Rationale: Many bacterial pathogens causing respiratory infections in children are common residents of the respiratory tract. Insight into bacterial colonization patterns and microbiota stability at a young age might elucidate healthy or susceptible conditions for development of respiratory disease.
Objectives: To study bacterial succession of the respiratory microbiota in the first 2 years of life and its relation to respiratory health characteristics.
Methods: Upper respiratory microbiota profiles of 60 healthy children at the ages of 1.5, 6, 12, and 24 months were characterized by 16S-based pyrosequencing. We determined consecutive microbiota profiles by machine-learning algorithms and validated the findings cross-sectionally in an additional cohort of 140 children per age group.
Measurements and Main Results: Overall, we identified eight distinct microbiota profiles in the upper respiratory tract of healthy infants. Profiles could already be identified at 1.5 months of age and were associated with microbiota stability and change over the first 2 years of life. More stable patterns were marked by early presence and high abundance of Moraxella and Corynebacterium/Dolosigranulum and were positively associated with breastfeeding in the first period of life and with lower rates of parental-reported respiratory infections in the consecutive periods. Less stable profiles were marked by high abundance of Haemophilus or Streptococcus.
Conclusions: These findings provide novel insights into microbial succession in the respiratory tract in infancy and link early-life profiles to microbiota stability and respiratory health characteristics. New prospective studies should elucidate potential implications of our findings for early diagnosis and prevention of respiratory infections.
Clinical trial registered with www.clinicaltrials.gov (NCT00189020).
Microbial colonization begins directly after birth and is followed by progressive assembly of species into a complex and dynamic microbial community. Although microbiota development during this early period in life may be very important for respiratory health, little is known yet about upper respiratory microbiota dynamics and their relation to susceptibility to respiratory infections.
This study shows that from 1.5 months on, distinct microbiota profiles can already be identified in the respiratory tract of healthy infants and that these early microbiota profiles correlate to microbiota stability and change over time. Moraxella and Corynebacterium/Dolosigranulum-dominated communities were more stable over the first 2 years of life and were associated with lower rates of parental-reported respiratory infections.
Children are prone to development of respiratory infections due to an immature immune system. Consequently, they have high colonization rates of potential pathogens (pathobionts) in the respiratory tract, which is an important prerequisite for infection (1). From the nasopharyngeal site, pathobionts can spread directly to cause otitis media or pneumonia or invade the bloodstream to cause sepsis or meningitis. Most studies of microbial and respiratory infections have typically focused on colonization patterns of individual pathobionts in childhood in cross-sectional settings (2, 3). It remains unclear, however, what mechanisms define if carriers of these pathobionts progress toward disease. Several studies have investigated the role of genetic susceptibility, immune mechanisms, and environmental factors as contributors to development of disease (4, 5).
Recent work has shown the importance of the human microbiota for our health. Commensals play an important role in immune maturation, enhancement of the mucosal barrier function, and colonization resistance for pathogens (6, 7). Given the important functions of colonizing microbes, establishment of this symbiotic relation between host and microbiota in early life might be crucial for optimal health later in life (8). Subsequently, changes in bacterial community characteristics, and changes in their functional ability are increasingly associated with diseases (7, 9).
Nothing is known yet about nasopharyngeal microbiota dynamics and stability over time, its relation with susceptibility to respiratory infections, and the influence of environmental exposures on the development of this community. We therefore studied the respiratory microbiota dynamics in 60 children at 1.5, 6, 12, and 24 months of age in relation to host characteristics and environmental exposures. Cross-sectional findings were validated in an additional 140 children per age group. Our goals were to identify microbiota profiles in young children and to link these profiles to microbial change or stability over time and respiratory health characteristics.
Upper respiratory (nasopharyngeal) samples were obtained from a randomized controlled trial that studied the efficacy of reduced-dose schedules of seven-valent pneumococcal conjugate vaccine (PCV-7) on nasopharyngeal pneumococcal carriage in healthy children in the Netherlands (10). For the current study, we selected children who were born between October 2005 and January 2006 and analyzed samples from 60 children obtained at 1.5, 6, 12, and 24 months of age.
In addition, per time point approximately 140 additional children were selected for whom we did not have all four consecutive samples analyzed as validation cohort for biomarker species analyses (see later).
Detailed metadata were obtained by questionnaires, and all children were considered healthy and nonfebrile during sampling moments. We did ask parents retrospectively how many (mild) respiratory tract infections had occurred during the time between sampling moments.
Apart from PCV-7, all children received vaccinations against diphtheria, tetanus, pertussis, polio, and Haemophilus influenzae type B at 2, 3, 4, and 11 months of age and against rubella, mumps, measles, and Neisseria meningitidis group C at 14 months of age, according to the Dutch National Immunization program. The randomized controlled trial (NCT00189020) was approved by an acknowledged Dutch National Ethics Committee (Stichting Therapeutische Evaluatie Geneesmiddelen) and undertaken in accordance with European Statements for Good Clinical Practice, which included the provisions of the Declaration of Helsinki of 1989. Before enrollment, written informed consent was obtained from both parents of each participant.
