- Email: [email protected]
Association of serum bilirubin in newborns affected by jaundice with gut microbiota dysbiosis
Accepted Manuscript Association of serum bilirubin in newborns affected by jaundice with gut microbiota dysbiosis
Shaoming Zhou, Zhangxing Wang, Fusheng He, Huixian Qiu, Yan Wang, Huihui Wang, Jianli Zhou, Jiaxiu Zhou, Guoqiang Cheng, Wenhao Zhou, Ruihuan Xu, Mingbang Wang PII: DOI: Reference:
S0955-2863(18)30264-X doi:10.1016/j.jnutbio.2018.09.016 JNB 8065
To appear in:
The Journal of Nutritional Biochemistry
Received date: Revised date: Accepted date:
29 March 2018 10 September 2018 12 September 2018
Please cite this article as: Shaoming Zhou, Zhangxing Wang, Fusheng He, Huixian Qiu, Yan Wang, Huihui Wang, Jianli Zhou, Jiaxiu Zhou, Guoqiang Cheng, Wenhao Zhou, Ruihuan Xu, Mingbang Wang , Association of serum bilirubin in newborns affected by jaundice with gut microbiota dysbiosis. Jnb (2018), doi:10.1016/j.jnutbio.2018.09.016
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Association of serum bilirubin in newborns affected by jaundice with gut microbiota dysbiosis Shaoming Zhou1*#; Zhangxing Wang2*; Fusheng He3*; Huixian Qiu4*; Yan Wang3; Huihui Wang5; Jianli Zhou1; Jiaxiu Zhou6; Guoqiang Cheng7; Wenhao Zhou7; Ruihuan Xu8#; Mingbang Wang9#;
PT E
D
MA
NU
SC
RI
PT
1. Division of Gastroenterology, Shenzhen Children’s Hospital, Shenzhen, Guangdong, 518038, China; 2. Division of Neonatology, Shenzhen Longhua People's Hospital, Guangdong, 518109, China; 3. Imunobio, Shenzhen, Guangdong, China, 518001; 4. Division of Neonatology, Longgang Central Hospital of Shenzhen, Guangdong, 518116, China; 5. Division of clinical nutrition, Shenzhen Children’s Hospital, Shenzhen, Guangdong, 518038, China; 6. Division of psychology, Shenzhen Children’s Hospital, Shenzhen, Guangdong, 518038, China; 7. Division of Neonatology, Children’s Hospital of Fudan University, National Center for Children's Health, Shanghai, 201102, China; 8. Clinical Laboratory, Longgang Central Hospital of Shenzhen, Guangdong, 518116, China; 9. Xiamen branch, Shanghai Key Laboratory of Birth Defects, Division of Neonatology, Children’s Hospital of Fudan University, National Center for Children's Health, Shanghai, 201102, China;
CE
* Contributed equally.
#Corresponding Author: Dr. Mingbang Wang ([email protected]), Dr.
AC
Ruihuan Xu ([email protected]) or Dr. Shaoming Zhou ([email protected]);
Author Contributions: All authors contributed to conception and design of the project, including collection of samples, clinical evaluation, sequencing and data analysis. M.W., R.X and S.Z. drafted and revised the manuscript. All authors approved the final version of the manuscript. Conflict of Interest Disclosures：There is no conflict of interest to declare.
ACCEPTED MANUSCRIPT BACKGROUND & AIMS Breast milk jaundice (BMJ) is common and benign, but neonatal cholestasis (NC) is rare and not benign, so early differentiation between NC and non-NC jaundice is important and may facilitate diagnosis and treatment. Gut microbiota plays an important role in enterohepatic circulation, which in turn plays an
PT
important role in the secretion of bilirubin. We aimed to determine the composition of
RI
gut microbiota in patients with NC and BMJ, and to identify the gut microbiota
SC
composition associated with NC and BMJ.
NU
METHODS Data on age, gender, delivery, feeding mode, serum total bilirubin, direct bilirubin, and liver function were collected for NC patients, BMJ patients and healthy
MA
controls ， respectively. Shotgun metagenomic sequencing and metagenome-wide
D
association were performed.
PT E
RESULTS Forty NC patients, 16 patients affected by BMJ, and 14 healthy controls (CON) without jaundice were enrolled. A significant increase in species richness,
CE
especially Bacteroides, was found in NC patients. The abundances of potentially pathogenic species and KEGG orthologies (KOs) of virulence factor genes were
AC
positively correlated with serum bilirubin level. The abundances of nine species of Bifidobacterium and three KOs of galactose metabolism were significantly decreased in the jaundice group (NC and BMJ) and were negatively correlated with serum bilirubin level.
CONCLUSIONS The gut microbiota in NC patients is characterized by a significant increase in species richness, possibly due to the proliferation of potentially pathogenic
ACCEPTED MANUSCRIPT species.
Additionally, the gut microbiota in jaundice patients is characterized by a decreased abundance of Bifidobacterium. Decreased Bifidobacterium has been associated with
PT
elevated bilirubin and abnormal gut microbiota galactose metabolic pathway. Further,
RI
ten bacteria species were identified as potential biomarker of jaundice.
SC
Keywords
MA
NU
Neonatal cholestasis; breast milk jaundice; gut microbiota; shotgun metagenomics
Key Points
D
Question Is there any alteration of gut microbiotain neonatal cholestasis patients?
PT E
Does gut microbiota have any involvement in the occurrence of neonatal cholestasis
CE
or breast milk jaundice?
Findings The alteration of gut microbiota in neonatal cholestasis patients mainly
AC
manifested as a significant increase in species richness and an increased abundance of potentially pathogenic species, while the main manifestation in jaundice patients was a significant decrease in Bifidobacterium which may be involved in the metabolism of bilirubin through the galactose metabolic pathway.
Meaning The results suggest that an imbalance of gut microbiota exist in neonatal cholestasis and breast milk jaundice patients, primarily in the form of a substantial
ACCEPTED MANUSCRIPT reduction in the abundance of Bifidobacterium, suggesting the possibility of intervention treatment for neonatal cholestasis and breast milk jaundice by
PT
supplementing probiotics.
RI
Introduction
SC
Physiological jaundice and breast milk jaundice (BMJ) commonly occur in the neonatal period due to an increase in serum unconjugated bilirubin, with most cases
NU
being benign [1, 2]. However, neonatal cholestasis (NC) or cholestatic jaundice
MA
possibly affects 1/2500 newborns, characterized by an increase in serum conjugated bilirubin or direct bilirubin. It is important to differentiate NC from noncholestatic
D
jaundice because NC is more likely to involve a severe condition such as abnormal
PT E
liver function and require prompt diagnosis and treatment [3, 4]. The most commonly known causes of NC are biliary atresia and neonatal hepatitis, followed by preterm
CE
birth and inborn errors of metabolism [5]. The pathogenesis of NC involves the liver failing to adequately secrete bilirubin from the blood into the bile [5]. Given that the
AC
gut microbiota plays an important role in human health, its disruption in early infancy may have a significant impact on the maturation of the immune system and on the health status in adulthood [6]; enterohepatic circulation also plays an important role in the excretion of bilirubin [7]. Moreover, the gut microbiota plays an important role in liver disease through the gut–liver axis [8]. Therefore, it is important to understand the composition of the gut microbiota in patients with NC and BMJ. In the present
ACCEPTED MANUSCRIPT study, we recruited 40 NC patients and 16 affected by BMJ, all presenting jaundice, and 14 gender- and age-matched healthy controls. Stool samples were collected from the subjects and shotgun metagenomic sequencing was performed, the purpose of which was to characterize the composition of the gut microbiota of patients with NC
PT
and BMJ and to search for potential gut microbiota markers associated with NC and
SC
RI
jaundice.
NU
Methods
MA
Participants
The participants were recruited from Shenzhen Children’s Hospital (Shenzhen, China).
PT E
D
The diagnostic criteria used for neonatal cholestasis (NC) were as prescribed in the Guideline for the Evaluation of Cholestatic Jaundice in Infants from the North
CE
American Society for Pediatric Gastroenterology, Hepatology and Nutrition [9, 10]. The inclusion and exclusion criteria for participants are detailed in the
AC
Supplementary Appendix. Written informed consent was provided by each child’s parents. The protocol of this study was in accordance with the Declaration of Helsinki and was approved by the Human Ethics Committee of Shenzhen Children’s Hospital. The stool samples were collected when infants were exhibiting jaundice.
ACCEPTED MANUSCRIPT Shotgun metagenomic sequencing
Shotgun metagenomic sequencing was performed in accordance with the protocol used in a previous study [11]. In short, fecal samples were collected and stored at
PT
−80°C prior to DNA extraction. DNA was extracted from fecal samples using a StoolGen DNA kit (CWBiotech Co., Beijing, China). Shotgun metagenomic libraries
RI
were constructed with a TruSeq DNA Sample Preparation kit (Illumina, San Diego,
SC
CA, USA). The libraries (insert size 200–500 bp) were sequenced using an Illumina
NU
Hiseq2500 sequencer (Illumina, San Diego, CA, USA) for 2G raw data (13 million reads with 150 base pairs) output for each sample. The gut microbiota composition
D
MA
was obtained using MEGAN5 [12] and is detailed in the Supplementary Appendix.
