We investigated five mouse (C57/Bl6) stool samples (designated as M1-M5), five human stool samples (designated as H1-H5), and one ZymoBIOMICS Gut Microbiome Standard (Cat#D6331) from Zymo Research. The Zymo control sample is comprised of 21 different bacterial and fungal strains that mimic the human gut microbiome. Prior knowledge on the composition and proportions of various bacteria in this sample allowed for validation of our sequencing and data analysis pipeline. Stool samples were collected under sterile conditions and stored in DNA/RNA Shield, a nucleic acid stabilizing solution from Zymo Research (R1100). DNA/RNA Shield provides an accurate molecular signature of the sample at the time of collection by preserving nucleic acids at ambient temperature and inactivating organisms including infectious agents. Human stools were collected from healthy volunteers under UT Southwestern Institutional Review Board (IRB) Number STU-022011-211. All research protocols and experiment methods used in this study were approved by the IRB. All participants gave their written informed consent to participate in the research.

We used the ZymoBIOMICS™ DNA Miniprep Kit (D4300) for DNA extraction on all the study samples. Figure 1 illustrates the design and experimental workflow of the study. To make sure that each aliquot received unbiased representation of the sample, the specimen was first hand-mixed using the spoon provided in the DNA/RNA Shield Fecal Collection Tubes. Then once all the large clumps were dissolved in the specimen and the sample appeared to be more uniform in solution, 1 ml of aliquots were prepared for various bead-beating conditions. Similarly, 75 ul of ZymoBIOMICS Gut Microbiome Standard (D6331) and 925 ml of DNA/RNA Shield was aliquoted into each of four separate tubes to test with four bead-beating conditions. Each sample was aliquoted into a ZR BashingBead lysis tube (0.1 and 0.5 mm beads). Next, each sample tube was tightly closed and loaded onto the PowerLyzer 24 Homogenizer (110/220 V) for bead-beating. We selected four different bead-beating time points as illustrated in Figure 1 : 0 min (no bead-beating at all), 1 min (one cycle of shaking for 1 min), 4 min (two cycles of 2 min shaking, with a 2 min pause after each cycle), and 9 min (6 cycles of 1 min 30 s, with a 2 min pause after each cycle). Each of these samples were bead-beaten at a speed of 2200 RPM and were maintained at a temperature of 20°C throughout the bead-beating process. Following beat-beating and lysis, DNA was purified using the ZymoBIOMICS protocol, and 100 ul was eluted for downstream experiments. The DNA concentration was measured using the Picogreen method (Invitrogen Quant-iT™ Picogreen dsDNA Assay Kit Reference No. P11496 on Perkin Elmer 2030 Multilabel Reader Victor X3), and the DNA integrity number (DIN) was determined on 4150 Tapestation from Agilent using Agilent’s gDNA Screen Tape (Reference No. 5067-5365) and Agilent’s gDNA Reagents (Reference No. 5067-5366).

We used 10-20 ng of stool-purified DNA to amplify hypervariable region V3-V4 of the bacterial 16S rRNA gene using the Zymo Research Quick-16S NGS Library Prep Kit (Catalog No. D6400). A single amplicon of about 460 bp was amplified using the Quick-16S Primer Set V3-V4 included in the library prep kit. PCR was stopped using the Reaction Clean-Up Solution per the kit’s recommendations. Next, Zymo Research Index Primer Set A was ligated to the amplicon for multiplexing. After the addition of barcodes, the final library size was 596 bp. Then the libraries were then brought to a volume of 50 ul with the addition of DNase/RNase-free water and purified using 56 ul of Beckman Coulter’s AMPure XP beads (Catalog No. A63881). The final libraries were eluted using 27.5 ul of Tris-HCl buffer (10mM, pH 8.5). Quality and quantity of each sequencing library were assessed using Agilent’s 4150 Tapestation using gDNA Screen Tape and gDNA Reagents both from Agilent and picogreen measurements, respectively. The libraries were then pooled in equal concentrations according to picogreen measurements. Each pool was quantified using the KAPA Biosystems Library Quant Kit (illumina) ROX Low qPCR Mix (Reference No. 07960336001) on an Applied Biosystems 7500 Fast Real-Time PCR system. According to the qPCR measurements, 6 pM of pooled libraries was loaded onto a MiSeqDX flow cell and sequenced using MiSeq Reagent Kit v3 600 Cycles PE (Paired end 300 bp). Raw FASTQ files were demultiplexed based on unique barcodes and assessed for quality.

