roseburia spp. and Anaerostipes spp., restriction of Lactobacillus spp. and a functional shift towards an increased capacity to degrade complex polysaccharides⁶¹. Although breast milk may temporarily suppress maturation of the gut microbiome during infancy, children who are predominantly breastfed develop a more mature gut microbiome by the second year of life⁴⁶.

Intrapartum antibiotics have been associated with gut microbiome instability and account for substantial variation in the composition of the gut microbiome even among vaginally delivered neonates during the first week of life^28,46. Exposure is associated with reduced relative abundance of Bacteroides and reduced levels of the SCFA propionate at birth⁶², as well as increased relative abundance of potentially pathogenic Proteobacteria⁶³. Although antibiotic-induced gut microbiome instability in children and adults appears to grossly resolve within a month of exposure⁶⁴, the interaction with other host and environmental factors and the impact of the timing of exposure and repeated administration may potentiate the magnitude and duration of their effects on the gut microbiome community^64,65. Postnatal antibiotic exposure has been shown to delay the compositional and functional maturation of the gut microbiome, which is manifested through a reduced relative abundance of Lachnospiraceae (for example, Dialister spp. and Lachnospira spp.) and Ruminococcaceae and increased relative abundance of Veillonella⁴⁶.

Gut microbiome maturation can be further influenced by the immediate social environment, geography and diet. A lack of older siblings is associated with reduced gut microbiome α-diversity and reduced relative abundance of Faecalibacterium^30,66, as well as delayed maturation of the gut microbiome by the age of 12 months³³. Similarly, delayed entry into daycare is associated with reduced relative abundance of Lachnospiraceae, Ruminococcaceae and Prevotella spp. and delayed maturation of the gut microbiome³¹. Furthermore, the gut microbiome composition of children living in urban areas in Africa is more similar to European children living in urban areas compared with children living in African villages⁶⁷. School-age children living in urban areas show reduced relative abundance of bacteria (for example, Prevotella spp.) capable of fermenting complex carbohydrates for the production of SCFAs and reduced stool levels of all major SCFAs⁶⁷. These observations have been linked to a shift in dietary habits towards reduced fibre consumption, increased food variety and increased calorie intake⁶⁷.

The aforementioned findings provide evidence for the potential impact of early-life exposures with established epidemiological links to BCP-ALL on the compositional and functional maturation of the gut microbiome. The common denominator among these adverse exposures appears to be the delayed expansion of key, evolutionary-conserved bacterial taxa⁴⁶ during critical periods of immune system development, which can have a catalytic role in the development of aberrant immune responses⁶⁸. Collectively, these observations emphasize the need to meticulously characterize potential differences in the gut microbiome of children with leukaemia and healthy children.

Gut dysbiosis in children with ALL

We searched the MEDLINE and Embase databases from inception until 20 October 2022 to identify existing clinical studies examining the diversity and composition of the gut microbiome of children with ALL. We only included studies in which sample collection occurred before the administration of any systemic chemotherapy, given that chemotherapy can have a substantial impact on the diversity and composition of the gut microbiome⁶⁹. None of the identified studies differentiated between different subtypes of childhood ALL. The search strategy and a summary of the results of the database search is provided in Supplementary Table 2. Differences in the diversity and composition (relative abundance of taxa) of the gut microbiome between healthy children and children with ALL, as well as levels of statistical significance, were directly extracted from the published results of selected studies. Only for the study of Liu et al.⁷⁰, which reported differences in relative abundance at the level of bacterial species, we used publicly available individual participant data (taxon-relative abundance) to conduct a linear discriminant analysis of effect size (LEfSE) at other taxonomic levels (Supplementary methods).

Gut microbiome diversity at the time of diagnosis

We identified six case–control studies investigating gut microbiome α-diversity at the time of childhood ALL diagnosis (Table 1). The Shannon diversity index was reported to be lower in children and adolescents with ALL compared with healthy siblings or unrelated children, reaching statistical significance (P < 0.05) in four studies^70–74 (Fig. 2, top panel). Other measures of α-diversity, such as the inverse Simpson index and Chao1, showed a similar pattern (Fig. 2, top panel). Matching of cases and controls by age and antibiotic exposure varied across studies. In the study of Bai et al.⁷⁴, antibiotic exposure was associated with a diminished difference in α-diversity between ALL cases and controls, when compared with participants without antibiotic exposure in the past 90 days before sample collection. The reduction in Shannon diversity index between children with ALL and healthy children was not statistically significant in the study of Liu et al.⁷⁰, in which participants were free of antibiotics for 90 days before specimen collection. However, the mean age of participants was greater compared with other cohorts, raising the possibility that previous differences in α-diversity during the first 5 years of life were no longer detectable⁷⁰.

