Background

The intestinal barrier plays an important role in the defense against infections, and nutritional, endocrine, and immune functions. The gut microbiota playing important role in development of the gastrointestinal tract can impact intestinal permeability and immunity during early life, but data concerning this problem are scare.

Methods

We analyzed the microbiota in fecal samples (101 samples in total) collected longitudinally over 24 months from 21 newborns to investigate whether the markers of small intestinal paracellular permeability (zonulin) and immune system development (calprotectin) are linked to the gut microbiota. The results were validated using data from an independent cohort that included the calprotectin and gut microbiota in children during the first year of life.

Results

Zonulin levels tended to increase for up to 6 months after childbirth and stabilize thereafter remaining at a high level while calprotectin concentration was high after chilbirth and begun to decline from 6 months of life. The gut microbiota composition and the related metabolic potentials changed during the first two years of life and were correlated with zonulin and calprotectin levels. Feacal calprotectin correlated inversely with alpha diversity (Shannon index, r =-0.30, FDR P (Q)=0.039). It also correlated with seven taxa; i.a. negatively with Ruminococccaceae (r=-0.34, Q=0.046), and Clostridiales (r=-0.34, Q=0.048) and positively with *Stapylococcus* (r=0.38, Q=0.023) and Staphylococcaceae (r=0.35, Q=0.04), whereas zonulin correlated with 19 taxa; i.a. with Bacillales (r=-0.52, Q=0.0004), Clostridiales (r=0.48, Q=0.001) and the* Ruminococcus *(*torques* group) (r=0.40, Q=0.026). When time intevals were considered only changes in abundance of the* Ruminococcus *(*torques* group) were associcated with changes in calprotectin (β=2.94, SE=0.8, Q=0.015). The dynamics of stool calprotectin was negatively associated with changes in two MetaCyc pathways: pyruvate fermentation to butanoate (β=-4.54, SE=1.08, Q=0.028) and *Clostridium acetobutylicum* fermentation** **(β=-4.48, SE=1.16, Q=0.026).

Conclusions

The small intestinal paracellular permeability, immune system-related markers and gut microbiota change dynamically during the first two years of life. The *Ruminococcus *(*torques* group) seems to be especially involved in controlling paracellular permeability. *Staphylococcus, *Staphylococcaceae, Ruminococcaceae, and Clostridiales, may be potential biomarkers of the immune system. Despite observed correlations their clear causation and health consequences were not proven. Mechanistic studies are required.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

This is a list of supplementary files associated with this preprint. Click to download.

- SuppFigure1.tif
Sample selection and availability (PMU cohort) From 100 healthy, full-term newborns during the period from March 2015 to April 2016, 18 mother + child pairs were initially selected with the highest number of samples available. Since only in two cases the delivery was natural and neither the mother nor the child was treated with antibiotics, it was decided to supplement this cohort with six pairs of mother + child who were not given antibiotics and the delivery was natural. Out of the 24 newborns that were selected in this way, three newborns (F19, F22, F24) were excluded due to inadequate number of samples. Twenty-one newborns (101 samples in total, green tiles) were included, in whom at least four longitudinal stool samples were available.

- SuppFigure2.tif
Study flow chart, including zonulin and calprotectin (PMU cohort) The number of samples for downstream analyses might differ due to results out of determination limits (zonulin 800 ng/mL and calprotectin 2100 ug/mL), technical problems (small volume of collected stool specimen, inadequate amount of DNA, sequencing depth), participant attrition.

- SuppFigure3.tif
Stool calprotectin level by time in the HMS cohort Likelihood ratio test, df=3, P=2.56e-10, adjusted for mode of delivery Notched boxplot with variable widths proportional to the square-roots of the number of observations in the groups; FDR adjusted p-values < 0.05 are shown, P2 - 10th day, P3 - 1st month, P4 - 6th month, P5 - 12th month.