Eight hundred samples were processed for 454 GS-FLX-Titanium sequencing of the 16S small subunit ribosomal DNA gene (11). In short, bacterial DNA was extracted using a phenol/bead-beating and a magnetic bead separation method. Quantity of bacterial DNA was measured for each sample using real-time polymerase chain reaction and universal primer-probe sets targeting 16S rDNA gene. An amplicon library was generated by amplification of the V5–V7 hypervariable region of this gene. The amplicon library was sequenced unidirectionally, and sequences were processed and classified using modules implemented in Mothur V.1.20.0 software platform (12). In short, sequences were de-noised, trimmed, and checked for quality and chimeras. The remaining high-quality sequences were aligned and classified using the RDP-II naive Bayesian classifier (13). Aligned sequences were subsequently clustered into operational taxonomic units (OTUs, defined by 97% similarity) using the average linkage clustering method. For each of the samples, rarefaction curves were plotted and community diversity parameters calculated. The Shannon diversity index is a diversity measure that takes into account both relative abundance and evenness of species in a sample. Good’s coverage was calculated for every sample to determine sufficiency of sequence depth. For all samples, we calculated the presence and relative abundance of all OTUs.
We used a coregularized spectral clustering algorithm (14, 15) to analyze changes in consecutive respiratory microbiota over time in 60 healthy children. The method stems from a recently proposed class of multiview clustering algorithms that have been reported to notably outperform standard techniques (e.g., k-means, hierarchical clustering, etc.) in clustering accuracy and stability. Multiview algorithms (closely related to cluster ensembles  and consensus techniques ) aim to combine multiple clustering hypotheses for increased accuracy and are not limited to a single similarity measure, thus leading to robust and reliable results (18–20). Furthermore, our algorithm allows identification of the optimal number of clusters via construction of co-occurrence matrices and probabilistic cluster assignments (see online supplement) (21). Next to the optimal number of clusters, our method also allows detection of biomarker species by using greedy forward feature subset selection approach (22). In all statistical analyses we used a normalized microbial abundance OTU matrix as obtained after processing sequence data using the Mothur package (12). More detailed information about the clustering and biomarker selection approach can be found in the online supplement.
Furthermore, to study dynamics of microbial diversity in healthy children we computed relative change in microbiota composition over time. Formally, the relative change is the difference in microbial abundance, for example, the difference in microbial abundance between 1.5 and 6 months divided by the microbial abundance at 1.5 months, computed per OTU. We considered consecutive intervals 1.5 to 6 months, 6 months to 12 months, and 12 months to 24 months. To study groups of children who exhibit similar patterns of microbial change over time, we applied our algorithm to these newly constructed datasets. Per consecutive interval and over all intervals we also computed magnitude of change by using L2 norm on rows of the relative change matrix.
To validate whether our clustering model captures the information contained in the dataset and additional metadata we randomly permuted its columns and rows given the input data matrix of microbial abundances. This corresponds to the situation when various relative abundances of species are randomly assigned to the children contained in dataset. Once the dataset is permuted, we performed the clustering analysis as described above. We repeated the randomization 1,000 times and recorded the squared error of the models averaged over the clusters for every repetition. We noted that the difference between the permuted and original data is statistically significant (P < 0.001), as none of the 1,000 random permutations had a lower squared error than the original data set.
To validate our findings in a cross-sectional manner, we performed forward selection biomarker algorithms on nasopharyngeal microbiota compositions of 137 to 143 additional children per time point participating in the randomized controlled trial (10). The microbiota of these children were previously analyzed as part of two other studies (23, 24).
Characteristics of the 60 children are listed in Table 1. Number of children born by cesarean section, as well as antibiotic consumption before sampling, was low. We obtained a total of 650,971 high-quality sequences (mean, 2,724; range, 439–7,516 sequences/sample) that could be assigned to 10 different phyla (each with >10 attributed sequences in the overall population) and 314 OTUs (each with ≥3 attributed sequences in the population). Sequence depth was sufficient to obtain high sequence coverage (mean, 99.7%; median, 99.7%; range, 98–100%). We observed no significant difference in Shannon diversity in samples over time on average 0.86 (SD, 0.47) Bacterial load as measured by universal 16S quantitative polymerase chain reaction also did not differ significantly over time (geomean, 13.1 pg/μl; 95% confidence interval, 10.7–16.0 pg/μl).
