ORIGINAL CONTRIBUTIONS
nature publishing group
1
Fecal Microbiota in Pediatric Inflammatory Bowel Disease and Its Relation to Inflammation Kaija-Leena Kolho, MD, PhD1, 6, Katri Korpela, MSc2, 6, Tytti Jaakkola, MD1, Madharasi V.A. Pichai, MSc3, Erwin G. Zoetendal, PhD4, Anne Salonen, PhD2 and Willem M. de Vos, PhD2, 4, 5 OBJECTIVES:
Inflammatory bowel disease (IBD) is considered to result from interplay between host and intestinal microbiota. While IBD in adults has shown to be associated with marked changes in the intestinal microbiota, there are only a few studies in children, and particularly studies focusing on therapeutic responses are lacking. Hence, this prospective study addressed the intestinal microbiota in pediatric IBD especially related to the level of inflammation.
METHODS:
In total, 68 pediatric patients with IBD and 26 controls provided stool and blood samples in a tertiary care hospital and 32 received anti-tumor necrosis factor-α (anti-TNF-α ). Blood inflammatory markers and fecal calprotectin levels were determined. The intestinal microbiota was characterized by phylogenetic microarray and qPCR analysis.
RESULTS:
The microbiota varied along a gradient of increasing intestinal inflammation (indicated by calprotectin levels), which was associated with reduced microbial richness, abundance of butyrate producers, and relative abundance of Gram-positive bacteria (especially Clostridium clusters IV and XIVa). A significant association between microbiota composition and inflammation was indicated by a set of bacterial groups predicting the calprotectin levels (area under curve (AUC) of 0.85). During the induction of anti-TNF-α , the microbial diversity and similarity to the microbiota of controls increased in the responder group by week 6, but not in the non-responders (P160 susceptibility loci to date (5). Several environmental risk factors have been identified including the absence of breastfeeding, the number of siblings, a family history of IBD, household characteristics, and dietary patterns (6). Most likely, the microbial environment affects the genetically determined immune responses orchestrating the IBD-related
1
Children’s Hospital, Helsinki University Central Hospital, University of Helsinki, Helsinki, Finland; 2Immunobiology Program, Department of Bacteriology and Immunology, University of Helsinki, Helsinki, Finland; 3Faculty of Medical and Health Sciences, University of Auckland, Auckland, New Zealand; 4Laboratory of Microbiology, Wageningen University, Wageningen, The Netherlands; 5Department of Veterinary Biosciences, University of Helsinki, Helsinki, Finland; 6Share the authorship for first author. Correspondence: Kaija-Leena Kolho, MD, PhD, Children’s Hospital, Helsinki University Central Hospital, University of Helsinki, PO Box 281, Helsinki FIN-00029 HUS, Finland. E-mail:kaija-leena.kolho@helsinki.fi Received 31 December 2014; accepted 14 April 2015
© 2015 by the American College of Gastroenterology
The American Journal of GASTROENTEROLOGY
PEDIATRICS
see related editorial on page x
PEDIATRICS
2
Kolho et al.