PT E
Impact of clinical indices on gut microbiota composition
CE
To evaluate the clinical indices that may impact gut microbiota composition, a Permutational Multivariate Analysis of Variance (PERMANOVA) using R’s Adonis
AC
function was also performed. We then compared gut microbiota diversity between subgroups with/without the studied clinical indices. In short, the ecological diversity of samples was computed using VEGAN [13] with a taxonomic profiling dataset at the species level. For alpha diversity-based comparison, both inverse Simpson’s and Shannon’s diversity indices were used, a boxplot was drawn, and a P value <0.05 in Wilcoxon’s rank sum test was considered significant. For beta diversity-based comparison, Bray–Curtis distance was used, and both nonmetric multidimensional
ACCEPTED MANUSCRIPT scaling (NMDS) and principal component analysis (PCA) of sample beta diversity were used. Furthermore, sparse partial least squares discriminant analysis (sPLS-DA) was also performed to determine whether microbiota species profiling can distinguish
PT
the NC patients from healthy controls.
RI
Identifying significantly enriched species and KOs and association with clinical
SC
indices
NU
Both taxonomic profiling datasets at the species level and functional profiling datasets
MA
at the KEGG (Kyoto Encyclopedia of Genes and Genomes [14]) orthology (KO)-level were used for metagenome-wide association study (MWAS). Species-level and
D
KO-level datasets were normalized to 1 million aligned reads. MWAS of NC patients
PT E
and non-NC controls was performed using DEseq2 [15]. Significantly enriched species were selected using the following thresholds: false discovery rate
CE
(FDR)-adjusted P value ≤0.05 (Wilcoxon’s rank sum test), |log2foldchange| ≥1 according to DEseq2, and mean relative abundance ≥1×10−5. Significantly enriched
AC
KOs were selected according to the following thresholds: FDR-adjusted P value ≤0.05 (Wilcoxon’s rank sum test) and |log2foldchange| ≥1, according to DEseq2. To determine whether the candidate species and KOs were associated with clinical indices, we performed a correlation analysis using R cor test and linear regression analysis using R lm function.
ACCEPTED MANUSCRIPT Identifying species with potential to act as biomarkers of jaundice
To identify species with the potential for use as disease biomarkers, the 13 species which were significantly increased or decreased in jaundice patients (BMJ and NC) as
PT
compared to healthy controls (CON) without jaundice were used to train a support vector machine (SVM) model using the e1071 function in R package, the optimal
RI
training species were validated using the pROC (Receiver Operating Characteristic)
NU
SC
function in R package, and the area under the curve was calculated.
MA
Results
Shotgun metagenomic sequencing performance
PT E
D
The flowchart of the study design is shown in Figure 1. A total of 70 samples were included in this study, as shown in Table 1 (see Table S1 for details). Forty NC
CE
patients were recruited, including 15 NC patients who were breastfeeding and 25 who were not. All NC patients presented jaundice. Thirty non-NC controls were also
AC
recruited, including 16 breast milk jaundice (BMJ) controls presenting jaundice and 14 healthy controls (CON) not presenting it. All non-NC controls were breastfeeding. Shotgun metagenomic sequencing yielded an average of 2.4 Gb of clean data (16 million 150-bp reads) for each sample. For taxonomic level profiling, a total of 33 phyla, 54 classes, 115 orders, 231 families, 573 genera, and 3300 species were identified for functional level profiling. Seven KEGG level 1 pathways, 40 KEGG level 2 pathways, 279 KEGG level 3 pathways, and 5453 KEGG level 4 pathways or
ACCEPTED MANUSCRIPT functional KEGG Orthologies (KOs) were identified.
Impact of clinical indices on gut microbiota composition To understand the clinical indices that may affect gut microbiota composition, we
PT
performed PERMANOVA. The results (Supplementary Table 2) showed that NC
RI
had the most significant effect on gut microbiota composition (P = 0.0001), indicating
SC
that the disease itself affected the composition of the gut microbiota. PCA results showed that, although it is difficult to distinguish between BMJ and CON, the NC and
NU
non-NC groups (BMJ and CON) can be distinguished very clearly based on the gut
MA
microbiota composition (Figure 2a), which is consistent with the results of the NMDS and PCoA (Supplementary Figure 1B). At the same time, we found that
D
breastfeeding also had a significant effect on the gut microbiota composition (P =
PT E
0.0002), indicating that breastfeeding is also an important variable in this study. The sPLS-DA analysis of gut microbiota composition showed that the presence or absence
CE
of breastfeeding can clearly distinguish the samples (Supplementary Figure 2). We
AC
also found that some important clinical indices, such as total bilirubin (TBIL), and bilirubin and total bilirubin ratios (DBIL/TBIL), also significantly influenced gut microbiota composition (P values of 0.0449 and 0.0002, respectively). Other clinical phenotypes, such as age and gender, had no significant effect on gut microbiota composition (P value > 0.05). Notably, there was also a tendency for the mode of delivery to affect the gut microbiota composition (P = 0.0799). sPLS-DA analysis also showed a trend of the samples being clearly divided into two groups based on the
ACCEPTED MANUSCRIPT mode of delivery (Supplementary Figure 2).
Gut microbiota species richness significantly increased in NC patients We compared the gut microbiota diversity among NC patients, BMJ, and CON, and
PT
found no significant differences in Shannon’s diversity and inverse Simpson’s
RI
diversity between NC patients and non-NC controls (Supplementary Figure 1A).
SC
However, there was a significant difference in gut microbiota species richness between NC patients and non-NC controls. Specifically, the Chao 1 abundance index
NU
of the NC group was significantly higher than those of the BMJ and CON groups
MA
(Figure 2b). To gain more insight into the specific types of bacteria that increased significantly in NC patients, we compared the overall abundance of high-level
D
bacteria between NC patients and non-NC controls and found that the NC patients had
PT E
a decreased abundance of the phylum Actinobacteria and an increase of the phylum Bacteroidetes compared with the non-NC controls; these findings were independent of
CE
the breastfeeding status (Figure 3). These results may also mean that there was a
AC
significant difference in gut microbiota composition between the NC patients and the non-NC controls.
MWAS revealed NC-related species or functional genes A total of 132 species were shown to be associated with NC. Among these, 19 species were significantly decreased in NC patients; more than 50% (10/19) of these were Bifidobacterium species, including nine Bifidobacterium longum strains and one
ACCEPTED MANUSCRIPT Bifidobacterium bifidum species. A total of 113 species were significantly increased in NC patients (Supplementary Table 3), including five species of Streptococcus, three species of Mycobacterium, two strains of Rhodopseudomonas palustris, two strains of Veillonella parvula, and two strains of Niastella koreensis. In addition, a total of 81
PT
KOs were associated with NC (Supplementary Table 4). Seven of them were
RI
significantly decreased in NC patients, with most (5/7) being related to energy
SC
metabolism, including K02003, K15598, aldehyde dehydrogenase (NAD+) [EC (Enzyme Commission number): 1.2.1.3], amylosucrase [EC: 2.4.1.4], and
NU
rhamnosyltransferase [EC: 2.4.1.-]. The other 74 Kos were significantly increased in
MA
NC patients, including those associated with pathogen invasion, such as pilus assembly protein CpaC, pilus assembly protein FimV, spore germination protein D,
D
spore germination protein KC, type III secretion protein SctC, type IV pilus assembly
PT E
protein PilQ, and type I polyketide synthase AVES.
CE
MWAS revealed jaundice-associated species or functional gene markers
AC
Considering that jaundice was found in both BMJ and NC groups, the gut microbiota of the BMJ and NC groups was compared with that of the CON group. Thirteen species were associated with jaundice (Supplementary Table 5), nine of which showed a significant decrease in the jaundice group (BMJ and NC), with most (6/9) being members of the genus Bifidobacterium, including four Bifidobacterium longum strains (Figure 2c and Supplementary Figure 3A) and two Bifidobacterium breve strains. We also analyzed the higher taxa of these Bifidobacterium species, such as the
ACCEPTED MANUSCRIPT genus Bifidobacterium, the family Bifidobacteriaceae, and the phylum Actinobacteria, and found that these taxa were also significantly decreased in the jaundice groups (BMJ and NC) (Supplementary Figure 3B). Other species also exhibited significant decreases in the jaundice group, including Citrobacter koseri ATCC BAA-895,
PT
Mobiluncus curtisii ATCC 43063, and Parabacteroides distasonis; four species
RI
showed significant decreases in the jaundice group (BMJ and NC), including two
SC
Clostridium botulinum strains, Providencia stuartii MRSN 2154 and Streptococcus thermophilus. In addition, a total of 37 KOs were associated with jaundice
NU
(Supplementary Table 6), most of which (36/37) were significantly decreased in the
pathway genes,
MA
jaundice groups (NC and BMJ). Of these, nine were glucose metabolism-related namely, 6-phosphogluconate
dehydrogenase [EC: 1.1.1.44]
D
glucokinase [EC: 2.7.1.2], glucose-6-phosphate 1-dehydrogenase [EC: 1.1.1.49],
PT E
glucose-6-phosphate 1-epimerase [EC: 5.1.3.15], phosphoglycerate mutase [EC: 5.4.2.1], thiamine pyrophosphokinase [EC: 2.7.6.2], UDP-galactopyranose mutase
CE
[EC: 5.4.99.9], and UDP-glucose 4-epimerase [EC: 5.1.3.2]. This implies that the
AC
jaundice group may have weaker microbial glucose metabolism than the control group. Acetylornithine/N-succinyl diaminopimelate aminotransferase [EC: 2.6.1.11, 2.6.1.17] showed a significant increase in the jaundice groups (NC and BMJ) compared with the level in CON.