Table 1.

Case–control studies investigating differences in gut microbiome diversity and composition between healthy children and children with acute lymphoblastic leukaemia at the time of diagnosis

Study	Country	Age of casesa	Cases (N)	Ab (time)b % of n	Type of control	Age of controla	Control (N)	Ab (time)b % of n	Seq	HVR	Reported levels of taxonomy	Method of analysis of relative abundance	Reporting threshold for relative abundance	FDR
De Pietri et al., 2020 (ref. 71)	USA	3.7	51	NR	HS	6.8	18	No (30 days)	16s rRNA	V3–V4	NR	NR	NR	NR
Gao et al., 2020 (ref. 72)	China	6.3	18	No (14 days)	UHC	5.9	18	No (14 days)	16s rRNA	V3–V4	Genus	WMW	FC > 1.5	<0.05
Chua et al., 2020 (ref. 75)	Malaysia	2–6	7	Yes (30 days), 100%	UHC	2–6	7	No (30 days)	16s rRNA	V4	Phylum, Genus	DESeq2	FC > 4	<0.1 BHP
Rajagopala et al., 2020 (ref. 73)	USA	5.0	29	Yes (60 days), 61%	HS	6.0	23	No (60 days)	16s rRNA	V4	Phylum, Genus	DESeq2	FC > 1.5	<0.05 BHP
Bai et al., 2017c (ref. 74)	China	5.5	20	Yes (30 days), 100%	UHC	4.3	16	Yes (30 days), 100%	16s rRNA	V3–V4	Phylum, Class, Order, Family, Genus	LEfSE	LDA > 2	<0.05
		5.6	10	No (90 days)	UHC	4.5	17	No (90 days)
Liu et al., 2020 (ref. 70)	China	7.2	58	No (90 days)	UHC	7.6	23	No (90 days)	16s rDNA micro -array	V1–V9	Phylumd, Classd, Orderd, Familyd, Genusd, Species	LEfSE	LDA > 2	<0.05

Ab, antibiotics; BHP, Benjamini–Hochberg procedure; DESeq2, differential gene expression analysis based on the negative binomial distribution; FC, fold change; FDR, false discovery rate; HS, healthy sibling; HVR, hypervariable region (the region of the 16s rRNA gene targeted for sequencing); LDA, linear discriminant analysis score; LEfSE, linear discriminant analysis (LDA) of effect size; N, sample number; NR, not reported; Seq, sequencing platform; UHC, unrelated healthy child; WMW, Wilcoxon–Mann–Whitney test. ^aAge reported as arithmetic mean or range (in years). ^bTime interval within which patients were exposed to antibiotics before sample collection or antibiotic-free time interval before sample collection. ^cThe study included two independent cohorts of antibiotic-exposed and antibiotic-unexposed participants. ^dFor this taxonomic level, we used participant-level data provided by the study to determine, through linear discriminant analysis of effect size, differences in relative abundances between study groups.

Fig. 2. Case–control studies investigating differences in gut microbiome diversity between children with newly diagnosed acute lymphoblastic leukaemia and healthy children.

We systematically searched the MEDLINE and Embase databases from inception until 20 October 2022 to identify existing clinical studies examining the diversity and composition of the gut microbiome of children with newly diagnosed acute lymphoblastic leukaemia (ALL). The search strategy and the criteria for study selection are provided in Supplementary Table 2 and in Supplementary methods. Differences in α-diversity (top panels) and β-diversity (bottom panels) between children with ALL and healthy children at the time of diagnosis (de Pietri et al. 2020 (ref. ⁷¹), Gao et al. 2020 (ref. ⁵⁶), Chua et al. 2020 (ref. ⁷⁵), Rajagopala et al. 2020 (ref. ⁷³), Bai et al. 2017 (ref. ⁷⁴) and Liu et al. 2020 (ref. ⁷⁰)). Ab, antibiotics; HS, healthy sibling; NR, not reported; UHC, unrelated healthy children.