inflammation, although the specific role of different bacteria or alterations in bacterial diversity is unresolved (7). Antibiotics modify the intestinal microbiota, notably in early life (8) and there is evidence that antibiotic use associates with the development of CD (9,10). It is well established that the pediatric form of IBD has a more aggressive disease course than IBD in adults. This is reflected in the high use of corticosteroids and in the high frequency of surgeries (11,12). The diagnostic evaluation includes ileocolonoscopy but for monitoring disease activity the use of endoscopies is restricted in pediatric patients due to their invasiveness. Therefore, fecal surrogate markers of inflammation and especially fecal calprotectin are used more and more in clinical practice in the follow-up of IBD patients. Recently, for calprotectin a negative predictive value of 0.95 (95% confidence interval 0.86–0.99) with a cutoff of 100 μ g/g for the presence of intestinal inflammation was reported in pediatric IBD (13) being in line with adult data (14). The recent guidelines for treating pediatric CD (11) or UC (12) are mainly based on the data extrapolated from adult patients. In moderate to severe CD, and comparable to adults recently also in UC, the use of the tumor necrosis factor-α (TNF-α ) antagonist agent infliximab is approved for children older than 6 years of age and restricted to hospital use. The other TNF-α antagonist agent, adalimumab, has been used for study purposes (15,16), but only recently approved for clinical use in pediatric CD. Thus, the use of TNF-α blockers in pediatric IBD is likely to further increase in the future. A considerable proportion of the patients (almost every other patient), however, lose their therapeutic response and need dose escalation already during the first year of therapy (17–19). Presently, the loss of therapeutic effect cannot be foreseen and this strongly affects the costs associated with the use of these biologicals. Several studies have addressed the intestinal microbiota of IBD patients by using 16S rRNA-based high throughput approaches, including next-generation sequencing and phylogenetic microarray analysis (20). Decreased bacterial diversity and changes in the abundance of certain bacterial groups have been observed in the fecal samples and colonic biopsies of IBD patients, in comparison with the samples of healthy subjects in studies with unrelated subjects, monozygotic twins, and multi-centered studies (7,21–23). Several bacterial groups have been associated with IBD, including Escherichia and Bacteroides spp. as well as bacteria belonging to the Clostridiales. Faecalibacterium prausnitzii, an abundant intestinal bacterium belonging to the Clostridium cluster IV, was found to be frequently reduced in abundance in CD, and a reduction in the mucus-degrader Akkermansia muciniphila has been reported in CD and UC (22,24–26). A recent study related the gut microbiota to therapeutic responses and an association was described between fecal microbiota and the response to steroid therapy in pediatric patients hospitalized for severe UC (27). This suggests that characterization of gut microbiota may provide an additional tool in assessment of therapeutic responses warranting further studies. We hypothesized that alterations in gut microbiota associate with the level of The American Journal of GASTROENTEROLOGY
inflammation in pediatric IBD. Hence, we conducted a prospective study where we analyzed the fecal microbiota in pediatric IBD with special reference to its association with disease activity and therapeutic responses to the TNF-α antagonist agents, both related to fecal calprotectin values. The results using deep phylogenetic microarray analysis indicate that fecal microbiota tightly associated with the level of inflammation (reflected in fecal calprotectin). Intriguingly, responders and non-responders to anti-TNF-α therapy could be distinguished based on their fecal microbiota composition, and the response associated with simultaneous changes in microbial diversity and composition.
METHODS We invited 9- to 18-year-old Finnish-speaking patients with IBD followed up at Children’s Hospital, Helsinki, Finland to this prospective study on fecal microbiota performed between June 2011 and January 2013. The families of patients scheduled for pediatric gastroenterologist, colonoscopy, or antiTNF-α therapy were approached by a letter or contacted by the IBD study nurse describing the study protocol and requesting stool and blood samples at their next regular visit. All IBD diagnoses were endoscopy based and fulfilling the revised Porto criteria (28). All patients were requested to provide two stool samples for study purposes: one for the measurement of fecal calprotectin and one for microarray analyses. Samples were frozen at −20 °C within 4 h of defecation and stored at −70 °C until analyzed (see below). Those who were entering anti-TNF-α therapy were also requested to bring additional stool samples at weeks 2 and 6 after the start of the therapy and once during the maintenance therapy. Controls were non-IBD patients either scheduled for checkup for juvenile idiopathic arthritis or healthy children recruited from the Helsinki region that were recruited the same way as the patients. The study groups comprised 68 patients with IBD and 26 controls, of whom 18 presented with juvenile idiopathic arthritis (Table 1). The number of patients on TNF-α blockers was 32 (on infliximab n=31; on adalimumab n=1; diagnosis Crohn’s n=27; UC/IBDU=5). Of these 11 were introduced to the therapy during the study period (at baseline on glucocorticoids and/or azathioprine). The other patients were medicated with 5-aminosalicylic acid (n=23) and/or azathioprine (n=9), methotrexate (n=4), glucocorticoids (n=2), and none (n=6). All control children had low levels of fecal calprotectin (3-fold or normalization) in fecal calprotectin values and mild or no clinical disease activity (31). The fecal calprotectin level was measured in a routine clinical laboratory using a quantitative enzyme immunoassay (PhiCal Test, Calpro AS, Oslo, Norway), and the values of 98% Pearson similarity) were used for further analysis. In total, 280 microarray experiments were conducted in this study; two samples were of insufficient quality for the analyses.