Significantly enriched species/KOs are associated with clinical indices To further screen for gut microbiota associated with clinical indices, we performed a
ACCEPTED MANUSCRIPT regression analysis of NC- and jaundice-related species and KOs with important clinical indices and found that a total of 16 species were associated with DBIL (Figure 4a). Nine of these were Bifidobacterium, namely, eight Bifidobacterium longum strains and one Bifidobacterium bifidum BGN4 (Figure 2c and
PT
Supplementary Figure 3a). A total of 19 KOs were also associated with DBIL
RI
(Figure 4b), of which 15 and 4 were negatively and positively correlated with it,
SC
respectively. Of the KOs negatively correlated with DBIL, six were glucose metabolism-related genes, including phosphoglycerate mutase [EC: 5.4.2.1], UDP 4
epimerase
5.1.3.2],
1-dehydrogenase
β-galactosidase
[EC:
1.1.1.49],
[EC:
3.2.1.23],
phosphogluconate
MA
glucose-6-phosphate
[EC:
NU
glucose
dehydrogenase [EC: 1.1.1.44], UDP galactopyranose mutase [EC: 5.4.99.9], and
D
glucokinase [EC: 2.7.1.2]. Additionally, cell filamentation protein, a possible bacterial
PT E
virulence factor-related protein, was positively correlated with DBIL. Further, some species and KOs were found to be correlated with DBIL/TBIL and ALT
AC
CE
(Supplementary Figure 4).
Significantly enriched species are potential biomarkers of jaundice In order to further determine the species that can be used as biomarkers of NC patients, the 13 species with significant differences in abundance between jaundice patients (BMJ and NC) and healthy controls (CON) without jaundice were used to train the SVM model. The results showed that a clear distinction between jaundice patients and non-jaundice controls could be achieved based on 10 bacteria species
ACCEPTED MANUSCRIPT (Supplementary Table 7), with an area under ROC curve (AUC) value of 0.91 (Figure 5), meaning that these 10 species are effective biomarkers of jaundice.
Discussion
PT
Neonatal cholestasis can cause a shortage of bile acids in extrahepatic regions,
RI
especially in the gut. This deficiency in the gut may promote the translocation of
SC
bacteria, as bile acids can not only inhibit microbes from entering enterocytes [16], but also directly inhibit the proliferation of pathogenic microbes [17]. In accordance
NU
with this, in the present study, we found that cholestasis was accompanied by a
MA
significant increase in gut microbiota richness. We also found that a total of 113 species were upregulated in NC patients. In particular, we found that potentially
D
pathogenic species, such as five species of Streptococcus, three species of
PT E
Mycobacterium, two strains of Rhodopseudomonas palustris, and two strains of Veillonella parvula, were upregulated in NC patients. Studies have shown that
CE
pathogenic microbes can kill commensal bacteria via virulence factors such as type VI
AC
secretion system (T6SS), whereas commensal bacteria can inhibit T6SS-mediated killing of commensal bacteria by modifying bile acids. A lack of intestinal bile acids will lead to disequilibrium in this regard, exacerbating the reduction of beneficial bacteria [18]. In agreement with this speculation, we found that potentially pathogenic species as well as gut microbiota functional genes that were linked to virulence factors, such as pilus assembly protein CpaC, pilus assembly protein FimV, spore germination protein D, spore germination protein KC, type III secretion protein SctC,
ACCEPTED MANUSCRIPT type IV pilus assembly protein PilQ, and type I polyketide synthase AVES, were significantly upregulated in NC patients. In particular, type III secretion protein SctC was positively correlated with serum bilirubin levels. We also found that a large number of commensal bacteria, such as Bifidobacterium and Bifidobacteriaceae, were
PT
significantly downregulated in NC patients, and the decrease of Bifidobacterium is
RI
associated with an increase in serum bilirubin levels. In summary, we conclude that
SC
the occurrence of NC is closely related to the large-scale growth of potentially pathogenic bacteria and the large reduction in beneficial bacteria such as
NU
Bifidobacterium.
MA
The occurrence of neonatal cholestasis is also accompanied by jaundice. The main manifestation of jaundice is an increase in serum bilirubin. It is well known that
D
bilirubin formed by the destruction of erythrocytes first enters the liver and then
PT E
combines with UDP-glucuronic acid (UDPGA) to form water-soluble conjugated bilirubin. Subsequently, bilirubin is secreted into the biliary tract and enters the
CE
intestine for excretion [19]. Gut microbiota is important in this context as it may play
AC
a major role in bilirubin efflux. Using the germ-free multidrug resistance 2 knockout mouse model, Tabibian and coworkers found that germ-free mice showed upregulated serum bilirubin levels [20], which provides evidence that the gut microbiota may participate in bilirubin metabolism. Moreover, Tuzun et al. indicated that downregulation of members of the Bifidobacterium genus, such as B. adolescentis, B. bifidum, and B. longum, was associated with an increase in serum bilirubin level [21]. This is consistent with our results, as we also found that a decrease in members of the
ACCEPTED MANUSCRIPT Bifidobacterium genus was associated with an increase in serum direct bilirubin level. Based on previous research and our present findings, we speculated that the decrease in Bifidobacterium may have been directly or indirectly caused by the increase in direct bilirubin. We conducted a comparative analysis of gut microflora functional
PT
levels between jaundice (NC and BMJ) groups and the group without jaundice (CON)
RI
as well as an abundance of galactose metabolism-related genes downregulated in the
SC
jaundice groups. Bifidobacterium is well known to be directly involved in the utilization of galactooligosaccharides (GOSs) [22], and transforming GOSs to
NU
galactose and UDP-glucose through the galactose metabolism pathway, while
MA
UDP-glucose, a product of galactose metabolism, is a direct precursor to UDPGA. Combined with the previous studies, it may be conjectured that Bifidobacterium is
D
involved in the metabolism of bilirubin by the galactose metabolism pathway. The
PT E
decrease in Bifidobacterium, coupled with the decreased abundance of galactose metabolism pathway-related genes, leads to a shortage of intestinal UDP-glucose and
CE
affecting the downstream glucuronic acid pathway to form conjugated bilirubin. At
AC
the same time, we also identified a decreased gut microbiota abundance of glucokinase-generated species; this may have exacerbated the decrease in UDP-glucose in the pentose phosphate pathway. Finally, we found a decrease in the abundance of phosphogluconate dehydrogenase (PGM)-associated gut microbiota, which may indicate abnormalities in the ATP-related glycolysis pathway and affect the glucuronic acid pathway that requires ATP (Supplementary Figure 5). Breastfeeding is considered to be a risk factor for BMJ, however, whether
ACCEPTED MANUSCRIPT breastfeeding plays a role in the pathogenesis of NC has not been well studied. In the present study, we found that the breast-fed and non-breast-fed NC patients had similar microbiota profiles at the phylum level; this may imply that the bile acid pool and enterohepatic circulation form a major determinant, or one that has greater influence
PT
than diet, of gut microbiota composition.
RI
There are some strengths in the present study. The first strength is the study
SC
design, which included three groups and the categorization of cases with/without NC and with/without jaundice to identify gut microbiota composition associated with NC
NU
and jaundice, respectively. Second, in contrast to the conventional 16S rDNA
MA
sequencing, the shotgun metagenomic sequencing method applied here enabled us to study the gut microbiota composition at the functional level. Finally, the analysis of
D
the correlation between the selected differential metagenomic components and
PT E
important clinical phenotypes was also helpful in determining the gut microbiota composition associated with NC and jaundice. However, there are also some
CE
limitations to the present study. Firstly, the sample size was not large enough.