|1.5 mo||6 mo||12 mo||24 mo|
|Exclusively breastfed||37 (61.7)||30 (50)||NA||NA|
|Allergen-free feeding||4 (6.7)||4 (6.7)||5 (8.3)||0|
|Month of sampling||Oct–Jan||Mar–Jun||Oct–Jan||Sept–Jan|
|Vaginal birth*||51 (92.7)||NA||NA||NA|
|PCV-7 vaccination†||0||30 (50)||30 (50)||30 (50)|
|Presence of siblings in household||38 (63.3)||38 (63.3)||38 (63.3)||41 (68.3)|
|Day care attendance‡||0||26 (43.3)||31 (51.7)||34 (56.7)|
|Smoke exposure||4 (6.7)||5 (8.3)||4 (6.7)||4 (6.7)|
|Eczema||1 (1.7)||9 (15)||6 (10)||4 (6.7)|
|Mild respiratory symptoms§||14 (23.2)||8 (13.3)||24 (40)||20 (33.3)|
|Otitis media||0||1 (1.7)||1 (1.7)||0|
|Previous no. of RTIs = 1||0||11 (18.3)||20 (33.3)||16 (26.7)|
|Previous no. RTIs > 1||—||2 (3.4)||7 (11.7)||8 (13.4)|
|Previous episodes of wheezing||0||7 (11.7)||6 (10)||10 (16.7)|
|Previous use of bronchodilators||0||2 (3.3)||5 (8.3)||8 (13.3)|
|Antibiotics consumption within last mo||||0||0||3 (5)||2 (3.3)|
We used coregularized spectral clustering of the nasopharyngeal microbiota per time point to distinguish specific clusters. For every sample, the probability of belonging to a certain cluster was calculated and enabled us to distinguish eight specific clusters of individuals (4–5 per time point; see Figures E1 and E2 and Table E1 in the online supplement). To determine which species contributed to the clustering, we used a forward selection biomarker algorithm. In total, 10 different biomarker species were observed and depicted in Figure 1. Ten biomarker species were identified: the first- and second-highest ranking Moraxella OTUs (Moraxella 1 and 2), Haemophilus, Streptococcus, Staphylococcus, the first- and second-highest ranking Corynebacterium (Corynebacterium 1 and 2), Dolosigranulum, Prevotella, and Bacteroidetes unclassified.
At 1.5 months of age, we observed clusters mostly dominated by either (1) Streptococcus (STP), (2) Moraxella (MOR), (3) Staphylococcus (STP), (4) Corynebacterium (COR), or (5) Corynebacterium and/or Dolosigranulum (CDG). The Streptococcus-dominated profile was most evenly distributed with additional high abundances of several other genera compared with the other clusters. Interestingly, in the Staphylococcus- and Streptococcus-dominated clusters, the presence and relative abundance of Dolosigranulum and Corynebacterium was lowest. After 1.5 months of age, the Staphylococcus-dominated cluster had disappeared, and a Haemophilus-dominated cluster (HPH) had emerged. Furthermore, the Corynebacterium/Dolosigranulum-dominated cluster was gradually replaced by a Moraxella/Dolosigranulum-dominated cluster (MDG).
Next to the aforementioned profiles, we observed from 6 months on two distinct Moraxella-dominated profiles. To determine more accurately species level, we clustered the unique sequences belonging to these OTUs in a phylogenetic dendrogram with representative sequences of Moraxella species from the RDP database (Figure E3). The sequences of Moraxella 2 were highly (100%) similar to Moraxella lincolnii (ML) whereas Moraxella OTU 1 was not classifiable up to species level, although culture results suggest that this OTU represents Moraxella catarrhalis. Because this could not be confirmed with the 16S sequences, we will still refer to this OTU as Moraxella.
All microbiota profiles of the approximately 140 cross-sectionally per time point sampled individuals were clustered using the previously described coregularized spectral clustering method, and biomarker species were determined using the statistical machine learning biomarker detection algorithm. Eight biomarker species were detected in both the 60 children and the approximately 140 children and are depicted with the relative abundance per cluster in Table E2. We did observe similar clustering between the 60 children and the 140 children. However, in contrast to the cohort of 60 children, in the cohorts of 140 children per time point we identified at 1.5, 6, and 24 months of age 10, 1, and 3 children, respectively, who depicted a microbiota profile different from previously identified clusters. Furthermore, at 1.5 months of age, two children already depicted a Moraxella 2 profile, whereas in the primary cohort this profile was observed only at 6 months of age or later.
At 12 months, coregularized spectral clustering identified a cluster of three individuals who belonged to the staphylococcal-dominated profile, which in the primary study was observed at 1.5 months only. Last, we observed that ML and Moraxella OTU 1 co-occurred more often in the M2 profile than was observed in the 60 children. Overall, profiling of the 60-children samples longitudinally and the 140-children samples cross-sectionally was highly similar, as were the biomarkers detected by the biomarker detection algorithm.
To detect intraindividual changes in relative abundance and presence of microbes over time, we calculated a relative change matrix per OTU per timeframe (1.5–6 mo, 6–12 mo, 12–24 mo). From this relative change matrix per timeframe we calculated the magnitude of microbiota change (norm value) using L2 norm analyses. The higher this norm value, the higher the magnitude of change. We observed the highest and most uniform norm values in children between 1.5 to 6 months (mean, 72.9; SD, 24.4), whereas the highest variation in norm values was observed between 12 and 24 months (mean, 52.1; SD, 31.9), followed by 6 to 12 months (mean, 47.1; SD, 28.8).