To determine the impact of bead-beating intensity on species-level diversity in the gut microbiome, the full length 16S rRNA gene was sequenced on subset samples (one Zymo, two Mouse, and three human samples) using long-read sequencing technology from Oxford Nanopore. About 20 ng of DNA was used to amplify and barcode the entire (1.5 Kb) segment of the 16S rRNA gene using Nanopore’s 16S Barcoding Kit (SQK-16S024). The PCR product was then purified using Beckman Coulter’s AMPure XP beads (Catalog No. A63881). The purified amplicon libraries were quantified using the Qubit dsDNA HS Assay kit (Catalog No. Q32854) on an Invitrogen Qubit 4 fluorometer. Then sequencing adapters were ligated to the pooled barcoded reads according to the manufacturer’s instructions. Sequencing was performed using R9.4.1 flow cell for 24 h on the MinION (Oxford Nanopore Technologies). Nanopore sequence data was analyzed with EPI2ME Agent v2020.2.10. The 16S sequences were assigned taxonomy using the What’s in my pot? (WIMP) workflow as illustrated in Figure 1 .

Samples with more than 50 K QC pass short sequencing reads from MiSeqDx were used for 16S OTU analysis. Taxonomic classification and operational taxonomic units (OTUs) abundance analysis was performed using the CLC Bio microbial genomics module (https://www.qiagenbioinformatics.com/plugins/clc-microbial-genomics-module/). Individual sample reads were annotated with the Greengenes v13 database using a 97% similarity index. Alpha and beta diversity analysis was done to understand within- and between-treatment group diversity, respectively. Nanopore 16S data were analyzed using the EPI2ME pipeline and WIMP workflow from Oxford Nanopore Technology (ONT). Raw FASTQ files from Illumina and Nanopore sequencing have been deposited in the Sequence Read Archive (SRA) with accession no. PRJNA685188 (v3-v4 amplicon data).

To compare the microbiome diversity between samples and treatments, we applied PERMANOVA analysis (PERmutational Multivariate ANalysis Of VAriance, also known as non-parametric MANOVA (Tang et al., 2016) available in CLC Bio Microbial Genomics Module 20.0). This measures effect size and significance on beta diversity for variables. The significance is obtained by a permutation test. In addition, abundances across various bead-beating conditions were compared using a linear model differential abundance test. This tool models each feature (e.g., OUT or an organism) as a separate generalized linear model (GLM). Microbiome compositions were compared across time points using DEseq (DESeq2_1.26.0), in R version 3.6.1. Data in Figures 3 – 5 represent relative abundance of species determined based on the number of reads detected for that species. Total observed reads for a species were normalized to 1 and then relative abundance at each treatment point was calculated. Statistical analysis was performed using GraphPad Prism software version 8 (GraphPad). T-test was performed and a two-tailed p value of <0.05 was considered significant. MinION sequencing data were analyzed using EPI2ME Agent v2020.2.10; 16S sequencing reads were assigned taxonomy using the WIMP workflow.

As shown in Supplementary Figure 1 , total DNA yield was significantly (t-test p<0.05) high in the samples beaten for 4 and 9 min ( Supplementary Figure 1A ). DNA integrity number (DIN) was high in samples beaten for 1 or 4 min ( Supplementary Figure 1B ). The number of pass filter sequencing reads were similar across the treatments ( Supplementary Figure 1C ). We also compared the total number of high-confidence OTUs annotated in all the samples in the amplicon and full length sequencing data. Median values of these data are compared using a t-test and presented in scatter plots in Supplementary Figure 1D . Overall, the data suggest that bead-beating has no significant impact on the number of sequence reads or OTUs ( Supplementary Figure 1D ).

