Page 12 - Read Online

Lai et al. Microbiome Res Rep 2024;3:21 https://dx.doi.org/10.20517/mrr.2023.76 Page 5 of 15

(Bioo Scientific, Austin, TX, USA). Illumina NovaSeq 6000 (Illumina Inc., San Diego, CA, USA) from
Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) was used for paired-end sequencing.
Experiments were performed in triplicate, with samples 1, 2, and 3 representing the biological replicates.

Fastp (https://github.com/OpenGene/fastp, version 0.20.0) was used to cut the adapter of the raw reads and
remove low-quality reads (read length less than 50 bp, average base mass value less than 20, or reads contain
N base). Host DNA sequences were removed using the BWA software (http://bio-bwa.sourceforge.net,
version 0.7.9a). The high-quality reads were then assembled into contigs using MEFAHIT (https://github.
com/voutcn/megahit, version 1.1.2), and then the screened contigs (≥ 300 bp) were further used for ORF
prediction [ORF finder (https://www.ncbi.nlm.nih.gov/orffinder/)]. SOAPaligner software (https://help.rc.
ufl.edu/doc/SOAPaligner) was used to calculate gene abundance information. Raw and annotated data for
metagenomics are shown in Supplementary Table 3. The annotation of the amino acid sequence of the non-
redundant gene catalog based on the NCBI NR database was obtained by Diamond (https://v2.
pseudomonas.com/blast/setpdiamond), as well as the calculation of species abundance. In addition, the
corresponding function of genes was revealed by comparing it with the KEGG database (https://www.kegg.
-5
jp/) with an e-value cut-off of 1e , and the abundance of function was calculated based on KO, Pathway,
EC, and Module.

Statistical analyses
Differential metabolites were considered by VIP > 1 (the variable importance in projection was obtained
based on PLS-DA), P-value < 0.05 (the statistical significance was calculated based on T-test), and FC ≥ 2 or
FC ≤ 0.5 (fold change of metabolites). PCA and PLS-DA were performed with SIMCA-P (Umetrics,
Sweden). Volcano plots were used to filter metabolites of interest based on related parameters. Results are
expressed as mean ± standard deviation (SD). Data were analyzed using one-way analysis of variance
(ANOVA) and Duncan’s test in the SPSS 19.0 software (SPSS Inc., Chicago, IL., USA) at a significance level
of P < 0.05. For the microbial community, similarities analysis (ANOSIM), barplot, and heatmap were
produced in the Vegan package of R (version 3.3.1). Venn plot and LEfSe (linear discriminant analysis effect
size) were used to recognize the common and/or unique species.

RESULTS AND DISCUSSION
Differences in microbial counts and physicochemical properties
The microbial counts could reflect the activity of fermentation, in which lactic acid bacteria and yeast
counts are shown in Table 1. Both of these organisms were the main fermentative candidates in non-salt
Suancai. The count of LAB in MLS-A was slightly lower than that in MF-A and MR-A. It was probably
associated with the content of isothiocyanates in different varieties of Brassica ingredients belonging to the
Cruciferae plants, which possess antibacterial properties . The LAB counts in samples were consistent with
[21]
that of Laotan Suancai, another well-known Chinese traditional fermented vegetable , which indicated that
[22]
the system of non-salt Suancai was also a typical LAB fermentation. However, the tendency of the yeast
counts was opposite. The yeast count in MLS-A was relatively high at 5.04 log CFU/mL, while that was 2.34
[23]
and 2.67 log CFU/mL in MF-A and MR-A, respectively. Yeast can promote flavor in Paocai , while the
massive amount of film yeast is the inducer of pellicle formation . In addition, yeast counts of samples
[24]
made by Manjing (Brassica rapa L.) were roughly the same and relatively low.

RS was exhausted in all samples after 15 days of fermentation, indicating that the microbes grew well and
reached the end of fermentation. In general, microorganisms can utilize RS and produce acids through
carbon cycle . As shown in Table 1, the substrate of ingredients had no effects on TA content, while it
[25]
varied with the suppliers. MF-B showed the lowest TA content, consistent with its LAB count. It is worth