AC
Secondly, there was a lack of assessment of some of the variables affecting the gut microbiota composition, such as antibiotic treatment, length of hospitalization, etc. Considering that there were no cases of jaundice in the healthy control group, we did not collect clinical indices such as serum bilirubin level for this group. Additionally, an independent validation cohort would be necessary to further evaluate whether Bifidobacterium is a biomarker of jaundice, and whether Bifidobacterium supplementation can mitigate the symptom of jaundice. Although it is hard to see an
ACCEPTED MANUSCRIPT immediate path to clinical use when fractionated bilirubin also accomplishes the same purpose quickly and at a significantly lower cost, the present study is of great significance in understanding the mechanisms of the occurrence of jaundice and in developing corresponding treatments. The search for markers related to pathological
PT
jaundice is of great significance, especially prior to the marked elevation of direct
SC
RI
bilirubin levels; further study in this regard needs to be carried out in the next step.
Conclusion
NU
In conclusion, the alteration of gut microbiota in NC patients mainly manifested as a
MA
significant increase in species richness. The main manifestation in jaundice patients is a significant decrease in Bifidobacterium, which may be involved in the metabolism
D
of bilirubin through the galactose metabolism pathway. The results of this study
PT E
suggest an imbalance in the gut microbiota of NC patients, which is associated with a substantial reduction in Bifidobacterium, suggesting the possibility of treating
AC
CE
jaundice through probiotic supplementation.
Acknowledgments We are sincerely grateful to the individuals who participated in this study.
Funding/Support: This project was supported by the National Natural Science Foundation of China (Program Nos. 81701351 and 81720108018) and Shenzhen Science
Technology
and
Innovation
Commission
(Program
Nos.
ACCEPTED MANUSCRIPT JCYJ20160429174706491 and JCYJ20170413093358429), the National Key Research and Development Program of China (program No. 2016YFC0905100 and 2017YFA0104204).
1.
PT
References Justinich, C.J. and J.S. Hyams, Neonatal Jaundice. 1998: Humana Press.
RI
293-300.
Mccandless, D.W., Breast Milk Jaundice. 2011: Humana Press. 115-120.
3.
Balistreri, W.F., Neonatal cholestasis. J Pediatr, 1985. 106(2): p. 171-84.
4.
Gotze, T., et al., Neonatal Cholestasis - Differential Diagnoses, Current
NU
SC
2.
Diagnostic Procedures, and Treatment. Front Pediatr, 2015. 3: p. 43. Hoerning, A., et al., Diversity of disorders causing neonatal cholestasis - the
MA
5.
experience of a tertiary pediatric center in Germany. Front Pediatr, 2014. 2: p.
Mueller, N.T., et al., The infant microbiome development: mom matters.
PT E
6.
D
65.
Trends Mol Med, 2015. 21(2): p. 109-17. 7.
Poland, R.L. and G.B. Odell, Physiologic jaundice: the enterohepatic
8.
CE
circulation of bilirubin. N Engl J Med, 1971. 284(1): p. 1-6. Visschers, R.G., et al., The gut-liver axis. Curr Opin Clin Nutr Metab Care,
9.
AC
2013. 16(5): p. 576-81. Fawaz, R., et al., Guideline for the Evaluation of Cholestatic Jaundice in Infants: Joint Recommendations of the North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition and the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition. J Pediatr Gastroenterol Nutr, 2017. 64(1): p. 154-168. 10.
Moyer, V., et al., Guideline for the evaluation of cholestatic jaundice in infants: recommendations
of
the
North
American
Society
for
Pediatric
Gastroenterology, Hepatology and Nutrition. J Pediatr Gastroenterol Nutr,
ACCEPTED MANUSCRIPT 2004. 39(2): p. 115-28. 11.
Zhou, S., et al., Diversity of Gut Microbiota Metabolic Pathways in 10 Pairs of Chinese Infant Twins. PLoS One, 2016. 11(9): p. e0161627.
12.
Huson, D.H., et al., MEGAN analysis of metagenomic data. Genome Res, 2007. 17(3): p. 377-86. Dixon, P., VEGAN, a package of R functions for community ecology. Journal
PT
13.
of Vegetation Science, 2003. 14(6): p. 927-930.
Kanehisa, M. and S. Goto, KEGG: kyoto encyclopedia of genes and genomes.
RI
14.
Nucleic Acids Res, 2000. 28(1): p. 27-30.
Love, M.I., W. Huber, and S. Anders, Moderated estimation of fold change
SC
15.
and dispersion for RNA-seq data with DESeq2. Genome Biol, 2014. 15(12): p.
16.
NU
550.
Wells, C.L., R.P. Jechorek, and S.L. Erlandsen, Inhibitory effect of bile on
MA
bacterial invasion of enterocytes: possible mechanism for increased translocation associated with obstructive jaundice. Crit Care Med, 1995. 23(2):
Buffie, C.G., et al., Precision microbiome reconstitution restores bile acid
PT E
17.
D
p. 301-7.
mediated resistance to Clostridium difficile. Nature, 2015. 517(7533): p. 205-8.
Bachmann, V., et al., Bile Salts Modulate the Mucin-Activated Type VI
CE
18.
Secretion System of Pandemic Vibrio cholerae. PLoS Negl Trop Dis, 2015.
19.
AC
9(8): p. e0004031. Billing, B.H., Twenty-five years of progress in bilirubin metabolism (1952-77). Gut, 1978. 19(6): p. 481-91. 20.
Tabibian, J.H., et al., Absence of the intestinal microbiota exacerbates hepatobiliary disease in a murine model of primary sclerosing cholangitis. Hepatology, 2016. 63(1): p. 185-96.
21.
Tuzun, F., et al., Breast milk jaundice: effect of bacteria present in breast milk and infant feces. J Pediatr Gastroenterol Nutr, 2013. 56(3): p. 328-32.
22.
Thongaram, T., et al., Prebiotic Galactooligosaccharide Metabolism by
ACCEPTED MANUSCRIPT Probiotic Lactobacilli and Bifidobacteria. J Agric Food Chem, 2017. 65(20): p.
AC
CE
PT E
D
MA
NU
SC
RI
PT
4184-4192.
ACCEPTED MANUSCRIPT Tables and legends
CON
BMJ
(Non-NC)
(Non-NC)
Participants(n=)
14
16
Gender(F/M)
3/11
2/14
Age(year)(mean±SD)
2.8±0.89
1.1±0.25
2.2±1.2
BF(n=;Yes/no)
14/0
16/0
15/25
VD(n=;Yes/no)
12/2
RI
Table 1. Characteristics of the Cohort
8/8
31/9
ALT(IU/l)(mean±SD)
NA
25±17
130±120
TBIL(umol/l)(mean±SD)
NA
210±66
170±78
DBIL(umol/l)(mean±SD)
NA
12±5.8
81±39
DBIL/TBIL(%)(mean±SD)
NA
6±3.6
50±11
Clinical indices
NC
PT
SC
NU
MA
40 21/19
Note: ALT, alanine transaminase, BF, breastfeeding, BMJ, breast milk jaundice; CON,
D
control; DBIL, direct bilirubin; F, female; M, male; NC, neonatal cholestasis; non-NC,
PT E
non-neonatal cholestasis; TBIL, total bilirubin; VD, vaginal delivery Figures and legends
CE
Figure 1. Flowchart of study design. ALT, alanine transaminase; BF, breastfeeding; BMJ, breast milk jaundice; CON, control; DBIL, direct bilirubin; FDR, false
AC
discovery rate; KEGG, Kyoto Encyclopedia of Genes and Genomes; MEGAN5, Metagenome Analyzer version 5; NC, neonatal cholestasis; NCBI, National Center for Biotechnology Information; non-NC, non-neonatal cholestasis; KO, KEGG Orthology; TBIL, total bilirubin
Figure 2. Altered gut microbiota composition in neonatal cholestasis patients. a, Neonatal cholestasis (NC) patients can be clearly separated from breast milk jaundice (BMJ) cases and healthy controls (CON) without jaundice according to gut
ACCEPTED MANUSCRIPT microbiota functional level composition. b, Increased species richness in gut microbiota of NC patients compared with that in BMJ and CON. c, Bifidobacterium longum significant decreased in NC. *, P value <0.05; **, P value <0.01, Wilcoxon’s rank sum test
PT
Figure 3. Impact of breastfeeding on gut microbiota of NC at phylum level. a, Comparison of high level taxa between subgroups CON (healthy controls), BMJ,
RI
NC_BF (NC with breastfeeding), NC_nonBF (NC without breastfeeding); b, comparison of phylum Actinobacteria between subgroups CON, BMJ, NC_BF and
SC
NC_nonBF; c, comparison of phylum Bacteroide between subgroups CON, BMJ, NC_BF and NC_nonBF; *P value<0.05; **, P value<0.01; ***, P value<0.001,
NU
Wilcoxon rank sum test; Bacteroidetes significantly increased in NC patients, and Actinobacteria significantly decreased in NC patients, independently of breastfeeding
MA
(BF) impact.