Quality pass sequencing reads were used to cluster OTUs in the study samples. Supplementary Figure 2 shows v3-v4 amplicon data-based phylum level OTUs in mouse and human stool samples. Alpha and beta diversity indices were determined for various bead-beating intensities ( Supplementary Figures 3A, B ). As shown in Supplementary Figure 3B , mouse samples were tightly clustered based on bead-beating condition. The Bray-Curtis method of beta-diversity assessment was used to evaluate dissimilarity index between communities. Shannon’s entropy and Simpson’s indices were reduced upon extensive bead-beating ( Supplementary Figures 3C, D ). However, overall phylogenetic diversity was higher in 4-min bead-beaten mouse stool ( Supplementary Figure 3E ). Abundance of OTUs in different bead-beating treatments was compared using generalized linear model differential abundance test on groups defined by bead-beating treatments ( Supplementary Table 1 ). In addition, PERMANOVA analysis was also used to measure significance on beta diversity ( Supplementary Table 2 ). Similar analysis was performed in human stool data and OTUs with statistically different abundances ( Supplementary Table 3 ). Supplementary Figures 3F, G show alpha- and beta diversity analysis in human stool samples. As shown, beta-diversity indices in human stool samples were very different from that of mice, as higher between sample diversity was observed ( Supplementary Figure 3G and Supplementary Table 4 ). We observed high Simpson’s index and Shannon entropy as well as phylogenetic diversity in 4- and 9-min bead-beaten samples in human stool ( Supplementary Figures 3H–J ), respectively. Overall, analysis showed that studied parameters were more consistent for the 4 min group in both mouse and human samples.

We tested and standardized our 16S sequencing and data analysis pipeline on a Zymo mock control sample. As shown in Supplementary Figure 4 and Supplementary Table 5 , our assay was able to recover all Gram-negative and Gram-positive strains in a mock sample in very similar proportions as pooled by Zymo Research (Pearson r=0.75-0.87, p<0.001), suggesting that our assay was capturing the read out quite accurately. As expected, bead-beating had a relatively moderate impact on the recovery of Gram-negative bacteria such as Escherichia, Prevotella, and Akkermansia ( Supplementary Figure 4A ). On the other hand, maximum abundance of Gram-positive bacteria such as Roseburia, Bifidobactarium, and Lactobacillus was only captured either in 4- or 9-min beaten samples ( Supplementary Figure 4B ). Consistent with literature that suggests that the complex cell wall in Gram-positive bacteria requires more intense lysis, we observed a strong correlation (Pearson’s r =0.91) between bead-beating intensity and recovery of Gram-positive strains in mock control ( Supplementary Figure 4C ). We used the same sequencing protocol and analysis pipeline to analyze the gut microbiome in mouse and human stool next.

Analysis of 16S rRNA v3-v4 amplicon sequencing data in mouse stool showed that the recovery of phylum Actinobacteria and Firmicutes was significantly (p<0.05) affected by bead-beating intensity ( Figure 2A ). As illustrated, maximum recovery of Actinobacteria and Firmicutes was observed in samples beaten for 4 and 9 min. On the other hand, the highest abundance of Proteobactaria and Bacteroidetes was observed in unbeaten or 1-min beaten samples ( Figure 2A ). Consistent with mouse stool data, Actinobacteria, Bacteroidetes, and Proteobacteria in human stool also showed similar sensitivity to bead-beating treatment, however human samples were more heterogeneous ( Figure 2B ). Interestingly, Firmicutes detected in mouse and human stool exhibited different sensitivity to bead-beating treatment ( Figures 2C, D ). As shown in Figure 2C , Firmicutes in mice stool were mostly comprised of Allobaculum and Clostridiales and showed a consistent pattern of recovery across all the samples ( Supplementary Table 6A ). On the other hand, Firmicutes in human stool were comprised of more diverse bacteria that showed large sample-to-sample heterogeneity in composition and abundances ( Figure 2D ). As shown, Allobaculum in mouse stool required extensive bead-beating for maximum recovery, whereas Firmicutes in human stool, i.e., Veillonella, Ruminococcus, and Acidaminococcus showed maximal recovery in unbeaten or 1-min beaten samples ( Supplementary Table 6B ). These differences in the composition of Firmicutes in mouse and human stool were the reasons for discordant results. Analysis of top bacteria in mouse stool data showed that recovery of Lactobacillus, Allobaculum, Bifidobacterium, Coriobacteriaceae, F16, and Clostridiales was significantly (p<0.05) affected by bead-beating treatment ( Supplementary Table 7 ). Similarly, comparison of OTUs in human stool samples revealed the differences in abundances between the four bead-beating conditions ( Supplementary Table 8 ). Analysis showed that recovery of Dorea, Blautia, Ruminococcus, Lactobacillus, and Bifidobacterium was significantly (p<0.05) affected by bead-beating intensity ( Supplementary Table 8 ). These data suggest that applying the same bead-beating treatment to mouse and human stool samples may obscure the actual diversity of the gut microbiome.