To obtain more insight into individual differences in change over time, we ordered individuals based on the cumulative norm values over time (Figure 2). From this cumulative norm, we defined four groups of infants (dotted lines in Figure 2) ordered from low to high change. Subsequently, we calculated the average relative abundance of the 10 biomarker species within these groups (Figure 3). We observed the lowest norm values (group 1; mean, 77.7; SD, 11.5) from 1.5 months on in infants who were colonized with Moraxella. The highest change (group 4; mean norm, 234.9; SD, 30.9) was associated with high(er) abundance of Streptococcus, Prevotella, and Bacteroidetes (unclassified) in the early phases and ML, Streptococcus, and/or Haemophilus in later phases (Figure 3).
We also calculated the directionality (increase or decrease) of change from the relative change matrices. To study groups of children who exhibit similar patterns of microbial change over time, we applied coregularized spectral clustering (Figure E4) to these relative change datasets and were able to identify that the change in microbiota represents specific patterns (Figure E5).
Given these patterns of change between two time points, we decided to follow the profile dynamics of individuals over time (Figure 4). Directionality of microbial change was depicted by a colored line when at least two individuals followed that route of change. Lines were colored according to the cluster the individuals originated from at 1.5 months of age. Lines are dotted gray if the children originated from various clusters, but followed the same change in profile over time.
We observed that 8 of 14 children with an early domination of Moraxella remained in this cluster until 24 months of age, which was reflected in an overall low cumulative norm for children in this cluster (mean norm over time, 121.2; SD, 48.1) compared with all other children (mean norm over time, 186.0; SD, 53.9; Mann-Whitney P < 0.001). Moreover, when an individual enters the Moraxella-dominated cluster at a later time point, he/she is also likely to remain in this cluster (i.e., 76% and 74% of infants in the Moraxella-dominated profile at 6 months and 12 months of age remain in that profile at 12 and 24 months of age, respectively).
The majority of infants dominated by Corynebacterium and Dolosigranulum at 1.5 months of age transferred to the Moraxella-dominated cluster (10/15) and to a lesser extent to the ML–dominated cluster (3/15) at 6 months of age and finally changed to the Moraxella-dominated cluster as well. Of the infants who started in the Streptococcus-dominated cluster at 1.5 months of age, only 5 of 12 and 2 of 12 remained there until 6 and 12 months, respectively, whereas the remaining children transferred randomly toward other profiles. The norm values for children in this cluster were correspondingly high (mean norm, 193.3; SD, 44.6; P = 0.06). Neither infants who started off in the Staphylococcus-dominated cluster at 1.5 months of age nor the children transferring through the Haemophilus-dominated profile had a preference for consecutive profiles. They never persisted there for more than one time point, resulting in a high magnitude of change (mean norm, 201.9; SD, 47.8; P = 0.01).
Using permutation tests and contingency analyses, we observed that exclusive breastfeeding was positively associated with the CDG profile at 1.5 months of age (permutation test, P < 0.001), with 12 of 15 children being breastfed in this group (Table 2).
|Age (mo)||Microbiota Profile||N||Breastfeeding||AB < 1 mo*||Consecutive Period(s) of URTI†|
|1.5||MOR||14||7 (50)||0||3 (21)|
|CDG||15||12 (80)||0||2 (13)|
|STR||12||5 (42)||0||6 (50)|
|STP||14||9 (64)||0||2 (14)|
|6||MOR||26||11 (42)||0||9 (35)|
|CDG||5||3 (60)||0||1 (20)|
|STR||14||8 (57)||0||10 (71)|
|ML||11||8 (73)||0||4 (36)|
|HPH||4||4 (25)||0||3 (75)|
|12||MOR||38||NA||1 (2.6)||11 (29)|
|STR||7||NA||1 (14)||4 (57)|
|HPH||8||NA||1 (12.5)||4 (50)|
Frequencies of parental-reported upper respiratory tract infections (URTI) in the consecutive 6-month period were lower in children with a CDG profile at 1.5 and 6 months of age compared with all other profiles (permutation test, P < 0.001). Similarly, parental-reported URTI frequencies were lower in the consecutive 6-month period in the children with a Moraxella-dominated profile at 1.5, 6, and 12 months of age compared with all other profiles (permutation test, P < 0.001), especially compared with the HPH and STR profiles (Table 2).
To eliminate the possible confounding effect of PCV-7 vaccination on microbiota composition, we studied this in our primary cohort of 60 children. PCV-7 vaccination showed no significant association with any of the microbiota profiles at 6, 12, and 24 months of age.
Our study of a cohort of 60 children is the first to describe the development of respiratory microbiota in healthy children over time. Our results demonstrate that already from 6 weeks of age distinct microbiota profiles exist and that early-life profiles are associated with microbiota stability and patterns of change over the first 2 years of life. More stable microbiota profiles in the first 2 years of life were characterized by an early colonization of Moraxella, and Dolosigranulum combined with Corynebacterium, whereas instability was associated with a trajectory from, through, or toward Haemophilus- and Streptococcus-dominated profiles. Moreover, these patterns of change seem to be associated with infant feeding and frequency of (mild) parental-reported respiratory infections, with especially the stable microbiota profiles being related with a history of breastfeeding and reduced numbers of consecutive respiratory infections.