D
Figure 4. Decrease of Bifidobacterium longum was associated with jaundice. a,
PT E
Bifidobacterium longum was negatively associated with serum bilirubin levels. b, Abundance of metabolism pathway-related genes in gut microbiota was negatively
CE
associated with serum bilirubin level. Figure 5. Receiver operating characteristic curve analysis results. With 10 species
AC
identified by SVM model training, an AUC value of 0.91 can be achieved to distinguish between jaundice patients (BMJ and NC) and non-jaundice (CON) controls.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Shaoming Zhou, Zhangxing Wang, Fusheng He, Huixian Qiu, Yan Wang, Huihui Wang, Jianli Zhou, Jiaxiu Zhou, Guoqiang Cheng, Wenhao Zhou, Ruihuan Xu, Mingbang Wang PII: DOI: Reference:
S0955-2863(18)30264-X doi:10.1016/j.jnutbio.2018.09.016 JNB 8065
To appear in:
The Journal of Nutritional Biochemistry
Received date: Revised date: Accepted date:
29 March 2018 10 September 2018 12 September 2018
Please cite this article as: Shaoming Zhou, Zhangxing Wang, Fusheng He, Huixian Qiu, Yan Wang, Huihui Wang, Jianli Zhou, Jiaxiu Zhou, Guoqiang Cheng, Wenhao Zhou, Ruihuan Xu, Mingbang Wang , Association of serum bilirubin in newborns affected by jaundice with gut microbiota dysbiosis. Jnb (2018), doi:10.1016/j.jnutbio.2018.09.016
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Association of serum bilirubin in newborns affected by jaundice with gut microbiota dysbiosis Shaoming Zhou1*#; Zhangxing Wang2*; Fusheng He3*; Huixian Qiu4*; Yan Wang3; Huihui Wang5; Jianli Zhou1; Jiaxiu Zhou6; Guoqiang Cheng7; Wenhao Zhou7; Ruihuan Xu8#; Mingbang Wang9#;
PT E
D
MA
NU
SC
RI
PT
1. Division of Gastroenterology, Shenzhen Children’s Hospital, Shenzhen, Guangdong, 518038, China; 2. Division of Neonatology, Shenzhen Longhua People's Hospital, Guangdong, 518109, China; 3. Imunobio, Shenzhen, Guangdong, China, 518001; 4. Division of Neonatology, Longgang Central Hospital of Shenzhen, Guangdong, 518116, China; 5. Division of clinical nutrition, Shenzhen Children’s Hospital, Shenzhen, Guangdong, 518038, China; 6. Division of psychology, Shenzhen Children’s Hospital, Shenzhen, Guangdong, 518038, China; 7. Division of Neonatology, Children’s Hospital of Fudan University, National Center for Children's Health, Shanghai, 201102, China; 8. Clinical Laboratory, Longgang Central Hospital of Shenzhen, Guangdong, 518116, China; 9. Xiamen branch, Shanghai Key Laboratory of Birth Defects, Division of Neonatology, Children’s Hospital of Fudan University, National Center for Children's Health, Shanghai, 201102, China;
CE
* Contributed equally.
#Corresponding Author: Dr. Mingbang Wang ([email protected]), Dr.
AC
Ruihuan Xu ([email protected]) or Dr. Shaoming Zhou ([email protected]);
Author Contributions: All authors contributed to conception and design of the project, including collection of samples, clinical evaluation, sequencing and data analysis. M.W., R.X and S.Z. drafted and revised the manuscript. All authors approved the final version of the manuscript. Conflict of Interest Disclosures：There is no conflict of interest to declare.
ACCEPTED MANUSCRIPT BACKGROUND & AIMS Breast milk jaundice (BMJ) is common and benign, but neonatal cholestasis (NC) is rare and not benign, so early differentiation between NC and non-NC jaundice is important and may facilitate diagnosis and treatment. Gut microbiota plays an important role in enterohepatic circulation, which in turn plays an
PT
important role in the secretion of bilirubin. We aimed to determine the composition of
RI
gut microbiota in patients with NC and BMJ, and to identify the gut microbiota
SC
composition associated with NC and BMJ.
NU
METHODS Data on age, gender, delivery, feeding mode, serum total bilirubin, direct bilirubin, and liver function were collected for NC patients, BMJ patients and healthy
MA
controls ， respectively. Shotgun metagenomic sequencing and metagenome-wide
D
association were performed.
PT E
RESULTS Forty NC patients, 16 patients affected by BMJ, and 14 healthy controls (CON) without jaundice were enrolled. A significant increase in species richness,
CE
especially Bacteroides, was found in NC patients. The abundances of potentially pathogenic species and KEGG orthologies (KOs) of virulence factor genes were
AC
positively correlated with serum bilirubin level. The abundances of nine species of Bifidobacterium and three KOs of galactose metabolism were significantly decreased in the jaundice group (NC and BMJ) and were negatively correlated with serum bilirubin level.
CONCLUSIONS The gut microbiota in NC patients is characterized by a significant increase in species richness, possibly due to the proliferation of potentially pathogenic
ACCEPTED MANUSCRIPT species.
Additionally, the gut microbiota in jaundice patients is characterized by a decreased abundance of Bifidobacterium. Decreased Bifidobacterium has been associated with
PT
elevated bilirubin and abnormal gut microbiota galactose metabolic pathway. Further,
RI
ten bacteria species were identified as potential biomarker of jaundice.
SC
Keywords
MA
NU
Neonatal cholestasis; breast milk jaundice; gut microbiota; shotgun metagenomics
Key Points
D
Question Is there any alteration of gut microbiotain neonatal cholestasis patients?
PT E
Does gut microbiota have any involvement in the occurrence of neonatal cholestasis
CE
or breast milk jaundice?
Findings The alteration of gut microbiota in neonatal cholestasis patients mainly
AC
manifested as a significant increase in species richness and an increased abundance of potentially pathogenic species, while the main manifestation in jaundice patients was a significant decrease in Bifidobacterium which may be involved in the metabolism of bilirubin through the galactose metabolic pathway.
Meaning The results suggest that an imbalance of gut microbiota exist in neonatal cholestasis and breast milk jaundice patients, primarily in the form of a substantial
ACCEPTED MANUSCRIPT reduction in the abundance of Bifidobacterium, suggesting the possibility of intervention treatment for neonatal cholestasis and breast milk jaundice by
PT
supplementing probiotics.
RI
Introduction
SC
Physiological jaundice and breast milk jaundice (BMJ) commonly occur in the neonatal period due to an increase in serum unconjugated bilirubin, with most cases
NU
being benign [1, 2]. However, neonatal cholestasis (NC) or cholestatic jaundice
MA
possibly affects 1/2500 newborns, characterized by an increase in serum conjugated bilirubin or direct bilirubin. It is important to differentiate NC from noncholestatic
D
jaundice because NC is more likely to involve a severe condition such as abnormal
PT E
liver function and require prompt diagnosis and treatment [3, 4]. The most commonly known causes of NC are biliary atresia and neonatal hepatitis, followed by preterm
CE
birth and inborn errors of metabolism [5]. The pathogenesis of NC involves the liver failing to adequately secrete bilirubin from the blood into the bile [5]. Given that the
AC
gut microbiota plays an important role in human health, its disruption in early infancy may have a significant impact on the maturation of the immune system and on the health status in adulthood [6]; enterohepatic circulation also plays an important role in the excretion of bilirubin [7]. Moreover, the gut microbiota plays an important role in liver disease through the gut–liver axis [8]. Therefore, it is important to understand the composition of the gut microbiota in patients with NC and BMJ. In the present
ACCEPTED MANUSCRIPT study, we recruited 40 NC patients and 16 affected by BMJ, all presenting jaundice, and 14 gender- and age-matched healthy controls. Stool samples were collected from the subjects and shotgun metagenomic sequencing was performed, the purpose of which was to characterize the composition of the gut microbiota of patients with NC
PT
and BMJ and to search for potential gut microbiota markers associated with NC and
SC
RI
jaundice.
NU
Methods
MA
Participants
The participants were recruited from Shenzhen Children’s Hospital (Shenzhen, China).
PT E
D
The diagnostic criteria used for neonatal cholestasis (NC) were as prescribed in the Guideline for the Evaluation of Cholestatic Jaundice in Infants from the North
CE
American Society for Pediatric Gastroenterology, Hepatology and Nutrition [9, 10]. The inclusion and exclusion criteria for participants are detailed in the
AC
Supplementary Appendix. Written informed consent was provided by each child’s parents. The protocol of this study was in accordance with the Declaration of Helsinki and was approved by the Human Ethics Committee of Shenzhen Children’s Hospital. The stool samples were collected when infants were exhibiting jaundice.
ACCEPTED MANUSCRIPT Shotgun metagenomic sequencing
Shotgun metagenomic sequencing was performed in accordance with the protocol used in a previous study [11]. In short, fecal samples were collected and stored at
PT
−80°C prior to DNA extraction. DNA was extracted from fecal samples using a StoolGen DNA kit (CWBiotech Co., Beijing, China). Shotgun metagenomic libraries
RI
were constructed with a TruSeq DNA Sample Preparation kit (Illumina, San Diego,
SC
CA, USA). The libraries (insert size 200–500 bp) were sequenced using an Illumina
NU
Hiseq2500 sequencer (Illumina, San Diego, CA, USA) for 2G raw data (13 million reads with 150 base pairs) output for each sample. The gut microbiota composition
D
MA
was obtained using MEGAN5 [12] and is detailed in the Supplementary Appendix.