Next, we clustered the genus-level OTUs in mouse and human stool based on their sensitivity to bead-beating intensity. We used ClustVis (Metsalu and Vilo, 2015) to stratify bacteria into clusters based on their abundance at various bead-beating conditions ( Figures 2E, F ). As shown in Figure 2E , this analysis stratified the top 22 OTUs into four major clusters in mouse stool. Cluster 1 was comprised of bacteria Dorea, Bifidobacterium, Lactobacillus, and Allobaculum that showed maximum abundance in the 9-min beaten sample as compared to the unbeaten sample. The second cluster included bacteria such as Prevotella and Bacteroides that showed maximum recovery at 1 or 4 min of bead-beating. Cluster 3 included bacteria that required minimal shaking, 1 min of or no bead-beating at all. The fourth cluster was comprised of organisms such as Helicobacter and Sutterella that showed maximum recovery in the unbeaten sample ( Figure 2E ).

Similar clusters of bacteria were also observed in human stool as well ( Figure 2F ). Cluster 1 included many of the common human gut commensals such as Dorea, Blautia, Bifidobacterium, and Lactobacillus. These organisms were very underrepresented in unbeaten samples. Cluster 2 constitutes the moderate group that certainly required some (1 min) bead-beating treatment as suggested by its reduced recovery in both 0 as well as in 9-min beaten samples. Interestingly, cluster 3 included some known human pathogens such as Klebsiella, Hemophilus, and Citrobacter. These organisms showed maximum recovery in unbeaten or 1-min beaten samples ( Figure 2F ). Cluster 4 showed maximum recovery of OTUs in unbeaten samples. This cluster included Faecalibacterium, Serratia, Veillonella, and Lachnospira ( Figure 2F ).

