Cell recovery and DNA extraction methods skew relative abundance estimations of foodborne pathogens and spoilage organisms on stainless steel surfaces Authors: Jingzhang Feng, Sarah E. Daly, Katerina Roth, Abigail B. Snyder Published: 2026 ABSTRACT ABSTRACTAmplicon sequencing investigations of surface microbiota in food facilities often report the relative abundance of bacteria and fungi. However, physiological differences among cell types can result in variable cell recovery and DNA yields, thereby skewing relative abundance estimates. Here, we evaluated (i) variations in cell recovery among different bacterial and fungal species after surface swabbing and (ii) the impact of DNA extraction protocols on relative abundance estimates from artificially inoculated stainless steel surfaces. Our results showed that Escherichia coli (Gram-negative cell), Listeria monocytogenes (Gram-positive), Bacillus cereus (bacterial spore), Alicyclobacillus suci (bacterial spore), Exophiala phaeomuriformis (fungal cell), Aspergillus fischeri (fungal spore) differed significantly (P < 0.05) in their recovery rates from stainless steel surfaces. Vegetative cells (E. coli and L. monocytogenes) exhibited lower average recovery rates from surface swabbing (2.9%–6.6%) than spores (35.2%–94.9%). Extending the bead-beating step in DNA extraction by 10 min generally improved yields though the impact varied by organism. For example, DNA yields of E. coli increased from 70 to 84 ng/mL while that of L. monocytogenes increased only from 23.2 to 29.2 ng/mL. Cell recovery and DNA extraction impacted relative abundance estimates from amplicon sequencing. Starting off at equal relative abundances of 25%, L. monocytogenes was underestimated (9%–17%) in downstream calculations, while B. cereus was overestimated (36%–44%). These results underscore the limitations of amplicon sequencing for microbiota characterization on food facility surfaces and highlight the need to improve current swabbing and DNA extraction methods.IMPORTANCEAmplicon sequencing has been used to characterize microbial communities on facility surfaces. However, few studies have evaluated the accuracy of the amplicon sequencing workflow for quantifying spoilage and pathogenic organisms in these microbial communities. Here, we assessed the accuracy of amplicon sequencing to evaluate the relative abundance of spoilage and pathogenic organisms commonly found in food-processing environments. The results revealed biases in relative abundances due to limitations in cell recovery and DNA extraction methods. These findings revealed the potential biases in surface microbiota characterization in food facilities and the need to refine current recovery and extraction methods to enhance the accuracy of microbiota characterization. INTRODUCTION Amplicon sequencing has been used to characterize microbiota on food facility surfaces (1–5). This process involves recovering cells from environmental surfaces with swabs, extracting DNA from the recovered cells, amplifying targeted DNA regions with polymerase chain reaction (PCR), and sequencing the DNA amplicons (6). Nucleotide sequences are then mapped to databases (e.g., NCBI RefSeq) to determine the taxonomy of organisms within the microbiota (7). Subsequently, the relative abundance of each taxon is calculated to reflect its proportion within the community (8). The relative abundances of spoilage (e.g., Alicyclobacillus suci, Exophiala phaeomuriformis, Aspergillus fischerii) and pathogenic (e.g., Escherichia coli, Listeria monocytogenes, Bacillus cereus) organisms are often compared among samples from various environmental conditions (9). Food producers are especially interested in characterizing and tracking the harborage points of pathogenic bacteria and spoilage organisms to assess and prevent product contamination risks. Food facility surfaces contain diverse spoilage and pathogenic organisms including Gram-positive and Gram-negative bacteria, yeasts, and molds. Due to differences in physiochemical properties (e.g., sizes, cell wall thickness), these organisms exhibit different surface attachment, which affect their removal from surfaces during recovery (10–12). As a result, different proportions of the cells of different microbes may be recovered from the surface via swabbing, causing the estimations of their relative abundances to be skewed (13–15). While some studies have compared the recovery of