PT E
Impact of clinical indices on gut microbiota composition
CE
To evaluate the clinical indices that may impact gut microbiota composition, a Permutational Multivariate Analysis of Variance (PERMANOVA) using R’s Adonis
AC
function was also performed. We then compared gut microbiota diversity between subgroups with/without the studied clinical indices. In short, the ecological diversity of samples was computed using VEGAN [13] with a taxonomic profiling dataset at the species level. For alpha diversity-based comparison, both inverse Simpson’s and Shannon’s diversity indices were used, a boxplot was drawn, and a P value <0.05 in Wilcoxon’s rank sum test was considered significant. For beta diversity-based comparison, Bray–Curtis distance was used, and both nonmetric multidimensional
ACCEPTED MANUSCRIPT scaling (NMDS) and principal component analysis (PCA) of sample beta diversity were used. Furthermore, sparse partial least squares discriminant analysis (sPLS-DA) was also performed to determine whether microbiota species profiling can distinguish
PT
the NC patients from healthy controls.
RI
Identifying significantly enriched species and KOs and association with clinical
SC
indices
NU
Both taxonomic profiling datasets at the species level and functional profiling datasets
MA
at the KEGG (Kyoto Encyclopedia of Genes and Genomes [14]) orthology (KO)-level were used for metagenome-wide association study (MWAS). Species-level and
D
KO-level datasets were normalized to 1 million aligned reads. MWAS of NC patients
PT E
and non-NC controls was performed using DEseq2 [15]. Significantly enriched species were selected using the following thresholds: false discovery rate
CE
(FDR)-adjusted P value ≤0.05 (Wilcoxon’s rank sum test), |log2foldchange| ≥1 according to DEseq2, and mean relative abundance ≥1×10−5. Significantly enriched
AC
KOs were selected according to the following thresholds: FDR-adjusted P value ≤0.05 (Wilcoxon’s rank sum test) and |log2foldchange| ≥1, according to DEseq2. To determine whether the candidate species and KOs were associated with clinical indices, we performed a correlation analysis using R cor test and linear regression analysis using R lm function.
ACCEPTED MANUSCRIPT Identifying species with potential to act as biomarkers of jaundice
To identify species with the potential for use as disease biomarkers, the 13 species which were significantly increased or decreased in jaundice patients (BMJ and NC) as
PT
compared to healthy controls (CON) without jaundice were used to train a support vector machine (SVM) model using the e1071 function in R package, the optimal
RI
training species were validated using the pROC (Receiver Operating Characteristic)
NU
SC
function in R package, and the area under the curve was calculated.
MA
Results
Shotgun metagenomic sequencing performance
PT E
D
The flowchart of the study design is shown in Figure 1. A total of 70 samples were included in this study, as shown in Table 1 (see Table S1 for details). Forty NC
CE
patients were recruited, including 15 NC patients who were breastfeeding and 25 who were not. All NC patients presented jaundice. Thirty non-NC controls were also
AC
recruited, including 16 breast milk jaundice (BMJ) controls presenting jaundice and 14 healthy controls (CON) not presenting it. All non-NC controls were breastfeeding. Shotgun metagenomic sequencing yielded an average of 2.4 Gb of clean data (16 million 150-bp reads) for each sample. For taxonomic level profiling, a total of 33 phyla, 54 classes, 115 orders, 231 families, 573 genera, and 3300 species were identified for functional level profiling. Seven KEGG level 1 pathways, 40 KEGG level 2 pathways, 279 KEGG level 3 pathways, and 5453 KEGG level 4 pathways or
ACCEPTED MANUSCRIPT functional KEGG Orthologies (KOs) were identified.
Impact of clinical indices on gut microbiota composition To understand the clinical indices that may affect gut microbiota composition, we
PT
performed PERMANOVA. The results (Supplementary Table 2) showed that NC
RI
had the most significant effect on gut microbiota composition (P = 0.0001), indicating
SC
that the disease itself affected the composition of the gut microbiota. PCA results showed that, although it is difficult to distinguish between BMJ and CON, the NC and
NU
non-NC groups (BMJ and CON) can be distinguished very clearly based on the gut
MA
microbiota composition (Figure 2a), which is consistent with the results of the NMDS and PCoA (Supplementary Figure 1B). At the same time, we found that
D
breastfeeding also had a significant effect on the gut microbiota composition (P =
PT E
0.0002), indicating that breastfeeding is also an important variable in this study. The sPLS-DA analysis of gut microbiota composition showed that the presence or absence
CE
of breastfeeding can clearly distinguish the samples (Supplementary Figure 2). We
AC
also found that some important clinical indices, such as total bilirubin (TBIL), and bilirubin and total bilirubin ratios (DBIL/TBIL), also significantly influenced gut microbiota composition (P values of 0.0449 and 0.0002, respectively). Other clinical phenotypes, such as age and gender, had no significant effect on gut microbiota composition (P value > 0.05). Notably, there was also a tendency for the mode of delivery to affect the gut microbiota composition (P = 0.0799). sPLS-DA analysis also showed a trend of the samples being clearly divided into two groups based on the
ACCEPTED MANUSCRIPT mode of delivery (Supplementary Figure 2).
Gut microbiota species richness significantly increased in NC patients We compared the gut microbiota diversity among NC patients, BMJ, and CON, and
PT
found no significant differences in Shannon’s diversity and inverse Simpson’s
RI
diversity between NC patients and non-NC controls (Supplementary Figure 1A).
SC
However, there was a significant difference in gut microbiota species richness between NC patients and non-NC controls. Specifically, the Chao 1 abundance index
NU
of the NC group was significantly higher than those of the BMJ and CON groups
MA
(Figure 2b). To gain more insight into the specific types of bacteria that increased significantly in NC patients, we compared the overall abundance of high-level
D
bacteria between NC patients and non-NC controls and found that the NC patients had
PT E
a decreased abundance of the phylum Actinobacteria and an increase of the phylum Bacteroidetes compared with the non-NC controls; these findings were independent of
CE
the breastfeeding status (Figure 3). These results may also mean that there was a
AC
significant difference in gut microbiota composition between the NC patients and the non-NC controls.
MWAS revealed NC-related species or functional genes A total of 132 species were shown to be associated with NC. Among these, 19 species were significantly decreased in NC patients; more than 50% (10/19) of these were Bifidobacterium species, including nine Bifidobacterium longum strains and one
ACCEPTED MANUSCRIPT Bifidobacterium bifidum species. A total of 113 species were significantly increased in NC patients (Supplementary Table 3), including five species of Streptococcus, three species of Mycobacterium, two strains of Rhodopseudomonas palustris, two strains of Veillonella parvula, and two strains of Niastella koreensis. In addition, a total of 81
PT
KOs were associated with NC (Supplementary Table 4). Seven of them were
RI
significantly decreased in NC patients, with most (5/7) being related to energy
SC
metabolism, including K02003, K15598, aldehyde dehydrogenase (NAD+) [EC (Enzyme Commission number): 1.2.1.3], amylosucrase [EC: 2.4.1.4], and
NU
rhamnosyltransferase [EC: 2.4.1.-]. The other 74 Kos were significantly increased in
MA
NC patients, including those associated with pathogen invasion, such as pilus assembly protein CpaC, pilus assembly protein FimV, spore germination protein D,
D
spore germination protein KC, type III secretion protein SctC, type IV pilus assembly
PT E
protein PilQ, and type I polyketide synthase AVES.
CE
MWAS revealed jaundice-associated species or functional gene markers
AC
Considering that jaundice was found in both BMJ and NC groups, the gut microbiota of the BMJ and NC groups was compared with that of the CON group. Thirteen species were associated with jaundice (Supplementary Table 5), nine of which showed a significant decrease in the jaundice group (BMJ and NC), with most (6/9) being members of the genus Bifidobacterium, including four Bifidobacterium longum strains (Figure 2c and Supplementary Figure 3A) and two Bifidobacterium breve strains. We also analyzed the higher taxa of these Bifidobacterium species, such as the
ACCEPTED MANUSCRIPT genus Bifidobacterium, the family Bifidobacteriaceae, and the phylum Actinobacteria, and found that these taxa were also significantly decreased in the jaundice groups (BMJ and NC) (Supplementary Figure 3B). Other species also exhibited significant decreases in the jaundice group, including Citrobacter koseri ATCC BAA-895,
PT
Mobiluncus curtisii ATCC 43063, and Parabacteroides distasonis; four species
RI
showed significant decreases in the jaundice group (BMJ and NC), including two
SC
Clostridium botulinum strains, Providencia stuartii MRSN 2154 and Streptococcus thermophilus. In addition, a total of 37 KOs were associated with jaundice
NU
(Supplementary Table 6), most of which (36/37) were significantly decreased in the
pathway genes,
MA
jaundice groups (NC and BMJ). Of these, nine were glucose metabolism-related namely, 6-phosphogluconate
dehydrogenase [EC: 1.1.1.44]
D
glucokinase [EC: 2.7.1.2], glucose-6-phosphate 1-dehydrogenase [EC: 1.1.1.49],
PT E
glucose-6-phosphate 1-epimerase [EC: 5.1.3.15], phosphoglycerate mutase [EC: 5.4.2.1], thiamine pyrophosphokinase [EC: 2.7.6.2], UDP-galactopyranose mutase
CE
[EC: 5.4.99.9], and UDP-glucose 4-epimerase [EC: 5.1.3.2]. This implies that the
AC
jaundice group may have weaker microbial glucose metabolism than the control group. Acetylornithine/N-succinyl diaminopimelate aminotransferase [EC: 2.6.1.11, 2.6.1.17] showed a significant increase in the jaundice groups (NC and BMJ) compared with the level in CON.
