Metaproteomics Approach To Study the Functionality of the Microbiota in the Human Infant Gastrointestinal Tract

A metaproteomics approach comprising two-dimensional gel electrophoresisand matrix-assisted laser desorption ionization-time of flight(mass spectrometry) was applied to the largely uncultured infantfecal microbiota for the first time. The fecal microbial metaproteomeprofiles changed over time, and one protein spot contained apeptide sequence that showed high similarity to those of bifidobacterialtransaldolases.

INTRODUCTION

Top

Introduction

The human gastrointestinal tract is rapidly colonized duringthe first days of life by microbes (4, 10) which ultimatelycontribute significantly to host nutrition and immunity amongother beneficial effects (1, 8). Analysis of 16S ribosomal DNAclone libraries revealed numerous undescribed species in intestinalsamples, emphasizing the importance of techniques that bypasscultivation of microbiota (3, 4, 15, 17, 20) such as proteomics.So far there have been two reports of the use of metaproteomicsto characterize complex bacterial ecosystems, illustrating thefeasibility of this approach in analyzing complex ecosystems(12, 19). In the present study we investigated the potentialuse of metaproteomics for characterization of the human fecalmicrobiota.

Dynamics of the intestinal microbiota.

Top

Dynamics of the intestinal...

The relatively simple infant fecal ecosystem was chosen forthis study as it is dominated by bifidobacteria, which wereused in this study to optimize the protein isolation and two-dimensional(2D) gel procedures (data not shown). Informed consent was obtainedfrom parents for use of the fecal samples of their infants.Infant fecal samples containing high bifidobacterial contentwere selected by bifidobacterium-specific fluorescent in situhybridization (6, 22). Infants A (8 days old) and B (117 daysold) harbored 45% and 63% bifidobacteria in the total microbiota,respectively. PCR-denaturing gradient gel electrophoresis (DGGE)(14, 21, 23) of the 16S rRNA gene was performed to monitor thetotal bacterial community (Fig. 1). The profiles were relativelysimple (Fig. 1) (4), as expected, and the predominance of bifidobacteriawas supported by the presence of abundant 16S rRNA gene ampliconsthat comigrated with the control Bifidobacterium longum strainand was confirmed by sequencing (14).

View larger version (123K):
[in this window]
[in a new window]

FIG. 1. PCR-DGGE profiles of regions V8 to V6 of 16S rRNA genes of fecal samples from two infants at different ages (infant A, 8 to 41 days; infant B, 128 to 191 days), reflecting the predominant bacterial community. The bands in the box represent amplicons of bifidobacterial species, including Bifidobacterium longum DSM 20219. *, start of weaning period for infant A. Sampling days for protein extraction are shown in bold. marker, mix of DNA size standards; bifspecific product, amplicon of 16S rRNA gene of B. longum DSM 20219.

Metaproteome production of infant fecal microbiota.

Top

Metaproteome production of...