- SuppFigure4.tif
Alpha and beta diversity over time in the HMS cohort A - Shannon alpha diversity by time, LRT, df=4, P< 2.2e-16, adjusted for mode of delivery, notched boxplot with variable widths proportional to the square-roots of the number of observations in the groups, B - Principal coordinate analysis plot with Bray-Curtis dissimilarity calculated from genus abundances, ellipses were drawn assuming a multivariate t-distribution, C - PCo1 scores by time, LRT, df = 4, P = 2.97e-14, adjusted for mode of delivery; D - PCo2 scores by time, LRT, df = 4, P = 7.31e-11; FDR adjusted p values < 0.05 are shown P1 - 1st day (meconium/the first stool), P2 - 10th day, P3 - 1st month, P4 - 6th month, P5 - 12th month

- SuppFigure5.tif
Gut microbiota composition change over time (PMU cohort) A linear mixed effects analysis followed by pairwise comparison of time points (adjusted for mode of delivery and breastfeeding time). The overall p-value - a likelihood ratio test (LRT) of nested models (FDR adjusted across genera); PXPY - contrast p values between the two time points (PX and PY), FDR adjusted for all possible contrasts, t.ratio - t statistics for the contrasts estimates (a positive value, colored blue, indicates a decrease abundance, a negative value, colored red, indicates increase in abundance). Taxa abundances (unrarefied) were transformed by generating 128 Monte Carlo instances of the Dirichlet distribution for each gut sample, followed by center-logtransform of each instance. A linear mixed effects analysis was performed for each instance separately and the results were averaged over 128 instances. P2 - 7th day, P3 - 1st month, P4 - 6th month, P5 - 12th month, P6 - 24th month

- SuppFigure6.tif
Gut microbiota composition change over time (HMS cohort) A linear mixed effects analysis followed by pairwise comparison of time points (adjusted for mode of delivery) for five taxonomic ranks (present in at least 10% samples), only taxons significantly associated with time are shown. The overall p-value - a likelihood ratio test (LRT) of nested models (FDR adjusted across taxons); PXPY - contrast p-values between the two time points (PX and PY), FDR adjusted for all possible contrasts, t.ratio - t statistics for the contrasts estimates (a positive value, colored blue, indicates a decrease in abundance, a negative value, colored red, indicates an increase in abundance). Taxa abundances (unrarefied) were transformed by generating 128 Monte Carlo instances of the Dirichlet distribution for each gut sample, followed by center-log transform of each instance. A linear mixed effects analysis was performed for each instance separately and the results were averaged over 128 instances. P2 - 10th day, P3 - 1st month, P4 - 6th month, P5 - 12th month

- SuppFigure7.tif
Predicted MetaCyc pathways that change significantly over time in both cohorts (PMU and HMS) Linear mixed effects analysis followed by pairwise comparison of time points (adjusted for mode of delivery and breastfeeding time (PMU) and mode of delivery only in the HMS cohort). The overall p-value - a likelihood ratio test (LRT) of nested models (FDR adjusted across pathways); PXPY - contrast p values between the two time points (X and Y), FDR adjusted for all possible contrasts, t.ratio - t statistics for the contrasts estimates. Pathway abundances (unrarefied) were transformed by generating 128 Monte Carlo instances of the Dirichlet distribution for each gut sample, followed by center-log transform of each instance. A linear mixed effects analysis was performed for each instance separately and the results were averaged over 128 instances. P2 - 7th day (PMU) or 10th day (HMS), P3 - 1st month, P4 - 6th month, P5 - 12th month, P6 - 24th month (PMU only).

- SuppFigure8.tif
Taxon change versus zonulin change (PMU cohort) Linear mixed effects models were used to test for an association between the taxon abundance change and zonulin change accounting for ten time point pairs (P2-P3, P3-P4, etc.) from the same subject, adjusted for mode of delivery and breastfeeding time. Two models were considered: without interaction (w/o int.) and with interaction between time point pair and taxon change (w/ int.). If the interaction term was significant - Q (w/ int.) < 0.05, individual p values (FDR adjusted) were interpreted. If the common slope model was chosen - Q (w/o int.) < 0.05 and Q (w /int) > 0.05, the same common slope coefficient β (coef.) across all time points and one p-value (P2-P3) were shown. Coefficients and Q values were averaged over 128 Monte Carlo instances of the Dirichlet distribution, followed by center-log transform of each instance.