Three studies investigating differences in β-diversity of the gut microbiome between healthy children and children with ALL at the time of diagnosis reported a statistically significant Bray–Curtis dissimilarity between the two groups^70,73,75 (Fig. 2, bottom panel). A consistent difference in gut microbiome β-diversity between children with ALL and healthy children, irrespective of previous antibiotic exposure, was also reported by Bai et al.⁷⁴ on the basis of the analysis of the weighted Unifrac distance. The reduced α-diversity and the persistence of significant differences in gut microbiome composition (β-diversity) observed in children with ALL compared with healthy children are in line with a stunted progression along the developmental and transitional phases of gut microbiome maturation.

Gut microbiome composition at the time of diagnosis

We identified four case–control studies of childhood ALL examining differences in gut microbiome composition at the level of phylum. All four studies revealed a significant reduction in the relative abundance of Firmicutes in children with ALL at the time of diagnosis compared with controls (Fig. 3). In three studies, a corresponding increase in the relative abundance of the phylum Bacteroidetes was reported in children with ALL compared with healthy children (Fig. 3). The latter observation was not replicated in our re-analysis of participant-level data from the study of Liu et al.⁷⁰, in which participants were older and free of antibiotics for 90 days before specimen collection. Differences at the levels of class, order and family are presented in Supplementary Fig. 2.

Fig. 3. Case–control studies investigating differences in gut microbiome composition at the level of phylum and genus between healthy children and children with newly diagnosed acute lymphoblastic leukaemia.

We systematically searched the MEDLINE and Embase databases from inception until 20 October 2022 to identify existing clinical studies examining the diversity and composition of the gut microbiome of children with newly diagnosed acute lymphoblastic leukaemia (ALL). The search strategy and the criteria for study selection are provided in Supplementary Table 2 and in Supplementary methods. Only for the study of Liu et al. (2020), we used publicly available participant-level data to analyse differences in relative abundance between study groups at the level of phylum and genus using linear discriminant analysis of effect size (LEfSE) as described in Supplementary methods. For all other studies, data were directly extracted as reported in individual publications (Gao et al. 2020 (ref. ⁵⁶), Chua et al. 2020 (ref. ⁷⁵), Rajagopala et al. 2020 (ref. ⁷³), Bai et al. 2017 (ref. ⁷⁴) and Liu et al. 2020 (ref. ⁷⁰)). a, Differences in the relative abundance of different bacterial phyla between children with ALL at the time of diagnosis and healthy children. b, Selected bacterial genera showing consistent differences in their relative abundance in children with ALL compared with healthy children. Only genera with consistent differences between study groups across three or more independent studies are presented. The complete list of genera identified by all studies (including known developmental trajectories) is provided in Supplementary Fig. 2.c, Physiological changes in the relative abundance of the five identified genera in healthy children during early childhood as reported by Roswall et al.⁶¹. Ab, antibiotics; FC, fold change; HS, healthy sibling; LDA, linear discriminant analysis score; NR, not reported; UHC, unrelated healthy children.

Five identified studies reported differences in gut microbiome composition at the level of genus between children with newly diagnosed ALL and healthy controls (Supplementary Fig. 2, which shows all identified taxa). Despite differences in antibiotic exposure, age matching of participants, geographical location, the sequencing platform used, the method of composition analysis and reporting thresholds, five genera showed consistent differences in relative abundance between children with ALL and healthy children across three or more studies (Fig. 3b). Roseburia, Dialister, Prevotella, Faecalibacterium and Anaerostipes showed reduced relative abundance in children with ALL at the time of diagnosis compared with healthy children. Two of the identified studies used machine learning to show that differences in the relative abundance at the level of genus can effectively discriminate between newly diagnosed children with ALL and healthy children, as evident by an area under the receiver operating characteristic curve (AUC) greater than 0.8 (refs. ^70,73).