Fecal samples were collected from the infants prior to weaning:for infant A, at days 8, 24, and 41, and for infant B, at days103, 117, and 144. Microbial cells were released from the fecesand washed as previously described (5, 22). The bead beatingmethod was confirmed to be applicable for protein extractionof infant fecal microbiota containing bifidobacteria (data notshown), as expected from previous systematic studies (5, 14,20, 23, 24). Total soluble protein was obtained by three treatmentsof 45 s of bead beating (FastPrep; Qbiogene), interspersed by1 min on ice, in 500-µl isoelectric focusing (IEF) buffer(Fluka, Switzerland) (10 M urea) containing 2% CHAPS (3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate)(Roche, Switzerland), 0.65 mM dithiothreitol (DTT) (Sigma, Switzerland),0.2% Biolytes 3/10 (Bio-Rad), Pefabloc Sc (Fluka), and glassbeads (Sigma) ( $≤$ 0.1 mm). Protein concentrations were determinedusing a 2-D Quant kit (Amersham Biosciences). Each protein sample(100 µg) was loaded onto 11-cm immobilized pH gradient(IPG) ReadyStrips (Bio-Rad) (pH 4 to 7) and rehydrated for 12h. IEF was carried out for 98,000 Vh (Protean IEF; Bio-Rad).Reduction and alkylation of proteins were performed prior toelectrophoresis in the second dimension by incubating for 10min in 6 M Urea-0.1 M Tris-HCl (pH 8.8)-2% (wt/vol) sodium dodecylsulfate (SDS) with 130 mM DTT followed by a 10 min incubationin the presence of 216 mM iodoacetamide (instead of DTT). TheIPG strip was positioned on an SDS-polyacrylamide electrophoresisgel (Criterion gel, Bio-Rad) (12.5% polyacrylamide) with 1%low melting agarose in 40 mM Tris-HCl (pH 6.8). Electrophoresiswas run at 100 V and 30 W. Silver staining was performed asdescribed previously (16). Protein maps were scanned with aGS-800 densitometer (Bio-Rad) and analyzed with PDQuest software(Bio-Rad). Triplicate 2D gels for each sample were grouped.For each infant, three groups, each representing one time pointin life, were compared using the quantitative function withinPDQuest software. The mean coefficient of variation (CV) (standarddeviation/mean x 100) is a quantitative index for variationof quantities among matched spots and was computed for gel-to-gelvariations within each replicate group. More than 200 proteinspots were visualized on each gel. A comparison of gels of infantA revealed changes in number and intensities of protein spotsduring the 33 days, although the patterns remained similar (Fig.2A to C). In order to clearly demonstrate the changes a sectionof the gel corresponding to each time point was enlarged (Fig.3). Comparisons of differential protein production values withpercent CV values for various time points for the marked spotsare illustrated with bar graphs (Fig. 3C). The metaproteomeprofiles of infant A (Fig. 2A to C) were different from thoseof infant B (see, e.g., Fig. 2D) in accordance with the uniquenessof each individual's microbiota (20), as was also observed inthe PCR-DGGE profiles obtained as described above.

View larger version (71K):
[in this window]
[in a new window]

FIG. 2. Silver-stained 2D gels of fecal samples from infant A at the ages of 8 days (A), 24 days (B), and 41 days (C) and from infant B at 117 days (D); data were obtained using IPG (pH 4 to 7). MW, molecular weight.

View larger version (44K):
[in this window]
[in a new window]

FIG. 3. (A) Silver-stained 2D gel of a fecal sample from infant A at the age of 41 days obtained using IPG (pH 4 to 7). The circled protein spots were excised, and mass fingerprints were obtained by MALDI-TOF (MS). Table 1 shows the quantity of the selected protein spots. MW, molecular weight. (B) Areas of interest enclosed in the boxes are enlarged from gels of samples from infant A at the ages of 8 (I), 24 (II), and 41 (III) days. The circled protein spots (1, 2, 7, and 11) gave good peptide sequencing results with MALDI-TOF (MS). (C) The diagrams demonstrate changes in intensity of the protein spots (1, 2, 7, and 11) (see panel BI) with percent CV values at ages 8, 24, and 41 days.

Identification of proteins.

Top

Identification of proteins.

A total of 55 protein spots were excised from the infant A 2Dgel at day 41 (Fig. 3A). Peptides were extracted after trypticdigestion and analyzed by matrix-assisted laser desorption ionization-timeof flight (mass spectrometry) [MALDI-TOF (MS)] (2) (AppliedBiosystems 4700 proteomics analyzer; Technology Facility, Departmentof Biology, The University of York, York, United Kingdom), whichprovided a catalogue of 21 good mass peptide fingerprints. Peptidesequences were searched using BLASTP, with the default settingsof the "Search for short, nearly exact matches" function, aswell as MS-Pattern in Protein Prospector 4.0.5 (The Universityof California). As expected, the peptides' mass spectra showedlow similarity to those of database entries and no similaritiesto those of human proteins. The mean differential protein amountsof the 21 spots with clean signals for days 8, 24, and 41 areindicated in Fig. 3A and presented in Table 1. MALDI-TOF (MS)of spots 1, 2, 7, and 11 was performed for de novo peptide sequencing(Fig. 3B), which resulted in determination of 11 N-terminalsequences that shared similarity to those of bacterial proteins,a viral protein, and four eukaryal proteins, but the collectivesequences were not significantly similar to any previously reported.