Gram-positive and Gram-negative bacteria using sponge swabs, differences in recovery among other cell morphologies have not been examined (13, 16). Therefore, it remains unclear to what extent sponge swabs can introduce recovery biases that skew relative abundance calculations. Further bias in relative abundance calculations can be introduced during DNA extraction. The distinct physicochemical properties of different cell types can impact DNA extraction and result in inconsistent DNA yields. Specifically, the structural differences between organisms can impact cell integrity, which affects the efficiency of lysis and DNA release. This issue is particularly prominent in spore-forming organisms since they contain a hard cell barrier that resist lysis (17, 18). However, commonly used DNA extraction methods (i.e., DNA PowerSoil Kit) can produce low DNA yields for spore-forming organisms (17, 19). Consequently, the relative abundance of spore-formers could be underestimated when using this method. Additionally, any modified methods need to maintain consistent DNA yield across organisms to prevent biases. Although some optimized methods have been proposed, little research has assessed the impact of these modified methods on relative abundances. Furthermore, few investigations have been carried out to evaluate method performance on spoilage organisms that can form spores in food facilities. In summary, issues with recovery and extraction yields can lead to inaccurate assessment of relative abundances of relevant spoilage and pathogenic microbes on surfaces in the food industry. This study aims to evaluate the effects of the sponge swab recovery method and modifies DNA extraction protocols on the relative abundances of spoilage and pathogenic organisms (Fig. 1). Understanding the impact of these methods can inform strategies to improve accuracy and understand the limitations of microbiota characterization. The experimental workflow to evaluate the impact of surface recovery and DNA extraction on downstream amplicon sequencing results of microbes from surfaces in food processing environments. MATERIALS AND METHODS Spore and vegetative cell cultures Alicyclobacillus suci and Bacillus cereus spore stocks Alicyclobacillus suci FSL W10-0048 was grown on Yeast-Starch-Glucose (YSG) (pH 4.0) agar at 45°C for 7 days to allow sporulation (20). Bacillus cereus FSL A1-001 was grown on brain heart infusion (BHI) agar at 35°C for 24 h, and then an isolated colony was transferred to AK#2 agar (BD, Thermo Fisher Scientific, Waltham, MA, USA) and incubated at 35°C for 7 days to allow sporulation. A. suci and B. cereus spores were harvested by adding 1 mL of sterile deionized water and gently scraping the surface of the plate with a plate spreader. The remaining vegetative cells were removed by treatments of 200 mg/mL lysozyme at 37°C for 40 min, 1% Sodium Lauryl Sulfate at 30°C for 20 min, and then an 80°C heat shock for 10 min. The presence of spores in the final suspension was confirmed by malachite green staining. Spore concentrations were confirmed by cell counts on hemacytometer and colony counts of heat shocked cultures (80°C for 10 min) on agars (YSG for A. scui, BHI for B. cereus). Exophiala phaeomuriformis cells and Aspergillus fisherii ascospores collection E. phaeomuriformis and (FSL E2-0572) A. fisherii (NRL B-2354) were grown separately on Acidic Potatoes Dextrose Agar (APDA) (pH 4.0) and incubated at 25°C for 28 days. Colonies were harvested with 1 × phosphate-buffered saline (PBS) and pelleted. Pellets were washed and resuspended again in 1 × PBS (21). The presence of fungal cells and ascospores in the final suspension was confirmed by phase contrast microscopy (AMScope, Irvine, CA, USA). Fungal culture concentrations were confirmed by the colony counts on APDA and cell counts using a hemacytometer. Escherichia coli and Listeria monocytogenes cultures Both E. coli FSL A1-0127 and L. monocytogenes FSL B2-0079 were incubated separately in Trypticase Soy Broth at 35°C for 24 h. Vegetative cells were pelleted, washed, and resuspended in 1 × PBS. Concentrations of vegetative cultures were confirmed by colony counts on BHI agar and cell counts using a hemacytometer. Assessing recovery rate of each cell culture from stainless steel coupon surfaces Each cell culture was adjusted to approximately 8 log CFU/mL, serially diluted, and plated on agars to confirm its concentration. Ten