Previous cross-sectional (25) and longitudinal studies (26, 27) have shown a specific trajectory of succession of gut microbiota patterns in infants and children. In our 2-year follow-up study, we also identified patterns of succession in respiratory microbiota. The high levels of Staphylococcus and Corynebacterium in the nasopharynx at 1.5 months of age might be explained by environmental encounters: both species are common colonizers of the adult human skin (28), whereas breast milk contains significant amounts of Corynebacterium (29). The decrease in Staphylococcus carriage over time in our study population is in line with former culture-based studies (30, 31), and these dynamics might reflect acquisition patterns of other bacteria like H. influenzae (32) and S. pneumoniae (33).
Interestingly, bacterial communities in the gut evolve toward an adult-like configuration within the first 3 years of life (34); the respiratory microbiota composition and diversity in children was at 2 years of age still highly different from adults (35). This might suggest that transition of respiratory microbiota communities toward a more adult-like configuration will take place over a longer period of childhood.
In addition, we observed distinct patterns of change in microbiota profiles over time, where early colonization with a Moraxella or Dolosigranulum/Corynebacterium-dominated profile resulted in more stable microbial colonization over time. Streptococcus- and Haemophilus-dominated profiles were marked by high levels of change and dispersion of profiles over time. Moreover, these patterns illustrate that microbiota composition in early infancy can define colonization trajectories in early childhood.
Moraxella seems to play a key role in this process. In the literature, carriage rates of M. catarrhalis detected by culture vary from 22% (36) to 70% (37), as detected in the original study cohort of 1,003 children. The increase in M. catarrhalis carriage between 1.5 to 6 months of age might be a consequence of increased acquisition rates due to the increased contact with other children, such as occurs after day-care attendance (36). M. catarrhalis is regarded a nosocomial pathogen that can cause mucosal disease like otitis media in children and lower respiratory tract infections (exacerbations) in patients with COPD (38). We, however, observed more stable profiles and documented lower frequency of parental-reported URTIs in children with Moraxella-dominated profiles. These findings might suggest that Moraxella is a cornerstone bacterium rather than a pathogen: stable microbiota profiles in the gut are also associated with health, in contrast to instable or dysbiotic microbiota (7, 9). The fact that almost 100% of children are colonized with Moraxella supports its role as symbiont, with the frequent identification of Moraxella in cultures from polymicrobial infections pointing toward a bystander effect (39–41) rather than true pathogenicity.
We observed two Moraxella profiles that related to different colonization patterns and phylogenetic annotations: ML and (less distinctive) Moraxella catarrhalis. Colonization dynamics and clinical significance of colonization by ML is not known yet in children (42). In light of the different phenotypic characteristics, it is worthwhile to explore the role of these individual species and strains in microbiota health and stability, especially considering their high prevalence and large variation in abundance.
Children with a Corynebacterium/Dolosigranulum-dominated profile at 1.5 months of age proceeded in almost all cases to a Moraxella-dominated profile at 6 months and older. Also, higher presence and abundance of Dolosigranulum and Corynebacterium in Moraxella-dominated profiles suggest that these profiles are preferentially related. The Corynebacterium/Dolosigranulum profile at 1.5 months of age, both in this cohort of 60 infants as well as in a larger cohort of 200 children (24), is strongly associated with receiving breastfeeding as well as with decreased numbers of parental-reported URTIs. Considering the protective effect of breastfeeding on respiratory infections in infancy, the strong correlation between Dolosigranulum and Corynebacterium colonization and breastfeeding and decreased URTIs, and the fact that Dolosigranulum is a lactic acid bacterium, we postulate that this profile might be beneficial for respiratory health. Our findings are supported by the observations of Laufer and colleagues, who also found that both Dolosigranulum and Corynebacterium are associated with a lower risk of acute otitis media in children older than 6 months of age (43).
In contrast, the Streptococcus- and Haemophilus-dominated profiles were associated with a high magnitude of microbial change and absence of clear and/or stable colonization patterns over time. Potentially, the lack of stability explains or is explained by the fact that previous studies have observed associations between the detection of respiratory viruses and the presence of H. influenzae and S. pneumoniae (44, 45). Several mechanisms of interaction seem to underlay these associations, because both viruses and bacteria can enhance each other’s epithelial attachment or pathogenicity (46, 47). In addition, Vissing and colleagues have shown that early neonatal airway colonization with pathogenic bacteria, like S. pneumonia and H. influenzae, is associated with increased risk of pneumonia and bronchiolitis in early life (48). Potentially, the lack of stability enhances the risk of development of respiratory infections, although this hypothesis should be confirmed in larger prospective studies.
We observed a virtual disappearance of the Staphylococcus profile at the age of 6 months. This decrease in Staphylococcus aureus carriage is in line with former cross-sectional culture-based studies (30, 31) and might reflect maturation of the immune system but also acquisition of other bacteria and interspecies competition (49, 50). In literature, both S. pneumoniae and H. influenzae carriage are negatively associated with S. aureus carriage (33), which might explain that a decline of Staphylococcus coincided with increased presence of Haemophilus at 6 months in our study.