Extensive colitis
15
Left-sided
11
Statistical analyses
6
All statistical analyses were conducted with the program R (37), using the packages vegan (38), gstat (39), nlme (40), and pROC (41). The microbiota data were converted into relative abundances. Diversity of the microbiota was calculated as the inverse Simpson diversity index, using the oligonucleotide-probe level data. Richness was calculated as the number of oligonucleotide probes with signal intensities higher than the median intensity over the whole data set. Similarity to control microbiota was calculated as the average correlation between the log-transformed oligonucleotide probe level data from the patient and the healthy control samples. Principal coordinate analysis was performed with the first samples of all individuals, using the log-transformed species-level data and Bray-Curtis distances (function “capscale” in the R-package vegan). The calprotectin, microbiota richness, abundance of groups associated with butyrate production, and the ratio between Grampositive and Gram-negative organisms in the principal coordinate space were calculated by inverse distance-weighted interpolation (function “idw” in the R-package gstat). The microbiota differences between the IBD and control cases were assessed using the genus-level data set and linear mixed effect models with individual as the random factor (function “lme” in the R-package nlme), since many individuals provided several samples. For each genus-level taxon, we fitted a model estimating the effect of diagnosis group (UC, CD, or control) on the relative abundance of the taxon. To study the association between the microbiota composition and calprotectin level, the calprotectin levels were modeled
Unclassified IBD (IBDU) Disease duration years, median (range)
3.5 (0–10.7)
Anti-TNF-α medication (no. of patients)
32 d
Introduction of therapy during the study period
11
Clinical disease activity Physician Global Assessment, median (range)
1 (1–3)
Remission according to PCDAI (Pediatric Crohn’s Disease Activity Index)
30/36
PUCAI (Pediatric Ulcerative Colitis Activity index)
17/32
a
Difference in the proportion of males, P=0.28 (proportion test). Difference in age between patients and controls P
nature publishing group
1
Fecal Microbiota in Pediatric Inflammatory Bowel Disease and Its Relation to Inflammation Kaija-Leena Kolho, MD, PhD1, 6, Katri Korpela, MSc2, 6, Tytti Jaakkola, MD1, Madharasi V.A. Pichai, MSc3, Erwin G. Zoetendal, PhD4, Anne Salonen, PhD2 and Willem M. de Vos, PhD2, 4, 5 OBJECTIVES:
Inflammatory bowel disease (IBD) is considered to result from interplay between host and intestinal microbiota. While IBD in adults has shown to be associated with marked changes in the intestinal microbiota, there are only a few studies in children, and particularly studies focusing on therapeutic responses are lacking. Hence, this prospective study addressed the intestinal microbiota in pediatric IBD especially related to the level of inflammation.
METHODS:
In total, 68 pediatric patients with IBD and 26 controls provided stool and blood samples in a tertiary care hospital and 32 received anti-tumor necrosis factor-α (anti-TNF-α ). Blood inflammatory markers and fecal calprotectin levels were determined. The intestinal microbiota was characterized by phylogenetic microarray and qPCR analysis.