drops (10 µL per drop) of the adjusted culture were inoculated onto a 1 ft × 1 ft sterilized stainless steel coupon (7 log CFU/coupon in total). Stainless steel coupons were dried under a fume hood for 20 min. Cells on stainless steel coupon surfaces were collected by 3 M swabs (Neogen, Lansing, MI, USA) premoistened with 10 mL of Dey-Engly broth (22). The handle of the 3 M swabs was snapped off, and 10 mL of 1 × PBS was added to the sponge. The sponge was stomached at 260 rpm for 2 min. Cell pellets were collected by transferring sponge buffers to a 50 mL Falcon tube and centrifuging at 6,500 relative centrifugal force (×g) for 15 min. Cell pellets were resuspended in 1 mL of 1 × PBS. One hundred microliters of each resuspended culture was serially diluted and plated on BHI to enumerate B. cereus, E. coli, and L. monocytogenes, plated on YSG to enumerate A. suci, and plated on APDA to enumerate E. phaeomuriformis and A. fischerii. BHI plates were incubated at 35°C for 24 h, YSG plates were incubated at 45°C for 48 h, and APDA plates were incubated at 25°C for 72 h. The recovery percentage for each culture was determined using the following equation: Determine the optimized method for DNA extraction on A. suci Four modified DNeasy PowerSoil Pro kit (Qiagen, Redwood City, CA, USA) protocols and the standard PowerSoil kit protocol (PS) were tested by extracting DNA from 5 log CFU/mL of A. suci spores. The cell concentration of the spore culture was confirmed using a hemacytometer prior to extraction. The modifications include (i) the standard protocol with an additional 10 min of beads beating (PS + 10), (ii) heat pre-treatment at 95°C for 10 min in CD1 buffer (PS95), (iii) heat pretreatment at 65°C for 10 min in CD1 (PS65), (iv) enzymatic pretreatment in lysis buffer containing 10 mg/mL lysozyme plus 5 mg/mL proteinase K at 37°C for 10 min (PSlys) (3). Each DNA extraction method was performed in triplicate. Extraction yields were quantified based on nucleic acid concentration (ng/mL) in the lysate using a Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Ct values of 16S V3-V4 amplicons from each lysate were determined by quantitative polymerase chain reactions (qPCRs). Specifically, 2 µL of each lysate was added to a master mix containing 12.5 µL of SYBR Green PCR Master mix (Thermo Fisher Scientific, Waltham, MA, USA), 1 µL each of 10 µM forward and reverse primer, 8.375 µL of ddH2O, and 0.125 µL of ROX control. qPCR was conducted in QuantiStudio 6 (Thermo Fisher Scientific, Waltham, MA, USA) with the following conditions: initial denaturation at 95°C for 2 min, 40 cycles of 95°C for 5 s, 60°C for 10 s. Three biological replicates of DNA extractions were performed. A newly prepared spore culture was used for each replicate, and each replicate was considered an independent experiment. Due to the variability in spore culture concentrations between trials, DNA concentrations in lysates varied across replicates and were, therefore, presented separately to better illustrate the differences between extraction methods. Evaluation of modified DNA extraction methods on other spores and vegetative cells The PS + 10 method demonstrated improved DNA yield from A. suci spore extractions and was, therefore, selected to evaluate extraction efficiency across all cultures. Suspensions with 5 log, 4 log, and 3 log CFU/mL of each bacterial culture, and suspensions with 6 log, 5 log, 4 log, and 3 log CFU/mL of each fungal culture were prepared. The cell concentration of each culture was confirmed by the cell count under hemacytometer. All suspensions were pelleted prior to extractions. DNA was extracted from pellets using either the PS or PS + 10 method. DNA yields were quantified by measuring DNA concentrations from 5 log CFU/mL of bacterial cultures and 6 log CFU/mL of fungal cultures using Qubit. Extracts from cultures with lower cell concentrations were below the quantification limit and were, therefore, not quantified. Three biological replicates of DNA extractions were performed. Newly prepared bacterial or fungal cultures were used for each replicate, and each replicate was considered an independent experiment. Due to the variability in culture concentrations between trials, DNA concentrations in lysates varied across replicates and were, therefore, presented separately to better illustrate differences between extraction methods for each organism. Extract DNA from microbiota