Finally, we observed no significant association between PCV-7 immunization and any of the microbiota profiles per time point. Potentially, our study cohort of 60 children is too small to find significant differences in microbiome composition regarding PCV-7 immunization, because in a larger cohort we previously observed significant effects on microbiota composition at 12 months of age (23). These effects mainly concerned bacterial shifts within the Streptococcus-dominated cluster as well as modest shifts toward the species H. influenzae and S. aureus.
Several limitations of our study should be recognized. Our study included only four sampling points in the first 2 years of life. New longitudinal studies with shorter time intervals are desired to further elucidate the succession of the respiratory microbiota in infants and young children. Furthermore, the study design allowed only retrospective and parental-reported clinical information and was not designed to evaluate clinical outcome measures. New prospective studies should be conducted to understand the role of the respiratory microbiota in development of disease and the consequences of perturbations to this community. Because of the size of the population studied and lack of data of other studies, data interpretation should be done with caution and regarded as preliminary. Furthermore, we were unable to analyze the different bacteria on the species level and are therefore limited in the confirmation of their taxonomic origin and phenotypic features. Strength of this study is that we are the first to study development of the microbiota in the respiratory tract in a large group of young children over time. Furthermore, we used robust machine learning algorithms that enabled detection of patterns of change and related those to the metadata.
In conclusion, we observed distinct patterns of nasopharyngeal microbial colonization in healthy children that were from infancy on associated with a more or less stable microbiota composition and susceptibility to respiratory infections. We recommend further studies that elaborate on the clinical impact of these colonization patterns in childhood to exploit its diagnostic and therapeutic potential.
The authors thank Elske J. M. van Gils, Gerwin D. Rodenburg, and the research team who performed the clinical trial. They also acknowledge the laboratory logistics organized and supervised by Jacob Bruin of the Streeklaboratorium Haarlem. They also thank Jordy Coolen and Martien P. M. Caspers for assistance in Mothur and other post-sequencing analyses. The authors thank all study and laboratory staff and collaborating institutes for their dedication to this project, and the children and their families who have made this study possible.
|1.||Bogaert D, De Groot R, Hermans PWM. Streptococcus pneumoniae colonisation: the key to pneumococcal disease. Lancet Infect Dis 2004;4:144–154.|
|2.||García-Rodríguez JA, Fresnadillo Martínez MJ. Dynamics of nasopharyngeal colonization by potential respiratory pathogens. J Antimicrob Chemother 2002;50:59–73.|
|3.||Mackenzie GA, Leach AJ, Carapetis JR, Fisher J, Morris PS. Epidemiology of nasopharyngeal carriage of respiratory bacterial pathogens in children and adults: cross-sectional surveys in a population with high rates of pneumococcal disease. BMC Infect Dis 2010;10:304.|
|4.||Spor A, Koren O, Ley R. Unravelling the effects of the environment and host genotype on the gut microbiome. Nat Rev Microbiol 2011;9:279–290.|
|5.||Benson AK, Kelly SA, Legge R, Ma F, Low SJ, Kim J, Zhang M, Oh PL, Nehrenberg D, Hua K, et al. Individuality in gut microbiota composition is a complex polygenic trait shaped by multiple environmental and host genetic factors. Proc Natl Acad Sci USA 2010;107:18933–18938.|
|6.||Turnbaugh PJ, Ley RE, Hamady M, Fraser-Liggett CM, Knight R, Gordon JI. The human microbiome project. Nature 2007;449:804–810.|
|7.||Cho I, Blaser MJ. The human microbiome: at the interface of health and disease. Nat Rev Genet 2012;13:260–270.|
|8.||Scholtens PAMJ, Oozeer R, Martin R, Amor KB, Knol J. The early settlers: intestinal microbiology in early life. Annu Rev Food Sci Technol 2012;3:425–447.|
|9.||Clemente JC, Ursell LK, Parfrey LW, Knight R. The impact of the gut microbiota on human health: an integrative view. Cell 2012;148:1258–1270.|
|10.||van Gils EJM, Veenhoven RH, Hak E, Rodenburg GD, Bogaert D, Ijzerman EPF, Bruin JP, van Alphen L, Sanders EAM. Effect of reduced-dose schedules with 7-valent pneumococcal conjugate vaccine on nasopharyngeal pneumococcal carriage in children: a randomized controlled trial. JAMA 2009;302:159–167.|