Implications for the two-hit model of ALL

Within the limitations imposed by differences in study methodology and the populations involved, the results of the identified studies suggest that the gut microbiome of children with ALL at the time of diagnosis exhibits reduced α-diversity and significantly different β-diversity compared with healthy children. They also show a decrease in the relative abundance of Firmicutes, the phylum with the most substantive expansion during the developmental phase of gut microbiome maturation (3–14 months). Furthermore, children with ALL at the time of diagnosis have reduced relative abundance of multiple genera that belong to older developmental trajectories; that is, genera which in healthy children exhibit low (<1%) relative abundance at birth and rapid expansion after weaning⁶¹ (Fig. 3c). Although these studies are somewhat preliminary with modest numbers and heterogenous design (including differences in sequencing methods, reporting thresholds and adjustment for confounders such as antibiotics), their results are in line with a suboptimal enrichment and delayed development of the gut microbiome postnatally in children with ALL and support the contention that immaturity of the gut microbiome is a key risk variable in the pathogenesis of the disease (Fig. 3). We also note that many of the affected taxa are well-known producers of SCFAs, which have a pivotal role in the regulation of gut immunity and the maintenance of an intact immune barrier^76–78. More specifically, Liu et al.⁷⁰ reported that newly diagnosed children with ALL exhibit reduced relative abundance of both butyrate-producing species (Roseburia faecis, R. intestinalis, R. inulinivorans, Anaerostipes hardus, Faecalibacterium prausnitzii, Eubacterium ramulus) and acetate-producing species (Prevotella maculosa, P. aurantiaca, Bacteroides uniformis, B. ovatus) (Supplementary Fig. 3).

The cross-sectional nature of the studies analysed, as well as their variable control for important confounders of gut microbiome composition, poses particular challenges to data interpretation and the inference of a causal relationship between gut microbiome maturation and progression of covert preleukaemic clones to overt leukaemia based solely on these preliminary data. For example, the chronological age of participants is a major determinant of gut microbiome maturation and could account for significant discrepancies in the relative abundance of different bacterial taxa between study groups⁴⁹. Similarly, recent antibiotic exposure (depending on the class of the antibacterial agent, duration of exposure and its precise timing relative to specimen collection) can cause transient disturbances in the composition of the gut microbiome^69,74,79 and has the potential to obfuscate study group differences in the relative abundance of a subset of implicated taxa. However, given that antibiotic-induced dysbiosis and reduction in SCFA-producing taxa in mice and humans are usually reversible within a few weeks of exposure^64,65, antibiotic exposure is unlikely to account for the consistent differences observed across all included studies. This notion is also supported by the results of two of the analysed case–control studies that revealed a consistent decrease in the relative abundance of Firmicutes, as well as major SCFA-producing genera in children with newly diagnosed ALL who were not exposed to antibiotics for at least 90 days before sample collection^70,74.

The consistent reduction in the relative abundance of specific bacterial taxa, and especially those belonging to older developmental trajectories across multiple studies despite variable matching of study groups (including different ethnicities, geographical locations and exposure to antibiotics) and analytic methods, supports a pervasive lag in the gut microbiome maturation of children with ALL compared with healthy children, which is likely to be long-standing and to originate from adverse exposures that occurred during the first year of life. Interestingly, longitudinal case–control studies in asthma have recently highlighted that delayed maturation of the gut microbiome (characterized by reduced relative abundance of the genera Roseburia, Dialister, Prevotella, Faecalibacterium and Blautia and increased abundance of Enterococcus), as well as reduced stool levels of butyrate by the age of 12 months, is associated with aberrant immune responses and an increased risk of asthma diagnosis by the age of 5 years^33–35. These observations suggest that gut microbiome immaturity during critical periods of immune priming increases the propensity for dysregulated immune responses⁸⁰, which in turn can pave the way for second chromosomal hits in preleukaemic clones upon exposure to common childhood infectious triggers⁷ (Fig. 4).

Fig. 4. The two-hit model of childhood B cell precursor-acute lymphoblastic leukaemia.

Proposed role of the gut microbiome in the two-hit model of childhood B cell precursor-acute lymphoblastic leukaemia (BCP-ALL) based on the results of currently available preliminary studies analysed in the present work. Chromosomal first hits are necessary events for the development of BCP-ALL, but are not sufficient to drive leukaemogenesis. The synergistic effect of adverse early-life exposures, such as caesarean section, reduced or absent breastfeeding, reduced dietary fibre^a, antibiotics^a, lack of older siblings and delayed entry into daycare, may lead to gut microbiome immaturity during a critical time in the development of the immune system. A deficiency in short-chain fatty acid (SCFA)-producing bacterial taxa can compromise gut microbiome-mediated immune priming leading to suppression of regulatory T (T_reg) cells and promotion of T helper 17 (T_H17)-dominated immune responses. SCFA deficiency can also compromise the integrity of the gut epithelial barrier and facilitate the systemic translocation of opportunistic pathogens, as well as increase susceptibility to viral infections. The resulting dysregulated pro-inflammatory immune responses against common infectious triggers can ultimately lead to an increased risk of leukaemic transformation in a small proportion (about 1%) of children with in utero-acquired preleukaemic clones. ETV6::RUNX1, ETS translocation variant 6::runt-related transcription factor 1. ^aThese variables have not been systematically evaluated as risk factors for ALL.

