Development of an enhanced metaproteomic approach for deepening the microbiome characterization of the human infant gut

Abstract

The establishment of early life microbiota in the human infant gut is highly variable and plays a crucial role in host nutrient availability/uptake and maturation of immunity. Although high-performance mass spectrometry (MS)-based metaproteomics is a powerful method for the functional characterization of complex microbial communities, the acquisition of comprehensive metaproteomic information in human fecal samples is inhibited by the presence of abundant human proteins. To alleviate this restriction, we have designed a novel metaproteomic strategy based on double filtering (DF) the raw samples, a method that fractionates microbial from human cells to enhance microbial protein identification and characterization in complex fecal samples from healthy premature infants. This method dramatically improved the overall depth of infant gut proteome measurement, with an increase in the number of identified low-abundance proteins and a greater than 2-fold improvement in microbial protein identification and quantification. This enhancement of proteome measurement depth enabled a more extensive microbiome comparison between infants by not only increasing the confidence of identified microbial functional categories but also revealing previously undetected categories.

Keywords: Metaproteome; double filtering; human infant gut; shotgun proteomics.

Figure 1

Workflow of the indirect double filtering (DF) method. Fecal raw material is suspended in cold PBS and passed through a 20 μm filter to remove large particles and intact human cells. The filtrate is homogenized and centrifuged to obtain a microbial cell pellet. The pellet is resuspended and passed through a 0.22 μm filter to collect microbial cells on the filter membrane. Collected cells are washed twice and subjected to SDS-based cell lysis and protein purification.

Figure 2

Rank–abundance plots of protein groups. Identified proteins are clustered into protein groups, and their spectral counts are balanced and normalized according to the approach specified in the Materials and Methods. Protein groups of (a) infant #UN1 and (b) infant #CA1 are ranked and plotted based on spectral counts. The indirect DF method facilitates an increasing number of identified protein groups. The two methods possess the same slope for top ranked groups but diverge at the group with fewer than 100 spectral counts. The indirect DF method has a shallower slope and thus provides more low-abundance protein group identifications.

Figure 3

Distributions of ScanRanker scores for collected mass spectra. ScanRanker scores are used to assess spectral quality for all collected mass spectra. Stack histograms are generated for ScanRanker scores of (a) infant #UN1 measured by the direct method, (b) infant #CA1 by the direct method, (c) infant #UN1 by the indirect DF method, and (d) infant #CA1 by the indirect DF method. The color denotes ScanRanker score distributions of unassigned (gray), assigned human (red), and assigned microbial (green) mass spectra in replicates. The indirect DF method enriches microbial mass spectra assignment as decreasing human mass spectra assignment.

Figure 4

Comparison of protein group identification and quantification results by two methods. The Venn diagram (a) shows unique and overlapped protein group identifications of infant #UN1 between the direct and indirect DF methods. Bar charts indicated human (red) versus microbial (green) protein group counts and spectral counts in the part of uniquely identified by the direct method (left), commonly identified (bottom), and uniquely identified by the indirect DF method (right). Scatter plots are constructed using log₂ spectral counts of microbial protein groups measured by two methods for infant #UN1 (b, r_s = 0.76) and infant #CA1 (c, r_s = 0.77). Solid line indicates a perfect correlation, and the dashed line indicates the offset owing to microbial protein enrichment. Microbial protein groups are enriched with a relatively high ranked correlation.

Figure 5

COG category analysis of microbial protein groups. Microbial protein groups are assigned into COG categories via rpsblast against the COG database from NCBI. Distributions of identified categories were constructed by category counts and spectra of infant #UN1 (a) and infant #CA1 (b). Abundant categories are numerically labeled.