- SuppFigure9.tif
Taxon change versus calprotectin change (PMU cohort) Linear mixed effects models were used to test for an association between the taxon abundance change and calprotectin change accounting for ten time point pairs (P2-P3, P3-P4, etc.) from the same subject, adjusted for mode of delivery and breastfeeding time. Two models were considered: without interaction (w/o int.) and with interaction between time point pair and taxon change (w/ int.). If the interaction term was significant - Q (w/ int.) < 0.05, individual p values (FDR adjusted) were interpreted. If the common slope model was chosen - Q (w/o int.) < 0.05 and Q (w /int) > 0.05, the same common slope coefficient β (coef.) across all time points and one p-value (P2-P3) were shown. Coefficients and Q values were averaged over 128 Monte Carlo instances of the Dirichlet distribution, followed by center-log transform of each instance.

- SuppFigure11.tif
Pathway change vs calprotectin change (HMS cohort) Linear mixed effects models were used to test for an association between the pathway abundance change and calprotectin change accounting for six time point pairs (P2-P3, P3-P4, etc.) from the same subject, adjusted for mode of delivery. Two models were considered: without interaction (w/o int.) and with interaction between time point pair and taxon change (w/ int.). If the interaction term was significant - Q (w/ int.) < 0.05, individual p values (FDR adjusted) were interpreted. If the common slope model was chosen - Q (w/o int.) < 0.05 and Q (w /int) > 0.05, the same common slope coefficient β (coef.) across all time points and one p-value (P2-P3) were shown. Coefficients and Q values were averaged over 128 Monte Carlo instances of the Dirichlet distribution, followed by center-log transform of each instance.

- graphicalabstract.png

Loading...

Posted 29 Jan, 2021

###### No community comments so far

###### Reviewer #3 agreed

On 26 Jan, 2021

###### Reviewer #2 agreed

On 25 Jan, 2021

###### Reviewers invited

Invitations sent on 25 Jan, 2021

###### Reviewer #1 agreed

On 25 Jan, 2021

###### Editor assigned

On 24 Jan, 2021

###### Editor invited

On 24 Jan, 2021

###### Submission checks complete

On 24 Jan, 2021

###### First submitted

On 22 Jan, 2021

Posted 29 Jan, 2021

###### No community comments so far

###### Reviewer #3 agreed

On 26 Jan, 2021

###### Reviewer #2 agreed

On 25 Jan, 2021

###### Reviewers invited

Invitations sent on 25 Jan, 2021

###### Reviewer #1 agreed

On 25 Jan, 2021

###### Editor assigned

On 24 Jan, 2021

###### Editor invited

On 24 Jan, 2021

###### Submission checks complete

On 24 Jan, 2021

###### First submitted

On 22 Jan, 2021

Background

The intestinal barrier plays an important role in the defense against infections, and nutritional, endocrine, and immune functions. The gut microbiota playing important role in development of the gastrointestinal tract can impact intestinal permeability and immunity during early life, but data concerning this problem are scare.

Methods

We analyzed the microbiota in fecal samples (101 samples in total) collected longitudinally over 24 months from 21 newborns to investigate whether the markers of small intestinal paracellular permeability (zonulin) and immune system development (calprotectin) are linked to the gut microbiota. The results were validated using data from an independent cohort that included the calprotectin and gut microbiota in children during the first year of life.

Results

Zonulin levels tended to increase for up to 6 months after childbirth and stabilize thereafter remaining at a high level while calprotectin concentration was high after chilbirth and begun to decline from 6 months of life. The gut microbiota composition and the related metabolic potentials changed during the first two years of life and were correlated with zonulin and calprotectin levels. Feacal calprotectin correlated inversely with alpha diversity (Shannon index, r =-0.30, FDR P (Q)=0.039). It also correlated with seven taxa; i.a. negatively with Ruminococccaceae (r=-0.34, Q=0.046), and Clostridiales (r=-0.34, Q=0.048) and positively with *Stapylococcus* (r=0.38, Q=0.023) and Staphylococcaceae (r=0.35, Q=0.04), whereas zonulin correlated with 19 taxa; i.a. with Bacillales (r=-0.52, Q=0.0004), Clostridiales (r=0.48, Q=0.001) and the* Ruminococcus *(*torques* group) (r=0.40, Q=0.026). When time intevals were considered only changes in abundance of the* Ruminococcus *(*torques* group) were associcated with changes in calprotectin (β=2.94, SE=0.8, Q=0.015). The dynamics of stool calprotectin was negatively associated with changes in two MetaCyc pathways: pyruvate fermentation to butanoate (β=-4.54, SE=1.08, Q=0.028) and *Clostridium acetobutylicum* fermentation** **(β=-4.48, SE=1.16, Q=0.026).