View this table:
[in this window]
[in a new window]

TABLE 1. Mean quantity of selected protein spots over time with % CV and experimental pI and MW values

One complete peptide fragment (ELAEATDFVDGR) of protein spot4 gave a result showing high-level database matches of 91% identity(i.e., one mismatch) with the Bifidobacterium infantis and B.longum NCC2705 transaldolases and 83% identity with the DJO10Atransaldolase. Production for this protein spot was relativelyhigh at day 8 but increased approximately 10-fold and 6-foldat days 24 and 41, respectively, possibly due to the increasein numbers and activity of bifidobacteria in the infants' microbiota,based on PCR-DGGE (Table 1 and Fig. 1). The data obtained forthe digestion sites of trypsin and the theoretical pI (4.87)and molecular mass (39.6 kDa) of the B. infantis transaldolaseprotein agreed with the position of the protein spot in the2D gel (Fig. 3A). This transaldolase gene is a common targetfor PCR used to detect and enumerate bifidobacteria (13). Transaldolasewas identified in a proteomic study by Vitali et al. (18) ofB. infantis B107, where it represented, together with nine otherproteins, the most abundant portion of the proteome. The dominanceof the (tentatively identified) bifidobacterial transaldolaseprotein in the feces of a newborn infant may explain its detectionby a metaproteomics approach.

Perspectives.

Top

Perspectives.

For the first time, reproducible 2D gels, extraction of proteins,and tentative identification using MALDI-TOF (MS) demonstratedthe applicability of the proteomics approach for the complexintestinal ecosystem. Currently insufficient microbiome sequenceinformation confounds identification of the proteins, but ongoingmetagenomic library analysis will enable meaningful identificationin time (7, 9, 11). Furthermore, peptide mass fingerprints fromsequence data will be sufficient to produce more statisticallyvalid database matches, as was recently demonstrated by Ramet al. (12), who matched 6,000 peptide fragments to DNA sequencesof an accompanying metagenomic library from a low-complexitynatural microbial biofilm. Metaproteomics approaches may becomea useful tool to monitor the functional products of the microbiotain feces over time as affected by dietary intervention, lengthlife, health, and disease.

Nucleotide sequence accession numbers and peptide sequences.

Top

Nucleotide sequence accession...

The bifidobacterial nucleotide sequences of partial 16S rRNAgenes have been deposited in the GenBank database under accessionno. DQ323457, DQ323458, and DQ323459; the 11 peptide sequencesare DLAVALSENKR, ATNSELMHVGVSR, ELAEATDFVDGR, and PSSKVGSGSSGAGALK(for spot 1), DVAPDLALMHTKLSR (for spot 2), ADNFEGDDR and TAFTGYETLR(for spot 7), and KTGPKLFAADEALK, HYGLASDALANGGCVDSVSDSPA, APMVALSELER,and TANSELLEAELAR (for spot 11).

ACKNOWLEDGMENTS

This work was supported by the Dutch Ministry of Economic Affairsthrough the Innovation Oriental Research Program on Genomics(IOP Genomics: IGE01016).

FOOTNOTES

* Corresponding author. Mailing address: Laboratory of Microbiology, Wageningen University, Hesselink van Suchtelenweg 4, 6703 CT Wageningen, The Netherlands. Phone: 31 317 483486. Fax: 31 317 483829. E-mail: Eline.klaassens{at}wur.nl.

Published ahead of print on 8 December 2006.

REFERENCES

Top