The 16S v3-v4 amplicon sequencing data were very limited in species-level taxonomic classification of detected OTUs ( Supplementary Table 9 ). About 35-40% of OTUs were classified up to a species level of taxonomy. The remaining 60-65% were only classified up to the phylum, class, order, family, or genus level ( Supplementary Table 10 ). So, next we performed full length 16S rRNA gene sequencing on two mice (M2, M3) and three human stool samples (H1, H2, H4). Long-read sequencing data was classified into various taxonomic ranks using the EPI2ME WIMP pipeline. NCBI taxonomy trees were generated based on the number of detected reads. The 16s v3-v4 amplicon and full length 16S rRNA sequencing data showed good correlation at the phylum level (Pearson r>0.70, p<0.01). Full length data characterized vast species-level diversity in mouse and human samples ( Figures 3 , 4 ). The 16S v3-v4 amplicon data annotated 14 and 10 OTUs to the species level of taxonomy in M2 and M3 mouse stool, respectively. On the other hand, full length sequences-based analysis revealed 98 and 96 bacterial species in M2 and M3 mouse stool, respectively. Similarly, amplicon data in H1, H3, and H4 samples annotated 10, 11, and 8 OTUs to species level respectively, whereas, full length sequence analysis assigned species rank to 155, 143, and 120 detected organisms in H1, H3, and H4 samples, respectively. These numbers were calculated based on a minimum of 15 long reads to support a taxon. Analysis of full length data showed that about 78% (75 out of 97) of the total observed species in M2 could be detected in the unbeaten sample, whereas only 59% (56 out of 95) of the total observed species were detected in the unbeaten M3 stool, suggesting potential differences in the proportion of bacteria that are more sensitive to bead-beating treatment. Next, to calculate the relative percentage of observed species at each time point, we normalized the total reads for the given species to 100 percent. Data presented in Figures 3 – 5 show normalized relative percentages of observed species. This allowed comparison of recovery in terms of observed abundance of the species at each bead-beating treatment. Furthermore, analysis showed that about 27% (26 out of 95) of bacteria in M3 stool showed maximum abundance in 4- or 9-min beaten samples as compared to 10% (10 out of 97) in M2 ( Figure 3B and Supplementary Figure 5 ). As shown in Supplementary Figure 5 , long-read sequencing data profiled species diversity in several major gut commensal genera, i.e., Bacteroides, Clostridium, Lactobacillus, Ruminococcus. We observed that some species in Roseburia, Blautia, and Ruminococcus showed variation in their recovery with respect to bead-beating treatment. Of these, R. gnavus and R. albus species of the Ruminococcus genus were particularly interesting as the maximum abundance of R. gnavus was detected in 4- and 9-min beaten samples, whereas R. albus species showed maximum abundance in 0- or 1-min beaten mouse samples ( Supplementary Figure 5H and Supplementary Table 11 ).

Next, full length 16S rRNA sequences-based microbiome analysis in three human stool samples (H1, H3, and H4) revealed vast species diversity (about an 8-10-fold increase compared to amplicon data) and sample-to-sample variation in the microbiome composition as suggested by variable number and types of detected bacteria ( Figure 4 and Supplementary Table 12 ). The number of bacteria detected in different bead-beating treatments varied from sample to sample ( Supplementary Table 12 ). It was observed that species of the genus Blautia, Streptococcus, and Ruminococcus exhibited variation in recovery with respect to bead-beating treatment ( Supplementary Figures 6A–E ). Full length sequencing data analysis showed significant diversity and heterogeneity in organisms detected in human stool ( Figure 4B ). For example, in sample H3, some clinically relevant microbes such as Citrobacter freundii and Klebsiella pneumoniae were also detected ( Supplementary Figures 6F–J ). We observed the highest abundance of these bacteria in the unbeaten sample ( Supplementary Figures 6F–J ). Similarly, species of Lactobacillus and Streptococcus in the H4 sample consistently showed highest abundance in the 9-min beaten sample ( Supplementary Figures 6K–N ). Although three samples are not enough for statistical comparisons, these data do explore microbial diversity and sample-to-sample heterogeneity in microbiota composition in human fecal material. Many OTUs in H1 and H4 showed maximum abundances in 4-9-min beaten samples. On the other hand, a large number of OTUs in H3 stool showed maximum abundance in the unbeaten sample ( Figure 4 and Supplementary Table 12 ). Next, we sorted bacterial species based on Gram-positive and Gram-negative classification and assessed their recovery across four bead-beating conditions in H3 stool. As expected, Gram-negative species showed maximum abundance in the unbeaten sample ( Supplementary Figure 7 ). We compared amplicon OTU composition in mouse and human stool across bead-beating time points using DEseq analysis ( Supplementary Tables 13 , Supplementary Table 14 ). We also compared abundances of various species detected in full length 16S data analysis across four bead-beating treatments and identified those that showed significant variation in recovery between four conditions ( Supplementary Table 15 ). As shown in Figures 5A–F , we observed variation in recovery of Blautia and Streptococcus species. For example, the abundance of Blautia hominis species showed no significant difference in recovery between the four bead-beating conditions, whereas Blautia luti and other species of this genus showed maximum abundance in the 9-min beaten sample ( Figures 5G, H and Supplementary Table 15 ). Similarly, abundance of Streptococcus parasanguinis species was not significantly different at the four different bead-beating conditions as was the case with Streptococcus thermophilus and other common species of this genus that exhibited significantly (p<0.01) higher abundance in the 9-min beaten sample ( Figures 5I, J , Supplementary Table 15 ). These data support our hypothesis that different species of a genus may exhibit variation in sensitivity to bead-beating treatment.