Significantly enriched species/KOs are associated with clinical indices To further screen for gut microbiota associated with clinical indices, we performed a
ACCEPTED MANUSCRIPT regression analysis of NC- and jaundice-related species and KOs with important clinical indices and found that a total of 16 species were associated with DBIL (Figure 4a). Nine of these were Bifidobacterium, namely, eight Bifidobacterium longum strains and one Bifidobacterium bifidum BGN4 (Figure 2c and
PT
Supplementary Figure 3a). A total of 19 KOs were also associated with DBIL
RI
(Figure 4b), of which 15 and 4 were negatively and positively correlated with it,
SC
respectively. Of the KOs negatively correlated with DBIL, six were glucose metabolism-related genes, including phosphoglycerate mutase [EC: 5.4.2.1], UDP 4
epimerase
5.1.3.2],
1-dehydrogenase
β-galactosidase
[EC:
1.1.1.49],
[EC:
3.2.1.23],
phosphogluconate
MA
glucose-6-phosphate
[EC:
NU
glucose
dehydrogenase [EC: 1.1.1.44], UDP galactopyranose mutase [EC: 5.4.99.9], and
D
glucokinase [EC: 2.7.1.2]. Additionally, cell filamentation protein, a possible bacterial
PT E
virulence factor-related protein, was positively correlated with DBIL. Further, some species and KOs were found to be correlated with DBIL/TBIL and ALT
AC
CE
(Supplementary Figure 4).
Significantly enriched species are potential biomarkers of jaundice In order to further determine the species that can be used as biomarkers of NC patients, the 13 species with significant differences in abundance between jaundice patients (BMJ and NC) and healthy controls (CON) without jaundice were used to train the SVM model. The results showed that a clear distinction between jaundice patients and non-jaundice controls could be achieved based on 10 bacteria species
ACCEPTED MANUSCRIPT (Supplementary Table 7), with an area under ROC curve (AUC) value of 0.91 (Figure 5), meaning that these 10 species are effective biomarkers of jaundice.
Discussion
PT
Neonatal cholestasis can cause a shortage of bile acids in extrahepatic regions,
RI
especially in the gut. This deficiency in the gut may promote the translocation of
SC
bacteria, as bile acids can not only inhibit microbes from entering enterocytes [16], but also directly inhibit the proliferation of pathogenic microbes [17]. In accordance
NU
with this, in the present study, we found that cholestasis was accompanied by a
MA
significant increase in gut microbiota richness. We also found that a total of 113 species were upregulated in NC patients. In particular, we found that potentially
D
pathogenic species, such as five species of Streptococcus, three species of
PT E
Mycobacterium, two strains of Rhodopseudomonas palustris, and two strains of Veillonella parvula, were upregulated in NC patients. Studies have shown that
CE
pathogenic microbes can kill commensal bacteria via virulence factors such as type VI
AC
secretion system (T6SS), whereas commensal bacteria can inhibit T6SS-mediated killing of commensal bacteria by modifying bile acids. A lack of intestinal bile acids will lead to disequilibrium in this regard, exacerbating the reduction of beneficial bacteria [18]. In agreement with this speculation, we found that potentially pathogenic species as well as gut microbiota functional genes that were linked to virulence factors, such as pilus assembly protein CpaC, pilus assembly protein FimV, spore germination protein D, spore germination protein KC, type III secretion protein SctC,
ACCEPTED MANUSCRIPT type IV pilus assembly protein PilQ, and type I polyketide synthase AVES, were significantly upregulated in NC patients. In particular, type III secretion protein SctC was positively correlated with serum bilirubin levels. We also found that a large number of commensal bacteria, such as Bifidobacterium and Bifidobacteriaceae, were
PT
significantly downregulated in NC patients, and the decrease of Bifidobacterium is
RI
associated with an increase in serum bilirubin levels. In summary, we conclude that
SC
the occurrence of NC is closely related to the large-scale growth of potentially pathogenic bacteria and the large reduction in beneficial bacteria such as
NU
Bifidobacterium.
MA
The occurrence of neonatal cholestasis is also accompanied by jaundice. The main manifestation of jaundice is an increase in serum bilirubin. It is well known that
D
bilirubin formed by the destruction of erythrocytes first enters the liver and then
PT E
combines with UDP-glucuronic acid (UDPGA) to form water-soluble conjugated bilirubin. Subsequently, bilirubin is secreted into the biliary tract and enters the
CE
intestine for excretion [19]. Gut microbiota is important in this context as it may play
AC
a major role in bilirubin efflux. Using the germ-free multidrug resistance 2 knockout mouse model, Tabibian and coworkers found that germ-free mice showed upregulated serum bilirubin levels [20], which provides evidence that the gut microbiota may participate in bilirubin metabolism. Moreover, Tuzun et al. indicated that downregulation of members of the Bifidobacterium genus, such as B. adolescentis, B. bifidum, and B. longum, was associated with an increase in serum bilirubin level [21]. This is consistent with our results, as we also found that a decrease in members of the
ACCEPTED MANUSCRIPT Bifidobacterium genus was associated with an increase in serum direct bilirubin level. Based on previous research and our present findings, we speculated that the decrease in Bifidobacterium may have been directly or indirectly caused by the increase in direct bilirubin. We conducted a comparative analysis of gut microflora functional
PT
levels between jaundice (NC and BMJ) groups and the group without jaundice (CON)
RI
as well as an abundance of galactose metabolism-related genes downregulated in the
SC
jaundice groups. Bifidobacterium is well known to be directly involved in the utilization of galactooligosaccharides (GOSs) [22], and transforming GOSs to
NU
galactose and UDP-glucose through the galactose metabolism pathway, while
MA
UDP-glucose, a product of galactose metabolism, is a direct precursor to UDPGA. Combined with the previous studies, it may be conjectured that Bifidobacterium is
D
involved in the metabolism of bilirubin by the galactose metabolism pathway. The
PT E
decrease in Bifidobacterium, coupled with the decreased abundance of galactose metabolism pathway-related genes, leads to a shortage of intestinal UDP-glucose and
CE
affecting the downstream glucuronic acid pathway to form conjugated bilirubin. At
AC
the same time, we also identified a decreased gut microbiota abundance of glucokinase-generated species; this may have exacerbated the decrease in UDP-glucose in the pentose phosphate pathway. Finally, we found a decrease in the abundance of phosphogluconate dehydrogenase (PGM)-associated gut microbiota, which may indicate abnormalities in the ATP-related glycolysis pathway and affect the glucuronic acid pathway that requires ATP (Supplementary Figure 5). Breastfeeding is considered to be a risk factor for BMJ, however, whether
ACCEPTED MANUSCRIPT breastfeeding plays a role in the pathogenesis of NC has not been well studied. In the present study, we found that the breast-fed and non-breast-fed NC patients had similar microbiota profiles at the phylum level; this may imply that the bile acid pool and enterohepatic circulation form a major determinant, or one that has greater influence
PT
than diet, of gut microbiota composition.
RI
There are some strengths in the present study. The first strength is the study
SC
design, which included three groups and the categorization of cases with/without NC and with/without jaundice to identify gut microbiota composition associated with NC
NU
and jaundice, respectively. Second, in contrast to the conventional 16S rDNA
MA
sequencing, the shotgun metagenomic sequencing method applied here enabled us to study the gut microbiota composition at the functional level. Finally, the analysis of
D
the correlation between the selected differential metagenomic components and
PT E
important clinical phenotypes was also helpful in determining the gut microbiota composition associated with NC and jaundice. However, there are also some
CE
limitations to the present study. Firstly, the sample size was not large enough.
AC
Secondly, there was a lack of assessment of some of the variables affecting the gut microbiota composition, such as antibiotic treatment, length of hospitalization, etc. Considering that there were no cases of jaundice in the healthy control group, we did not collect clinical indices such as serum bilirubin level for this group. Additionally, an independent validation cohort would be necessary to further evaluate whether Bifidobacterium is a biomarker of jaundice, and whether Bifidobacterium supplementation can mitigate the symptom of jaundice. Although it is hard to see an
ACCEPTED MANUSCRIPT immediate path to clinical use when fractionated bilirubin also accomplishes the same purpose quickly and at a significantly lower cost, the present study is of great significance in understanding the mechanisms of the occurrence of jaundice and in developing corresponding treatments. The search for markers related to pathological
PT
jaundice is of great significance, especially prior to the marked elevation of direct
SC
RI
bilirubin levels; further study in this regard needs to be carried out in the next step.