Adverse early-life exposures can compromise gut microbiome-mediated immune priming

The gut microbiome is a powerful mediator of the impact of early-life exposures on the development of the immune system, which also follows a stereotypical pattern of development during the first few months of life³². Νeonates normally possess a high relative frequency of myeloid-derived suppressor cells (MDSCs), CD4⁺ forkhead box P3 (FOXP3)⁺ regulatory T (T_reg) cells and regulatory B cells with a polyreactive immunoglobulin repertoire, as well as high IL-10 and IL-27 levels^32,81. This initial period of immune tolerance facilitates gut colonization with organisms that, in turn, are indispensable for subsequent development of the immune system and the maintenance of an intact gut barrier⁸⁰. Microbial signals at the level of the intestinal mucosa, such as microorganism-associated molecular patterns (MAMPs)⁶⁸ and microorganism-derived metabolites (for example, SCFAs)⁸², are thus utilized for mediating immune training^83,84. Principal aspects of this gut microbiome–immune system interplay include the induction of T_reg cells that orchestrate the complex immune network, as well as modulation of various aspects of B cell development^78,85–87.

Commensal gut microbiota have been shown to tightly regulate RAG-dependent editing in pro-B cells located in the lamina propria and to diversify the early-B cell receptor repertoire⁸⁸. Mice grown in germ-free facilities display smaller Peyer’s patches, reduced numbers of intestinal IgA-expressing B cells⁸⁹, decreased numbers of T_reg cells⁹⁰ and increased levels of IgE, which can be normalized by colonization with commensal bacteria during the first 4 weeks of life⁹¹. Similarly, perturbation of the gut microbiome by antibiotics can precipitate exaggerated T helper 17 (T_H17) immune responses to inhaled allergens in mice when exposure occurs during the first few postnatal days, but not when exposure happens in adult life⁶⁸. One of the best-studied examples of microbial-mediated immune training involves polysaccharide A, a Toll-like receptor (TLR) 2 ligand, of Bacteroides fragilis, which induces T_reg cells and suppresses pro-inflammatory T_H17 cell responses, thus promoting gut immune tolerance and host–microbial symbiosis⁹². Other Bacteroides species (for example, B. ovatus and B. uniformis), whose successful and timely colonization is determined by the mode of delivery²⁸, can have a profound impact on stool IgA levels during infancy⁹³ and thus shape the stability of other immunomodulatory bacterial taxa that colonize the infant gut at later stages of gut microbiome maturation⁹⁴. Elective C-section, an established risk factor for BCP-ALL^21,22,95, induces persisting compositional and functional instability in the gut microbiome (especially in Bacteroides spp.^28,55) that has been shown to suppress T_reg cell differentiation and compromise gut immune tolerance, leading to the promotion of pro-inflammatory responses^96,97. In this context, we note that children with ALL at the time of diagnosis were reported to have reduced relative abundance of B. vulgatus, B. ovatus and B. uniformis⁷⁰; that is, immunomodulatory bacterial species whose transmission from mother to child is disrupted by C-section²⁸.

Reduced or absent breastfeeding is another established risk factor for childhood ALL^23,24 with pronounced effects on the development of the gut microbiome–immune system axis. Maternal IgA was recently shown to set the initial homeostatic level of T_reg cells in the colon of offspring⁹⁸. Maternal IgA transferred through breastfeeding stabilizes the infant gut microbiome and maintains gut immune tolerance until the establishment of endogenous secretory IgA production^94,99, which begins after the first 30 days of life³². In parallel, human milk oligosaccharides (HMOs) facilitate the expansion of Bifidobacterium species, which have an instrumental role in the development of T cell-dependent IgA responses⁹⁹ and have been linked to higher numbers of memory B cells during infancy¹⁰⁰. Exclusive formula feeding is associated with lower numbers of T_reg cells and increased production of pro-inflammatory cytokines¹⁰¹. A low relative abundance of Bifidobacteria and depletion of HMO utilization genes in the gut microbiome of human infants was also recently shown to be associated with systemic inflammation and polarization of naive CD4⁺ T cells towards T_H17 cells, which could be reversed upon supplementation with Bifidobacterium infantis¹⁰². Geographical variations in the prevalence of specific Bifidobacterium spp. with different capacity to utilize HMOs have been associated with important differences in the overall composition of the gut microbiome community in the first year of life and may contribute to inefficient early immune priming and dysregulated immune responses in later life (for example, as encountered in allergy and autoimmunity)¹⁰³.