Conclusions

The small intestinal paracellular permeability, immune system-related markers and gut microbiota change dynamically during the first two years of life. The *Ruminococcus *(*torques* group) seems to be especially involved in controlling paracellular permeability. *Staphylococcus, *Staphylococcaceae, Ruminococcaceae, and Clostridiales, may be potential biomarkers of the immune system. Despite observed correlations their clear causation and health consequences were not proven. Mechanistic studies are required.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

This is a list of supplementary files associated with this preprint. Click to download.

- SuppFigure1.tif
Sample selection and availability (PMU cohort) From 100 healthy, full-term newborns during the period from March 2015 to April 2016, 18 mother + child pairs were initially selected with the highest number of samples available. Since only in two cases the delivery was natural and neither the mother nor the child was treated with antibiotics, it was decided to supplement this cohort with six pairs of mother + child who were not given antibiotics and the delivery was natural. Out of the 24 newborns that were selected in this way, three newborns (F19, F22, F24) were excluded due to inadequate number of samples. Twenty-one newborns (101 samples in total, green tiles) were included, in whom at least four longitudinal stool samples were available.

- SuppFigure2.tif
Study flow chart, including zonulin and calprotectin (PMU cohort) The number of samples for downstream analyses might differ due to results out of determination limits (zonulin 800 ng/mL and calprotectin 2100 ug/mL), technical problems (small volume of collected stool specimen, inadequate amount of DNA, sequencing depth), participant attrition.

- SuppFigure3.tif
Stool calprotectin level by time in the HMS cohort Likelihood ratio test, df=3, P=2.56e-10, adjusted for mode of delivery Notched boxplot with variable widths proportional to the square-roots of the number of observations in the groups; FDR adjusted p-values < 0.05 are shown, P2 - 10th day, P3 - 1st month, P4 - 6th month, P5 - 12th month.

- SuppFigure4.tif
Alpha and beta diversity over time in the HMS cohort A - Shannon alpha diversity by time, LRT, df=4, P< 2.2e-16, adjusted for mode of delivery, notched boxplot with variable widths proportional to the square-roots of the number of observations in the groups, B - Principal coordinate analysis plot with Bray-Curtis dissimilarity calculated from genus abundances, ellipses were drawn assuming a multivariate t-distribution, C - PCo1 scores by time, LRT, df = 4, P = 2.97e-14, adjusted for mode of delivery; D - PCo2 scores by time, LRT, df = 4, P = 7.31e-11; FDR adjusted p values < 0.05 are shown P1 - 1st day (meconium/the first stool), P2 - 10th day, P3 - 1st month, P4 - 6th month, P5 - 12th month

- SuppFigure5.tif
Gut microbiota composition change over time (PMU cohort) A linear mixed effects analysis followed by pairwise comparison of time points (adjusted for mode of delivery and breastfeeding time). The overall p-value - a likelihood ratio test (LRT) of nested models (FDR adjusted across genera); PXPY - contrast p values between the two time points (PX and PY), FDR adjusted for all possible contrasts, t.ratio - t statistics for the contrasts estimates (a positive value, colored blue, indicates a decrease abundance, a negative value, colored red, indicates increase in abundance). Taxa abundances (unrarefied) were transformed by generating 128 Monte Carlo instances of the Dirichlet distribution for each gut sample, followed by center-logtransform of each instance. A linear mixed effects analysis was performed for each instance separately and the results were averaged over 128 instances. P2 - 7th day, P3 - 1st month, P4 - 6th month, P5 - 12th month, P6 - 24th month