RESULTS:
The microbiota varied along a gradient of increasing intestinal inflammation (indicated by calprotectin levels), which was associated with reduced microbial richness, abundance of butyrate producers, and relative abundance of Gram-positive bacteria (especially Clostridium clusters IV and XIVa). A significant association between microbiota composition and inflammation was indicated by a set of bacterial groups predicting the calprotectin levels (area under curve (AUC) of 0.85). During the induction of anti-TNF-α , the microbial diversity and similarity to the microbiota of controls increased in the responder group by week 6, but not in the non-responders (P160 susceptibility loci to date (5). Several environmental risk factors have been identified including the absence of breastfeeding, the number of siblings, a family history of IBD, household characteristics, and dietary patterns (6). Most likely, the microbial environment affects the genetically determined immune responses orchestrating the IBD-related
1
Children’s Hospital, Helsinki University Central Hospital, University of Helsinki, Helsinki, Finland; 2Immunobiology Program, Department of Bacteriology and Immunology, University of Helsinki, Helsinki, Finland; 3Faculty of Medical and Health Sciences, University of Auckland, Auckland, New Zealand; 4Laboratory of Microbiology, Wageningen University, Wageningen, The Netherlands; 5Department of Veterinary Biosciences, University of Helsinki, Helsinki, Finland; 6Share the authorship for first author. Correspondence: Kaija-Leena Kolho, MD, PhD, Children’s Hospital, Helsinki University Central Hospital, University of Helsinki, PO Box 281, Helsinki FIN-00029 HUS, Finland. E-mail:kaija-leena.kolho@helsinki.fi Received 31 December 2014; accepted 14 April 2015
© 2015 by the American College of Gastroenterology
The American Journal of GASTROENTEROLOGY
PEDIATRICS
see related editorial on page x
PEDIATRICS
2
Kolho et al.
inflammation, although the specific role of different bacteria or alterations in bacterial diversity is unresolved (7). Antibiotics modify the intestinal microbiota, notably in early life (8) and there is evidence that antibiotic use associates with the development of CD (9,10). It is well established that the pediatric form of IBD has a more aggressive disease course than IBD in adults. This is reflected in the high use of corticosteroids and in the high frequency of surgeries (11,12). The diagnostic evaluation includes ileocolonoscopy but for monitoring disease activity the use of endoscopies is restricted in pediatric patients due to their invasiveness. Therefore, fecal surrogate markers of inflammation and especially fecal calprotectin are used more and more in clinical practice in the follow-up of IBD patients. Recently, for calprotectin a negative predictive value of 0.95 (95% confidence interval 0.86–0.99) with a cutoff of 100 μ g/g for the presence of intestinal inflammation was reported in pediatric IBD (13) being in line with adult data (14). The recent guidelines for treating pediatric CD (11) or UC (12) are mainly based on the data extrapolated from adult patients. In moderate to severe CD, and comparable to adults recently also in UC, the use of the tumor necrosis factor-α (TNF-α ) antagonist agent infliximab is approved for children older than 6 years of age and restricted to hospital use. The other TNF-α antagonist agent, adalimumab, has been used for study purposes (15,16), but only recently approved for clinical use in pediatric CD. Thus, the use of TNF-α blockers in pediatric IBD is likely to further increase in the future. A considerable proportion of the patients (almost every other patient), however, lose their therapeutic response and need dose escalation already during the first year of therapy (17–19). Presently, the loss of therapeutic effect cannot be foreseen and this strongly affects the costs associated with the use of these biologicals. Several studies have addressed the intestinal microbiota of IBD patients by using 16S rRNA-based high throughput approaches, including next-generation sequencing and phylogenetic microarray analysis (20). Decreased bacterial diversity and changes in the abundance of certain bacterial groups have been observed in the fecal samples and colonic biopsies of IBD patients, in comparison with the samples of healthy subjects in studies with unrelated subjects, monozygotic twins, and multi-centered studies (7,21–23). Several bacterial groups have been associated with IBD, including Escherichia and Bacteroides spp. as well as bacteria belonging to the Clostridiales. Faecalibacterium prausnitzii, an abundant intestinal bacterium belonging to the Clostridium cluster IV, was found to be frequently reduced in abundance in CD, and a reduction in the mucus-degrader Akkermansia muciniphila has been reported in CD and UC (22,24–26). A recent study related the gut microbiota to therapeutic responses and an association was described between fecal microbiota and the response to steroid therapy in pediatric patients hospitalized for severe UC (27). This suggests that characterization of gut microbiota may provide an additional tool in assessment of therapeutic responses warranting further studies. We hypothesized that alterations in gut microbiota associate with the level of The American Journal of GASTROENTEROLOGY
inflammation in pediatric IBD. Hence, we conducted a prospective study where we analyzed the fecal microbiota in pediatric IBD with special reference to its association with disease activity and therapeutic responses to the TNF-α antagonist agents, both related to fecal calprotectin values. The results using deep phylogenetic microarray analysis indicate that fecal microbiota tightly associated with the level of inflammation (reflected in fecal calprotectin). Intriguingly, responders and non-responders to anti-TNF-α therapy could be distinguished based on their fecal microbiota composition, and the response associated with simultaneous changes in microbial diversity and composition.