Diet plays a major part in determining the composition and function of the developing gut microbiome and in shaping its interactions with the host¹⁰⁴. SCFAs have emerged as being instrumental for gut immune homeostasis⁸². SCFAs, particularly butyrate, appear to be critical for the induction of IL-10-producing T_reg cells in the gut through inhibition of histone deacetylases (HDACs), which leads to the activation of the FOXP3 promoter on naive CD4⁺ T cells^105–108. Butyrate has also been shown to reduce the expression of co-stimulatory molecules on dendritic cells in response to MAMPs (for example, lipopolysaccharide (LPS))¹⁰⁹, as well as to promote the differentiation of naive B cells into regulatory B cells that are capable of suppressing inflammatory responses¹¹⁰. Fibre-derived butyrate and propionate production by gut microbiota is capable of directly suppressing AID through inhibition of HDACs and ameliorating B cell-mediated immunopathology in mouse models¹¹¹.

SCFA deficiency has been increasingly linked to impaired gut immune tolerance^76,112,113. Reduced dietary fibre consumption observed in urban societies has been associated with reduced stool levels of all major SCFAs⁶⁷, as well as aberrant activation of pro-inflammatory pathways¹¹⁴. In mice, a low-fibre diet during pregnancy and lactation leads to reduced SCFA levels and impaired thymic T_reg cell differentiation in the offspring¹¹⁵. Reduced postnatal fibre intake by the offspring has also been associated with reduced SCFA levels (especially butyrate) and induction of pro-inflammatory pathways⁶⁷. Although robust epidemiological evidence to support a role for reduced dietary fibre intake in the pathogenesis of childhood ALL is still lacking, reduced maternal fibre consumption during pregnancy has been linked to increased incidence of childhood ALL¹¹⁶. Interestingly, SCFA supplementation has also been shown to increase leptin sensitivity in mice fed a Western diet¹¹⁷. This may constitute an additional pathway through which diet and the gut microbiome can modulate the risk of childhood ALL, given that fasting may inhibit the transformation of preleukaemic clones through enhanced leptin receptor signalling in mouse models¹¹⁸.

Early-life family structure and social exposure have been associated with differential risk of developing infection-induced cancer¹¹⁹. The lack of older siblings and delayed entry into daycare have been associated with both delayed maturation of the gut microbiome^31,33 and increased risk of childhood BCP-ALL^{25–27,120–122}. Infants with older siblings exhibit increased gut microbiome α-diversity, faster gut microbiome maturation and earlier colonization with Faecalibacterium species⁵⁸, which are known to increase the T_reg/T_H17 cell ratio via inhibition of HDACs and to ameliorate gut inflammation^123,124. Living in a larger household and attending daycare has also been associated with an increased number of naive T_reg cells and reduced incidence of allergies by 12 months of age¹²⁵. Attending daycare facilities in which children are oriented to have a closer relationship to nature (for example, through extended contact with soil) can increase gut microbiome α-diversity, augment the number of T_reg cells and suppress T_H17 immune responses within 30 days of attendance¹²⁶.

In summary, social risk factors associated with increased risk of childhood ALL can compromise gut microbiome-mediated early immune priming and disrupt gut immune tolerance during critical periods of immune system development. These observations are in line with the finding that children who develop ALL have reduced levels of IL-10 and enhanced pro-inflammatory signatures at birth¹²⁷, which in turn have been linked to impaired B cell development and increased B cell DNA damage in mouse models of childhood ALL¹²⁸.

The gut microbiome and susceptibility to common infections

Emerging evidence suggests that the gut bacterial microbiome can also affect host susceptibility to viral infections, as well as clinical outcomes¹²⁹. The gut microbiota has been shown to enhance systemic antiviral immunity by regulating tonic interferon type I responses at distal sites through membrane vesicles containing bacterial DNA¹³⁰. In parallel, gut microbiome-derived SCFAs also contribute to the priming of antiviral immunity¹³¹. Increased abundance of SCFA-producing bacteria appears to offer protection against many common respiratory viruses including rhinovirus, respiratory syncytial virus, adenovirus, influenza¹³² and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)^133,134. Furthermore, the gut microbiome has emerged as an important determinant of the efficacy of oral and parenteral antiviral vaccines^135,136. Gut microbiome-derived SCFAs can facilitate antiviral immunity by regulating interferon responses and antiviral effector cells, including circulating monocytes and CD8⁺ T cells¹³⁷.