|11.||Biesbroek G, Sanders EAM, Roeselers G, Wang X, Caspers MPM, Trzciński K, Bogaert D, Keijser BJF. Deep sequencing analyses of low density microbial communities: working at the boundary of accurate microbiota detection. PLoS ONE 2012;7:e32942.|
|12.||Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB, Lesniewski RA, Oakley BB, Parks DH, Robinson CJ, et al. Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Appl Environ Microbiol 2009;75:7537–7541.|
|13.||Cole JR, Chai B, Marsh TL, Farris RJ, Wang Q, Kulam SA, Chandra S, McGarrell DM, Schmidt TM, Garrity GM, et al.; Ribosomal Database Project. The Ribosomal Database Project (RDP-II): previewing a new autoaligner that allows regular updates and the new prokaryotic taxonomy. Nucleic Acids Res 2003;31:442–443.|
|14.||Kumar A, Rai P, Daume H. Co-regularized multi-view spectral clustering. Adv Neural Inf Process Syst 2011;24:1413–1421.|
|15.||Tsivtsivadze E, Borgdorff H, van de Wijgert J. Neighborhood co-regularized multi-view spectral clustering of microbiome data. Second IAPR International Workshop, PSL 2013. DOI: 10.1007/978-3-642-40705-5_8.|
|16.||Strehl A, Ghosh J. Cluster ensembles—a knowledge reuse framework for combining multiple partitions. J Mach Learn Res 2003;3:583–617.|
|17.||Punera K, Ghosh J. Consensus-based ensembles of soft clusterings. Appl Artif Intell 2008;22:780–810.|
|18.||Zhou D, Burges C. Spectral clustering and transductive learning with multiple views. Proceedings of the 24th International Conference on Machine Learning 2007;1159–1166.|
|19.||de Sa VR. Spectral clustering with two views. ICML Workshop on Learning with Multiple Views. August 2005, Bonn, Germany.|
|20.||Tang W, Lu Z, Dhillon IS. Clustering with multiple graphs. IEEE International Conference on Data Mining (ICDM). December 2009, Miami, FL.|
|21.||Cornelisse LN, Tsivtsivadze E, Meijer M, Dijkstra TMH, Heskes T, Verhage M. Molecular machines in the synapse: overlapping protein sets control distinct steps in neurosecretion. PLoS Comput Biol 2012;8:e1002450.|
|22.||Guyon I, Elisseeff A. An introduction to variable and feature selection. J Mach Learn Res 2003;3:1157–1182.|
|23.||Biesbroek G, Wang X, Keijser BJ, Eijkemans RM, Trzciński K, Rots NY, Veenhoven RH, Sanders EA, Bogaert D. Seven-valent pneumococcal conjugate vaccine and nasopharyngeal microbiota in healthy children. Emerg Infect Dis 2014;20:201–210.|
|24.||Biesbroek G, Bosch AA, Wang X, Keijser BJ, Veenhoven RH, Sanders EA, Bogaert D. The impact of breastfeeding on nasopharyngeal microbial communities in infants. Am J Respir Crit Care Med 2014;190:298–308.|
|25.||Palmer C, Bik EM, DiGiulio DB, Relman DA, Brown PO. Development of the human infant intestinal microbiota. PLoS Biol 2007;5:e177.|
|26.||Koenig JE, Spor A, Scalfone N, Fricker AD, Stombaugh J, Knight R, Angenent LT, Ley RE. Succession of microbial consortia in the developing infant gut microbiome. Proc Natl Acad Sci USA 2011;108:4578–4585.|
|27.||Madan JC, Koestler DC, Stanton BA, Davidson L, Moulton LA, Housman ML, Moore JH, Guill MF, Morrison HG, Sogin ML, et al. Serial analysis of the gut and respiratory microbiome in cystic fibrosis in infancy: interaction between intestinal and respiratory tracts and impact of nutritional exposures. MBio 2012;3:e00251–12.|
|28.||Grice EA, Segre JA. The skin microbiome. Nat Rev Microbiol 2011;9:244–253.|
|29.||Hunt KM, Foster JA, Forney LJ, Schütte UME, Beck DL, Abdo Z, Fox LK, Williams JE, McGuire MK, McGuire MA. Characterization of the diversity and temporal stability of bacterial communities in human milk. PLoS ONE 2011;6:e21313.|
|30.||van Gils EJM, Hak E, Veenhoven RH, Rodenburg GD, Bogaert D, Bruin JP, van Alphen L, Sanders EAM. Effect of seven-valent pneumococcal conjugate vaccine on Staphylococcus aureus colonisation in a randomised controlled trial. PLoS ONE 2011;6:e20229.|
|31.||Kwambana BA, Barer MR, Bottomley C, Adegbola RA, Antonio M. Early acquisition and high nasopharyngeal co-colonisation by Streptococcus pneumoniae and three respiratory pathogens amongst Gambian new-borns and infants. BMC Infect Dis 2011;11:175.|
|32.||Sá-Leão R, Nunes S, Brito-Avô A, Alves CR, Carriço JA, Saldanha J, Almeida JS, Santos-Sanches I, de Lencastre H. High rates of transmission of and colonization by Streptococcus pneumoniae and Haemophilus influenzae within a day care center revealed in a longitudinal study. J Clin Microbiol 2008;46:225–234.|
|33.||Regev-Yochay G, Dagan R, Raz M, Carmeli Y, Shainberg B, Derazne E, Rahav G, Rubinstein E. Association between carriage of Streptococcus pneumoniae and Staphylococcus aureus in children. JAMA 2004;292:716–720.|