METHODS We invited 9- to 18-year-old Finnish-speaking patients with IBD followed up at Children’s Hospital, Helsinki, Finland to this prospective study on fecal microbiota performed between June 2011 and January 2013. The families of patients scheduled for pediatric gastroenterologist, colonoscopy, or antiTNF-α therapy were approached by a letter or contacted by the IBD study nurse describing the study protocol and requesting stool and blood samples at their next regular visit. All IBD diagnoses were endoscopy based and fulfilling the revised Porto criteria (28). All patients were requested to provide two stool samples for study purposes: one for the measurement of fecal calprotectin and one for microarray analyses. Samples were frozen at −20 °C within 4 h of defecation and stored at −70 °C until analyzed (see below). Those who were entering anti-TNF-α therapy were also requested to bring additional stool samples at weeks 2 and 6 after the start of the therapy and once during the maintenance therapy. Controls were non-IBD patients either scheduled for checkup for juvenile idiopathic arthritis or healthy children recruited from the Helsinki region that were recruited the same way as the patients. The study groups comprised 68 patients with IBD and 26 controls, of whom 18 presented with juvenile idiopathic arthritis (Table 1). The number of patients on TNF-α blockers was 32 (on infliximab n=31; on adalimumab n=1; diagnosis Crohn’s n=27; UC/IBDU=5). Of these 11 were introduced to the therapy during the study period (at baseline on glucocorticoids and/or azathioprine). The other patients were medicated with 5-aminosalicylic acid (n=23) and/or azathioprine (n=9), methotrexate (n=4), glucocorticoids (n=2), and none (n=6). All control children had low levels of fecal calprotectin (3-fold or normalization) in fecal calprotectin values and mild or no clinical disease activity (31). The fecal calprotectin level was measured in a routine clinical laboratory using a quantitative enzyme immunoassay (PhiCal Test, Calpro AS, Oslo, Norway), and the values of 98% Pearson similarity) were used for further analysis. In total, 280 microarray experiments were conducted in this study; two samples were of insufficient quality for the analyses.
Extensive colitis
15
Left-sided
11
Statistical analyses
6
All statistical analyses were conducted with the program R (37), using the packages vegan (38), gstat (39), nlme (40), and pROC (41). The microbiota data were converted into relative abundances. Diversity of the microbiota was calculated as the inverse Simpson diversity index, using the oligonucleotide-probe level data. Richness was calculated as the number of oligonucleotide probes with signal intensities higher than the median intensity over the whole data set. Similarity to control microbiota was calculated as the average correlation between the log-transformed oligonucleotide probe level data from the patient and the healthy control samples. Principal coordinate analysis was performed with the first samples of all individuals, using the log-transformed species-level data and Bray-Curtis distances (function “capscale” in the R-package vegan). The calprotectin, microbiota richness, abundance of groups associated with butyrate production, and the ratio between Grampositive and Gram-negative organisms in the principal coordinate space were calculated by inverse distance-weighted interpolation (function “idw” in the R-package gstat). The microbiota differences between the IBD and control cases were assessed using the genus-level data set and linear mixed effect models with individual as the random factor (function “lme” in the R-package nlme), since many individuals provided several samples. For each genus-level taxon, we fitted a model estimating the effect of diagnosis group (UC, CD, or control) on the relative abundance of the taxon. To study the association between the microbiota composition and calprotectin level, the calprotectin levels were modeled
Unclassified IBD (IBDU) Disease duration years, median (range)
3.5 (0–10.7)
Anti-TNF-α medication (no. of patients)
32 d
Introduction of therapy during the study period
11
Clinical disease activity Physician Global Assessment, median (range)
1 (1–3)
Remission according to PCDAI (Pediatric Crohn’s Disease Activity Index)
30/36
PUCAI (Pediatric Ulcerative Colitis Activity index)
17/32
a
Difference in the proportion of males, P=0.28 (proportion test). Difference in age between patients and controls P