- SuppFigure6.tif
Gut microbiota composition change over time (HMS cohort) A linear mixed effects analysis followed by pairwise comparison of time points (adjusted for mode of delivery) for five taxonomic ranks (present in at least 10% samples), only taxons significantly associated with time are shown. The overall p-value - a likelihood ratio test (LRT) of nested models (FDR adjusted across taxons); PXPY - contrast p-values between the two time points (PX and PY), FDR adjusted for all possible contrasts, t.ratio - t statistics for the contrasts estimates (a positive value, colored blue, indicates a decrease in abundance, a negative value, colored red, indicates an increase in abundance). Taxa abundances (unrarefied) were transformed by generating 128 Monte Carlo instances of the Dirichlet distribution for each gut sample, followed by center-log transform of each instance. A linear mixed effects analysis was performed for each instance separately and the results were averaged over 128 instances. P2 - 10th day, P3 - 1st month, P4 - 6th month, P5 - 12th month

- SuppFigure7.tif
Predicted MetaCyc pathways that change significantly over time in both cohorts (PMU and HMS) Linear mixed effects analysis followed by pairwise comparison of time points (adjusted for mode of delivery and breastfeeding time (PMU) and mode of delivery only in the HMS cohort). The overall p-value - a likelihood ratio test (LRT) of nested models (FDR adjusted across pathways); PXPY - contrast p values between the two time points (X and Y), FDR adjusted for all possible contrasts, t.ratio - t statistics for the contrasts estimates. Pathway abundances (unrarefied) were transformed by generating 128 Monte Carlo instances of the Dirichlet distribution for each gut sample, followed by center-log transform of each instance. A linear mixed effects analysis was performed for each instance separately and the results were averaged over 128 instances. P2 - 7th day (PMU) or 10th day (HMS), P3 - 1st month, P4 - 6th month, P5 - 12th month, P6 - 24th month (PMU only).

- SuppFigure8.tif
Taxon change versus zonulin change (PMU cohort) Linear mixed effects models were used to test for an association between the taxon abundance change and zonulin change accounting for ten time point pairs (P2-P3, P3-P4, etc.) from the same subject, adjusted for mode of delivery and breastfeeding time. Two models were considered: without interaction (w/o int.) and with interaction between time point pair and taxon change (w/ int.). If the interaction term was significant - Q (w/ int.) < 0.05, individual p values (FDR adjusted) were interpreted. If the common slope model was chosen - Q (w/o int.) < 0.05 and Q (w /int) > 0.05, the same common slope coefficient β (coef.) across all time points and one p-value (P2-P3) were shown. Coefficients and Q values were averaged over 128 Monte Carlo instances of the Dirichlet distribution, followed by center-log transform of each instance.

- SuppFigure9.tif
Taxon change versus calprotectin change (PMU cohort) Linear mixed effects models were used to test for an association between the taxon abundance change and calprotectin change accounting for ten time point pairs (P2-P3, P3-P4, etc.) from the same subject, adjusted for mode of delivery and breastfeeding time. Two models were considered: without interaction (w/o int.) and with interaction between time point pair and taxon change (w/ int.). If the interaction term was significant - Q (w/ int.) < 0.05, individual p values (FDR adjusted) were interpreted. If the common slope model was chosen - Q (w/o int.) < 0.05 and Q (w /int) > 0.05, the same common slope coefficient β (coef.) across all time points and one p-value (P2-P3) were shown. Coefficients and Q values were averaged over 128 Monte Carlo instances of the Dirichlet distribution, followed by center-log transform of each instance.

- SuppFigure11.tif
Pathway change vs calprotectin change (HMS cohort) Linear mixed effects models were used to test for an association between the pathway abundance change and calprotectin change accounting for six time point pairs (P2-P3, P3-P4, etc.) from the same subject, adjusted for mode of delivery. Two models were considered: without interaction (w/o int.) and with interaction between time point pair and taxon change (w/ int.). If the interaction term was significant - Q (w/ int.) < 0.05, individual p values (FDR adjusted) were interpreted. If the common slope model was chosen - Q (w/o int.) < 0.05 and Q (w /int) > 0.05, the same common slope coefficient β (coef.) across all time points and one p-value (P2-P3) were shown. Coefficients and Q values were averaged over 128 Monte Carlo instances of the Dirichlet distribution, followed by center-log transform of each instance.

- graphicalabstract.png

Loading...