|34.||Yatsunenko T, Rey FE, Manary MJ, Trehan I, Dominguez-Bello MG, Contreras M, Magris M, Hidalgo G, Baldassano RN, Anokhin AP, et al. Human gut microbiome viewed across age and geography. Nature 2012;486:222–227.|
|35.||Yan M, Pamp SJ, Fukuyama J, Hwang PH, Cho D-Y, Holmes S, Relman DA. Nasal microenvironments and interspecific interactions influence nasal microbiota complexity and S. aureus carriage. Cell Host Microbe 2013;14:631–640.|
|36.||Verhaegh SJC, Lebon A, Saarloos JA, Verbrugh HA, Jaddoe VWV, Hofman A, Hays JP, Moll HA, van Belkum A. Determinants of Moraxella catarrhalis colonization in healthy Dutch children during the first 14 months of life. Clin Microbiol Infect 2010;16:992–997.|
|37.||van Gils EJM, Veenhoven RH, Rodenburg GD, Hak E, Sanders EAM. Effect of 7-valent pneumococcal conjugate vaccine on nasopharyngeal carriage with Haemophilus influenzae and Moraxella catarrhalis in a randomized controlled trial. Vaccine 2011;29:7595–7598.|
|38.||Murphy TF, Parameswaran GI. Moraxella catarrhalis, a human respiratory tract pathogen. Clin Infect Dis 2009;49:124–131.|
|39.||Aebi C. Moraxella catarrhalis - pathogen or commensal? Adv Exp Med Biol 2011;697:107–116.|
|40.||Broides A, Dagan R, Greenberg D, Givon-Lavi N, Leibovitz E. Acute otitis media caused by Moraxella catarrhalis: epidemiologic and clinical characteristics. Clin Infect Dis 2009;49:1641–1647.|
|41.||Armbruster CE, Hong W, Pang B, Weimer KED, Juneau RA, Turner J, Swords WE. Indirect pathogenicity of Haemophilus influenzae and Moraxella catarrhalis in polymicrobial otitis media occurs via interspecies quorum signaling. MBio 2010;1:e00102–10.|
|42.||Vandamme P, Gillis M, Vancanneyt M, Hoste B, Kersters K, Falsen E. Moraxella lincolnii sp. nov., isolated from the human respiratory tract, and reevaluation of the taxonomic position of Moraxella osloensis. Int J Syst Bacteriol 1993;43:474–481.|
|43.||Laufer AS, Metlay JP, Gent JF, Fennie KP, Kong Y, Pettigrew MM. Microbial communities of the upper respiratory tract and otitis media in children. MBio 2010;2:e00245–10.|
|44.||van den Bergh MR, Biesbroek G, Rossen JW, de Steenhuijsen Piters WA, Bosch AA, van Gils EJ, Wang X, Boonacker CW, Veenhoven RH, Bruin JP, et al. Associations between pathogens in the upper respiratory tract of young children: interplay between viruses and bacteria. PLoS ONE 2012;7:e47711.|
|45.||Moore HC, Jacoby P, Taylor A, Harnett G, Bowman J, Riley TV, Reuter K, Smith DW, Lehmann D; Kalgoorlie Otitis Media Research Project Team. The interaction between respiratory viruses and pathogenic bacteria in the upper respiratory tract of asymptomatic Aboriginal and non-Aboriginal children. Pediatr Infect Dis J 2010;29:540–545.|
|46.||Kane M, Case LK, Kopaskie K, Kozlova A, MacDearmid C, Chervonsky AV, Golovkina TV. Successful transmission of a retrovirus depends on the commensal microbiota. Science 2011;334:245–249.|
|47.||Bosch AA, Biesbroek G, Trzciński K, Sanders EA, Bogaert D. Viral and bacterial interactions in the upper respiratory tract. PLoS Pathog 2013;9:e1003057.|
|48.||Vissing NH, Chawes BLK, Bisgaard H. Increased risk of pneumonia and bronchiolitis after bacterial colonization of the airways as neonates. Am J Respir Crit Care Med 2013;188:1246–1252.|
|49.||Peacock SJ, Justice A, Griffiths D, de Silva GDI, Kantzanou MN, Crook D, Sleeman K, Day NPJ. Determinants of acquisition and carriage of Staphylococcus aureus in infancy. J Clin Microbiol 2003;41:5718–5725.|
|50.||Chatzakis E, Scoulica E, Papageorgiou N, Maraki S, Samonis G, Galanakis E. Infant colonization by Staphylococcus aureus: role of maternal carriage. Eur J Clin Microbiol Infect Dis 2011;30:1111–1117.|
*These authors contributed equally to this work.
Supported by The Netherlands Organization for Scientific Research through NWO-VENI grant 91610121 and ZonMw grant 91209010. The randomized controlled trial (ClinicalTrials.gov NCT00189020) was financed by the Dutch Ministry of Health.
The sponsors had no role in study design, data collection, analysis and interpretation, or writing the report. The corresponding author had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Author Contributions: Conceived and designed the experiments: G.B., E.T., E.A.M.S., R.M., D.B., B.J.F.K. Performed the clinical study: R.H.V., E.A.M.S. Performed the experiments: G.B., B.J.F.K., D.B. Analyzed the data: G.B., E.T., B.J.F.K., D.B. Wrote the paper: G.B., E.T., E.A.M.S., B.J.F.K., D.B. Critical review of the paper: all authors.
This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org
Originally Published in Press as DOI: 10.1164/rccm.201407-1240OC on October 20, 2014