Three types of microbiota were prepared, each containing two bacterial spore cultures (i.e., A. suci and B. cereus), two vegetative cell cultures (E. coli and L. monocytogenes), and two fungal cultures (A. fischeri and E. phaeomuriformis). One microbiota contained an equal ratio of each culture. The second type of microbiota contained 6 log CFU/mL of each bacterial spore and fungal culture and 5 log CFU/mL of each vegetative cell culture, representing a high spore-to-vegetative cell ratio. The third type of microbiota contained 5 log CFU/mL of each bacterial spore and fungal culture, and 6 log CFU/mL of each vegetative cell culture, representing a low spore-to-vegetative cell ratio. Cell concentrations were confirmed by hemacytometer counts prior to creating the microbiota. Triplicates of each microbiota were prepared. Newly prepared bacterial or fungal cultures were used for each replicate, and each replicate was considered an independent experiment. Nucleic acids from each microbiota were extracted with the method PS + 10. Extraction lysates were sent to NovoGene Co. (Beijing, China) for 16S rRNA V3-V4 and ITS2 amplicon sequencing. Relative abundances of organisms were presented separately for each trial to illustrate differences between the experimentally calculated and theoretical relative abundance values. Assessing the amplification rate of 16S rRNA V3-V4 region for bacterial DNA and ITS2 regions for fungal DNA One milliliter of each bacterial and fungal culture was pelleted by centrifuging at 12,000 × g for 5 min. Each pellet was subjected to DNA extraction using the PowerSoil kit, and the resulting lysates were quantified with Qubit to determine DNA concentration. The DNA concentration of each lysate was adjusted to 200 ng/mL with ddH2O. Two microliters of each adjusted lysate was added to the qPCR master mix and performed qPCR with conditions described in the Determine the optimized method for DNA extraction on A. suci section. Recovering microbiota from surfaces and performing amplicon sequencing Triplicates of microbiota with 6 log CFU/mL of cells per culture were created. Cell concentrations were confirmed by hemacytometer prior to creating the microbiota. Microbiota cultures were centrifuged at 12,000 × g for 5 min to pellet cells. Pellets were resuspended in 100 µL of 1 × PBS, inoculated onto coupons, recovered with swabs, and pelleted as described in the Assessing recovery rate of each cell culture from stainless steel coupon surfaces secion. DNA was extracted from the pellet using the PS + 10 method. Extraction lysates were sent to NovoGene Co. (Beijing, China) for 16S rRNA V3-V4 and ITS2 amplicon sequencing. Assessing the impact of swabbing on relative abundance Viable cells of each bacterial and fungal species in the microbiota described in the Recovering microbiota from surfaces and performing amplicon sequencing section were enumerated before inoculation and after recovery using selective media to assess the impact of swab recovery on relative abundances. For enumeration, 100 µL of serially diluted culture was plated onto YSG to enumerate A. suci, on Violet Red Bile Glucose agar (VRBG) to enumerate E. coli, on LMPN to enumerate L. monocytogenes, and on APDA to enumerate E. phaeomuriformis and A. fischeri. An additional 100 µL aliquot was heat-shocked at 80°C for 10 min, serially diluted, and plated on BHI to enumerate B. cereus. Incubation conditions were BHI and LMPN at 35°C for 24 h, VRBG at 37°C for 24 h, APDA at 25°C for 72 h, and YSG at 45°C for 48 h. Relative abundance of each viable bacterial or fungal organism before inoculation and after recovery was calculated as follows: Newly prepared bacterial or fungal cultures were used for each replicate, and each replicate was considered an independent experiment. Relative abundances of organisms were presented separately for each trial to better illustrate differences between the experimentally calculated and theoretical relative abundance values. Sequence analysis Raw sequencing data were processed with QIIME 2 (22.2) (v. qiime-2022.2) (23). Sequences were quality checked with DADA2 (24). Paired-end reads were denoised, had chimeras removed, and were trimmed if their quality score was below 15. A feature table of Amplicon Sequence Variants (ASVs) was generated after denoising and was analyzed by a pre-trained classifier based on the SILVA 16S v3-v4 database to create a taxonomy table (25). The