Conclusion
NU
In conclusion, the alteration of gut microbiota in NC patients mainly manifested as a
MA
significant increase in species richness. The main manifestation in jaundice patients is a significant decrease in Bifidobacterium, which may be involved in the metabolism
D
of bilirubin through the galactose metabolism pathway. The results of this study
PT E
suggest an imbalance in the gut microbiota of NC patients, which is associated with a substantial reduction in Bifidobacterium, suggesting the possibility of treating
AC
CE
jaundice through probiotic supplementation.
Acknowledgments We are sincerely grateful to the individuals who participated in this study.
Funding/Support: This project was supported by the National Natural Science Foundation of China (Program Nos. 81701351 and 81720108018) and Shenzhen Science
Technology
and
Innovation
Commission
(Program
Nos.
ACCEPTED MANUSCRIPT JCYJ20160429174706491 and JCYJ20170413093358429), the National Key Research and Development Program of China (program No. 2016YFC0905100 and 2017YFA0104204).
1.
PT
References Justinich, C.J. and J.S. Hyams, Neonatal Jaundice. 1998: Humana Press.
RI
293-300.
Mccandless, D.W., Breast Milk Jaundice. 2011: Humana Press. 115-120.
3.
Balistreri, W.F., Neonatal cholestasis. J Pediatr, 1985. 106(2): p. 171-84.
4.
Gotze, T., et al., Neonatal Cholestasis - Differential Diagnoses, Current
NU
SC
2.
Diagnostic Procedures, and Treatment. Front Pediatr, 2015. 3: p. 43. Hoerning, A., et al., Diversity of disorders causing neonatal cholestasis - the
MA
5.
experience of a tertiary pediatric center in Germany. Front Pediatr, 2014. 2: p.
Mueller, N.T., et al., The infant microbiome development: mom matters.
PT E
6.
D
65.
Trends Mol Med, 2015. 21(2): p. 109-17. 7.
Poland, R.L. and G.B. Odell, Physiologic jaundice: the enterohepatic
8.
CE
circulation of bilirubin. N Engl J Med, 1971. 284(1): p. 1-6. Visschers, R.G., et al., The gut-liver axis. Curr Opin Clin Nutr Metab Care,
9.
AC
2013. 16(5): p. 576-81. Fawaz, R., et al., Guideline for the Evaluation of Cholestatic Jaundice in Infants: Joint Recommendations of the North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition and the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition. J Pediatr Gastroenterol Nutr, 2017. 64(1): p. 154-168. 10.
Moyer, V., et al., Guideline for the evaluation of cholestatic jaundice in infants: recommendations
of
the
North
American
Society
for
Pediatric
Gastroenterology, Hepatology and Nutrition. J Pediatr Gastroenterol Nutr,
ACCEPTED MANUSCRIPT 2004. 39(2): p. 115-28. 11.
Zhou, S., et al., Diversity of Gut Microbiota Metabolic Pathways in 10 Pairs of Chinese Infant Twins. PLoS One, 2016. 11(9): p. e0161627.
12.
Huson, D.H., et al., MEGAN analysis of metagenomic data. Genome Res, 2007. 17(3): p. 377-86. Dixon, P., VEGAN, a package of R functions for community ecology. Journal
PT
13.
of Vegetation Science, 2003. 14(6): p. 927-930.
Kanehisa, M. and S. Goto, KEGG: kyoto encyclopedia of genes and genomes.
RI
14.
Nucleic Acids Res, 2000. 28(1): p. 27-30.
Love, M.I., W. Huber, and S. Anders, Moderated estimation of fold change
SC
15.
and dispersion for RNA-seq data with DESeq2. Genome Biol, 2014. 15(12): p.
16.
NU
550.
Wells, C.L., R.P. Jechorek, and S.L. Erlandsen, Inhibitory effect of bile on
MA
bacterial invasion of enterocytes: possible mechanism for increased translocation associated with obstructive jaundice. Crit Care Med, 1995. 23(2):
Buffie, C.G., et al., Precision microbiome reconstitution restores bile acid
PT E
17.
D
p. 301-7.
mediated resistance to Clostridium difficile. Nature, 2015. 517(7533): p. 205-8.
Bachmann, V., et al., Bile Salts Modulate the Mucin-Activated Type VI
CE
18.
Secretion System of Pandemic Vibrio cholerae. PLoS Negl Trop Dis, 2015.
19.
AC
9(8): p. e0004031. Billing, B.H., Twenty-five years of progress in bilirubin metabolism (1952-77). Gut, 1978. 19(6): p. 481-91. 20.
Tabibian, J.H., et al., Absence of the intestinal microbiota exacerbates hepatobiliary disease in a murine model of primary sclerosing cholangitis. Hepatology, 2016. 63(1): p. 185-96.
21.
Tuzun, F., et al., Breast milk jaundice: effect of bacteria present in breast milk and infant feces. J Pediatr Gastroenterol Nutr, 2013. 56(3): p. 328-32.
22.
Thongaram, T., et al., Prebiotic Galactooligosaccharide Metabolism by
ACCEPTED MANUSCRIPT Probiotic Lactobacilli and Bifidobacteria. J Agric Food Chem, 2017. 65(20): p.
AC
CE
PT E
D
MA
NU
SC
RI
PT
4184-4192.
ACCEPTED MANUSCRIPT Tables and legends
CON
BMJ
(Non-NC)
(Non-NC)
Participants(n=)
14
16
Gender(F/M)
3/11
2/14
Age(year)(mean±SD)
2.8±0.89
1.1±0.25
2.2±1.2
BF(n=;Yes/no)
14/0
16/0
15/25
VD(n=;Yes/no)
12/2
RI
Table 1. Characteristics of the Cohort
8/8
31/9
ALT(IU/l)(mean±SD)
NA
25±17
130±120
TBIL(umol/l)(mean±SD)
NA
210±66
170±78
DBIL(umol/l)(mean±SD)
NA
12±5.8
81±39
DBIL/TBIL(%)(mean±SD)
NA
6±3.6
50±11
Clinical indices
NC
PT
SC
NU
MA
40 21/19
Note: ALT, alanine transaminase, BF, breastfeeding, BMJ, breast milk jaundice; CON,
D
control; DBIL, direct bilirubin; F, female; M, male; NC, neonatal cholestasis; non-NC,
PT E
non-neonatal cholestasis; TBIL, total bilirubin; VD, vaginal delivery Figures and legends
CE
Figure 1. Flowchart of study design. ALT, alanine transaminase; BF, breastfeeding; BMJ, breast milk jaundice; CON, control; DBIL, direct bilirubin; FDR, false
AC
discovery rate; KEGG, Kyoto Encyclopedia of Genes and Genomes; MEGAN5, Metagenome Analyzer version 5; NC, neonatal cholestasis; NCBI, National Center for Biotechnology Information; non-NC, non-neonatal cholestasis; KO, KEGG Orthology; TBIL, total bilirubin
Figure 2. Altered gut microbiota composition in neonatal cholestasis patients. a, Neonatal cholestasis (NC) patients can be clearly separated from breast milk jaundice (BMJ) cases and healthy controls (CON) without jaundice according to gut
ACCEPTED MANUSCRIPT microbiota functional level composition. b, Increased species richness in gut microbiota of NC patients compared with that in BMJ and CON. c, Bifidobacterium longum significant decreased in NC. *, P value <0.05; **, P value <0.01, Wilcoxon’s rank sum test
PT
Figure 3. Impact of breastfeeding on gut microbiota of NC at phylum level. a, Comparison of high level taxa between subgroups CON (healthy controls), BMJ,
RI
NC_BF (NC with breastfeeding), NC_nonBF (NC without breastfeeding); b, comparison of phylum Actinobacteria between subgroups CON, BMJ, NC_BF and
SC
NC_nonBF; c, comparison of phylum Bacteroide between subgroups CON, BMJ, NC_BF and NC_nonBF; *P value<0.05; **, P value<0.01; ***, P value<0.001,
NU
Wilcoxon rank sum test; Bacteroidetes significantly increased in NC patients, and Actinobacteria significantly decreased in NC patients, independently of breastfeeding
MA
(BF) impact.
D
Figure 4. Decrease of Bifidobacterium longum was associated with jaundice. a,
PT E
Bifidobacterium longum was negatively associated with serum bilirubin levels. b, Abundance of metabolism pathway-related genes in gut microbiota was negatively
CE
associated with serum bilirubin level. Figure 5. Receiver operating characteristic curve analysis results. With 10 species
AC
identified by SVM model training, an AUC value of 0.91 can be achieved to distinguish between jaundice patients (BMJ and NC) and non-jaundice (CON) controls.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5