In summary, the importance of gut SCFA-producing bacterial taxa in the pathogenesis of childhood ALL is likely to be twofold. First, a delayed enrichment of these taxa during the first year of life, as a result of the synergistic effect of adverse life exposures, can compromise early immune training and shift the T_H17/T_reg cell balance towards dysregulated, pro-inflammatory responses. Second, a persisting deficiency in SCFA-producing bacterial taxa over time can additionally compromise the integrity of the gut epithelial barrier and increase host susceptibility to opportunistic pathogens, as well as commonly encountered pathogens of childhood (for example, respiratory viruses). This notion is supported by studies of mouse models of acute leukaemia in which antibiotic-induced disruption of the gut microbiome accelerates disease development, potentially through translocation of pro-inflammatory bacterial products (for example, LPS) via a leaky gut epithelial barrier that has been deprived of SCFAs^138,139. Previous experimental studies have shown that preleukaemic clones residing in the bone marrow are resistant to suppression by pro-inflammatory cytokines (IL-6, IL-1β and tumour necrosis factor (TNF))¹⁴⁰ or apoptosis by transforming growth factor β (TGFβ) secreted by bone marrow mesenchymal stromal cells (BMMSCs)¹⁴¹, leading to their survival advantage over normal BCPs. At the same time, the genome of preleukaemic clones has been shown to be increasingly vulnerable to repeated inflammatory stimuli and pro-inflammatory cytokines, which can trigger the acquisition of second chromosomal hits¹³. Evidence suggests that the gut microbiome may directly modulate the ability of BMMSCs to induce the maturation or apoptosis of haematopoietic cell lines, given that BMMSCs from germ-free mice have been shown to secrete increased pro-inflammatory cytokines (for example, IL-23)¹⁴².

Future perspectives

The low incidence of ALL compared with other childhood diseases associated with altered gut microbiome diversity and composition, such as allergy, makes the design of prospective longitudinal studies particularly challenging. A more feasible approach with potential for clinical translation is the implementation of large, cross-sectional multicentre studies encompassing a range of geographical areas, age groups, early-life social exposures, antibiotic treatments and disease cytogenetics to establish ALL-associated gut microbiome signatures. Controlling for these variables will be of paramount importance for confirming the ALL-associated gut microbiome maturation delay that has been observed in the presented preliminary studies. Given the complexity of gut microbiome communities, the multiplex co-abundance patterns and the different dispersal properties of individual bacterial taxa across human communities⁵², the use of machine learning is going to be indispensable for identifying the dynamic imprint of early-life exposures on gut microbiome maturation^143,144. Provision of publicly available datasets accompanied by thorough patient metadata and utilization of standardized reporting tools, such as the Strengthening The Organization and Reporting of Microbiome Studies (STORMS) checklist¹⁴⁵, can facilitate comprehensive meta-analyses, as well as meaningful comparisons with other diseases of childhood, including allergies and auto-immune conditions.

A necessary extension of the studies reviewed earlier will be to analyse the gut microbiome of newly diagnosed children with BCP-ALL in comparison with other types of childhood acute leukaemia (for example, T cell-ALL and acute myeloid leukaemia) to confirm the anticipated selective impact that gut microbiome dysbiosis may have on causation of subtypes of leukaemia. If this is confirmed, it will further endorse the causal link. In parallel with these patient-based studies, mouse modelling of infection-driven ALL¹³⁸ may both confirm an association between microbiome status and risk of leukaemia and provide a test bed for prevention trials with faecal microbiome or specific bacterial transplants.

Conclusion

Herein, we summarized current progress in the emerging research field exploring the role of the gut microbiome in the pathogenesis of childhood ALL. These preliminary results are consistent with a delayed maturation of the gut microbiome in children with ALL, as detected at the time of diagnosis. This raises the possibility that adverse early-life exposures associated with BCP-ALL perturb the development of the gut microbiome–immune system axis away from evolutionary-conserved maturation trajectories. We propose that the resulting deficiency of specific SCFA-producing taxa at the early stages of gut microbiome development compromises immune network stabilization, increasing the risk that later infectious exposures prompt chronic inflammation and trigger ALL. The latter will occur infrequently and only in children who carry silent pre-malignant clones generated before birth. Pending confirmation in larger studies with harmonized study design and reporting, we anticipate that the gut microbiome maturity status will likely be established as a decisive factor in the pathogenesis of the major genetic subtypes of childhood BCP-ALL and offer opportunities for primary prevention^4,5.

