Page 19 - Read Online

Li et al. Microbiome Res Rep 2024;3:26 https://dx.doi.org/10.20517/mrr.2023.57 Page 13 of 16

RapidAIM protocol; (B) PC1 vs. PC2; (C) PC2 vs. PC3; (D) Hierarchical clustering of samples showing 72-hour samples of GutAlive
differentiated from 72-hour samples of our in-house buffer; (E) Comparison of the responses of 72-hour samples to kestose between the
(in-house) Buffer group and GutAlive group, Euler plot (area proportional Venn plot) showing numbers of significantly increased COGs
by t-test (P values-adjusted by FDR); (F) Pathways corresponding to significantly increased COGs in response to kestose. COGs only in
the GutAlive group (blue lines), only in the Buffer group (red lines), and shared responses (black lines) are shown.

were processed on the day of collection, the 72-hour sample showed a separation of the metaproteomics
profiles from the other groups, indicating a possible change in microbiome functionality that is specific to
kestose uptake during the storage [Figure 3B-D]. PERMANOVA analysis showed that the storage period
has a more significant impact in the GutAlive than in the in-house buffer group [Supplementary Tables 5
and 6]. Therefore, we performed differential protein abundance analyses between kestose- and blank
control- group samples of each preservation buffer, and we annotated proteins with COG and examined
pathway responses. We observed that the different preservation methods could influence different numbers
of COGs that exhibited a significant increase in the presence of kestose [Figure 3E], and the in-house buffer
is sensitive in observing more responded COGs. Next, we mapped the responded COGs to microbial
metabolism pathway maps in iPATH and found that in-house buffer storage method can better capture
microbial metabolic pathway responses [Figure 3F]. The result suggests that storage of samples in our in-
house buffer at 4 °C for 72 h prior to culturing/biobanking does not affect functional responses.

DISCUSSION
There have been various in vitro models to evaluate microbiome responses. Early in vitro gut microbiome
models were based on large-scale bioreactors that are low-throughput and, due to the large volume of
cultures, very costly owing to the considerable amount of compounds added. More recent advances in
modeling the gut ecosystem include realizing the culturing of complex human gut microbiome in anaerobic
intestine-on-a-chip models, enabling the observation of host-microbiome interactions . However, for the
[10]
purpose of high-throughput compound screening, these models are not easily adaptable. This study
describes the most recently optimized 2.0 version of RapidAIM, which consists of extensive details on stool
sample collection, biobanking, in vitro culturing and stimulation, microbiome sample processing,
metaproteomics measurement and data analysis. Using RapidAIM 2.0, we show consistent responses of
individual microbiomes to prebiotic kestose across five different biobanking workflows; we also show that
kestose had consistent functional effects across individuals and can be used as a positive control in the assay.

In addition to the recommendations described in the protocol in the Method section, we recommend the
following considerations for experimental design:

(1) Plate layout. The experimental design will be performed based on a 96-well format. Taking into
TM
consideration the use of TMT11plex , we recommend that an 8 rows × 10 columns plate layout is used for
each 96-well plate. The first column will later be used for the TMT reference sample, which will be
generated after the desalting step. The last column will be left blank throughout the experiment.

(2) Randomization. Compound treatments across all assay plates should be randomized. We provide the
[39]
“96-well plate randomizer” tool in our iMetaLab Suite to assist researchers with the study randomization
(https://shiny.imetalab.ca/96_well_randomizer/). Randomizing within and across sample plates will be
helpful to detect batch effects between plates, if any, and meet the criteria to apply batch removal tools .
[40]
Samples should be randomized again prior to LC-MS/MS analysis.