In mouse stool data, we observed that about 50% of Ruminococcus reads were annotated as R. gnavus that showed maximum abundance in 9-min beaten samples in all the mice samples. The remaining reads that could not be assigned any species showed relatively higher abundance in unbeaten or 1-min beaten samples, as shown in M2 and M3 ( Figure 6A ). Full length 16S rRNA sequencing in M2 and M3 samples revealed other species such as Ruminococcus champanellensis, Ruminococcus flavefaciens, and Ruminococcus albus that were rather more abundant than Ruminococcus gnavus ( Figure 6B ). Interestingly, we observed that abundance of Ruminococcus albus species in M3 long-read data was highest in the unbeaten sample. Similarly, amplicon data on human stool also showed a lot of Ruminococcus reads that were not classified into any species, consistent with mouse stool data. Although maximum abundance of Ruminococcus was captured in 9-min beaten samples, abundance of unclassified species was also observed at 1 and 4 min as well. The most interesting sample was H1, in which about 80% of Rumnicoccus reads could not be classified into any species, whereas the other two samples (H3 & H4) showed a high abundance of R. gnavus reads ( Figure 6C ). Full length 16S rRNA gene sequencing in these three samples further revealed the species-level diversity, especially in the H1 sample. As shown in Figure 6D , the H1 sample showed the presence of various Ruminococcus species including Ruminococcus albus, Ruminococcus bromii, Ruminococcus callidus, Ruminococcus champanellensis as well as Ruminococcus gnavus. More interestingly, consistent with M3 mouse data, the abundance of Ruminococcus albus in the H1 sample was also highest in the unbeaten sample ( Figure 6D ). Unlike H1, samples H3 and H4 were mostly enriched with Ruminococcus gnavus species, which is consistent with their amplicon data in Figure 6C . Next we performed multisequence alignments and phylogenetic analysis of various Ruminococcus species to explore the 16S gene regions of divergences between these species ( Figures 6E, F ). As shown, complete 16S gene sequence alignment between R. albus and R. gnavus species showed that most of the differences lay outside the v3-v4 hypervariable region ( Figure 6E ). Maximum likelihood phylogenetic analysis on complete 16S rRNA gene sequences showed a genetic relationship between four common Ruminococcus species ( Figure 6F ).

To assess the impact of bead-beating on the recovery of some clinically relevant microbes, we performed a bead-beating experiment on the ZymoBIOMICS microbial community standard (D6300) that represents a balanced mixture of common infectious microorganisms including Listeria monocytogenes, Bacillus subtilis, Staphylococcus aureus, Enterococcus faecalis, Lactobacillus fermentum, Salmonella enterica, Escherichia coli, and Pseudomonas aeruginosa. The v3-v4 amplicon sequencing analysis on this sample showed that Bacillus, Listeria, and Lactobacillus bacteria certainly required 4-9 min of bead-beating for maximal recovery ( Supplementary Figures 8A–C ). On the other hand, Salmonella, Pseudomonas, and Enterococcus showed maximum recovery in the unbeaten samples ( Supplementary Figures 8D–F ). The presented data are from two independent experiments on the same Zymo control DNA. Similarly, we also observed a variation in the recovery of clinically relevant microbes in human stool samples. As shown in Supplementary Figures 8G–J , Streptococcus, Dorea, Blautia, and Coporocuccus exhibited a variation in recovery with respect to bead-beating treatment. It was observed that abundance of these bacteria was significantly (p=0.05) higher in the 9-min beaten sample as compared to the unbeaten sample ( Supplementary Figures 8G–J ). On the other hand, Hemophilus and Citrobacter showed an opposite trend with maximum recovery in the unbeaten sample ( Supplementary Figures 8K–N ). However, these results are exploratory and need validation in larger sample cohorts in the future.