Finally, we note the parallels between the social risk factors and likely pivotal role of the early-life gut microbiome–immune system axis in ALL and other diseases that are increasingly more prevalent in young members of modern societies including allergies, type 1 diabetes and possibly other auto-immune diseases, such as multiple sclerosis⁴. These diseases have distinctive immunopathologies and background genetic risk variables, but they may share a common, immune priming deficiency that is microbiome dependent^4,146. This possibility requires further exploration but raises the prospect of a common prophylactic intervention strategy that could be risk-reducing for a rare disease such as ALL as well as more common debilitating illnesses of childhood. This speculative and ambitious vision is encouraged by recent clinical studies in which Bifidobacterium and Lactobacillus species have demonstrable risk-reducing efficacy in infants with sepsis¹⁴⁷, a pre-term birth¹⁴⁸ and allergies¹⁴⁹. Yet the future investigation of disease prevention via microbiome modification or boosting in infancy might benefit from more interaction between scientists and clinicians working on these different illnesses of childhood.

Glossary

α-Diversity

A measure of the number of different taxa (richness) and/or the degree of evenness in their relative abundance within a single sample.

β-Diversity

A measure of the degree of similarity or distance between the composition of the microbial communities of two samples.

Activation-induced cytidine deaminase

(AID). An enzyme required for somatic hypermutation and class-switch recombination of immunoglobulin genes during B cell maturation and immune response.

Area under the receiver operating characteristic curve

(AUC). An aggregate measure of the performance of a predictive model across all possible classification thresholds.

Bray–Curtis dissimilarity

A measure of β-diversity that quantifies the degree of dissimilarity in the composition of the microbial communities of two samples.

Chao1

A measure of α-diversity that takes into account the number of different taxa (richness) within a sample.

Human milk oligosaccharides

(HMOs). Human milk oligosaccharides are unconjugated complex glycans that have a central role in the development of the gut microbiome–immune system axis.

High hyperdiploidy

A genetic aberration characterized by chromosomal gains (>51 chromosomes) that is commonly found in preleukaemic clones of childhood B cell precursor ALL.

Inverse Simpson index

A measure of α-diversity that takes into account both richness and evenness within a sample, giving more weight to common taxa.

Leptin

A hormone produced by adipose tissue that has a central role in the regulation of energy balance and has widespread effects in multiple organ systems, including haematopoietic cells.

Linear discriminant analysis of effect size

(LEfSe). Determines the taxa most likely to explain differences between study groups. It uses standard statistical tests to detect taxa with significant difference in relative abundance between the groups, as well as additional tests to assess the biological significance and relevance of these taxa.

Microorganism-associated molecular patterns

(MAMPs). Molecular structures conserved among classes of microorganisms that can be recognized by pattern recognition receptors to elicit immune responses.

Neonatal Guthrie cards

Samples of dried blood routinely collected after birth via heel prick for the purpose of universal screening for genetic conditions.

Peyer’s patches

Gut-associated lymphoid tissue found in the small intestine that forms the interface of the gut microbiome-mediated immune system priming.

Shannon diversity index

A measure of α-diversity that takes into account both richness and evenness of taxa within a sample, giving more weight to rare taxa.

Short-chain fatty acids

(SCFAs). Metabolites produced by gut commensals through the fermentation of non-digestible fibre.

Weighted Unifrac distance

A measure of β-diversity that incorporates abundance information and places more weight to common species. By contrast, the unweighted Unifrac distance is a measure of β-diversity that takes into account the presence and absence of taxa.

Author contributions

The authors contributed equally to all aspects of the article.

Peer review

Peer review information

Nature Reviews Cancer thanks Martin Blaser, Stephen Sallan and Josef Vormoor for their contribution to the peer review of this work.

Data availability

The primary data that support the findings presented in this Perspective article, including the results of our re-analysis of participant-level data of the study of Liu et al.⁷¹, are available as Supplementary figures.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41568-023-00584-4.