abundance of each taxon was normalized by its 16S rRNA copy number using the q2-gcn-norm plug-in in QIIME 2. Microbial profiles were generated, and the relative abundance of each microbe was calculated with the phyloseq package in R (version 1.44.0) (26). Statistical analysis Significant differences in cell recovery among the six cell cultures were determined by ANOVA (n = 6) followed by Tukey’s post-hoc test. In the experiment assessing DNA extraction methods for A. suci, differences in DNA concentrations of lysates and Ct values between the standard method (PS) and the modified methods (PS + 10, PS65, PS95, PSLys) were evaluated using paired-end t-tests. In the experiment assessing DNA extraction across different bacterial and fungal cultures, differences in DNA concentrations of lysates between the PS and PS + 10 methods were evaluated by paired-end t-test. In the DNA extraction of mock microbiota, differences in relative abundances between sequencing results and theoretical values (i.e., calculated based on cell counts using a hemacytometer) were assessed using paired-end t-tests. In the experiment recovering microbiota from surface, differences in relative abundances between sequencing results and theoretical values were evaluated using paired-end t-test. Furthermore, differences in relative abundances of viable cells between the recovered and inoculated microbiota were evaluated using paired-end t-test. All statistical analyses were performed using the base function of R (version 1.44.0). RESULTS AND DISCUSSION A lower proportion of vegetative bacterial cells was recovered through surface swabbing than bacterial spores, fungal vegetative cells, and fungal ascospores The percent of recovered L. monocytogenes and E. coli was similar. Although the recovery of L. monocytogenes was numerically higher than E. coli, the difference was not statistically significant (P > 0.05) (Table 1). On average, 2.9% ± 3.0% of the inoculated E. coli cells and 6.6% ± 6.0% of the inoculated L. monocytogenes cells were recovered from surfaces. Both percentages were lower than those reported by Keeratipibul et al. (16) (E. coli: 52.3% ± 6.0%, L. monocytogenes: 73.7% ± 2.6%), but the L. monocytogenes recovery was similar to that observed by Gómez et al. (27) (L. monocytogenes: 2.81% ± 0.63%) (16, 27). The reason for the high recovery percentages reported by Keeratipibul et al. (16) may be due to the use of log-transformed numbers which inflates recovery percentages. TABLE 1The amount (log CFU/ft2) and proportion of the inoculated viable cells recovered from stainless steel surfaces with sponge swabbingbKingdomSpeciesMorphologyGram reactionAverage inoculated cells ± standard deviation(log CFU/ft2)Average recovered cells ± standard deviation(log CFU/ft2)Average %recovery ± standard deviationaBacteria E. coli VegetativeNegative7.25 ± 0.035.49 ± 0.542.9% ± 3.0%a L. monocytogenes VegetativePositive7.7 ± 0.216.34 ± 0.486.6% ± 6.0%a B. cereus SporesPositive6.82 ± 0.126.8 ± 0.1294.9% ± 5.0%b A. suci SporesPositive6.21 ± 0.345.96 ± 0.4635.3% ± 11%cFungi E. phaeomuriformis Yeast-like conidiaNA6.91 ± 0.126.82 ± 0.2480% ± 19%b A. fischerii AscosporesNA5.51 ± 0.094.72 ± 0.6136% ± 12%c a Values followed by same letters are not significantly different (P > 0.05). b Differences in the % cell recovery between the six organisms were evaluated using ANOVA. The amount (log CFU/ft2) and proportion of the inoculated viable cells recovered from stainless steel surfaces with sponge swabbingb Values followed by same letters are not significantly different (P > 0.05). Differences in the % cell recovery between the six organisms were evaluated using ANOVA. Notably, the number of cells that died-off on surfaces during drying could also bias the recovery percentages. Specifically, cells susceptible to desiccation, such as E. coli and L. monocytogenes, may have a lower recovery percentage due to their die-off. However, studies by Wilks et al. (28) and (29) reported a reduction of less than 1 log CFU of E. coli and L. monocytogenes cells on stainless steel surfaces within 180 min of incubation at room temperature (28). Since cells were dried for less than 30 min in this study, the number of dead cells is unlikely to substantially affect the recovery percentage of E. coli and L. monocytogenes. The recovery of other bacteria differed significantly from each other (Table 1). On average, sponge swabs recovered 94.9% ± 5.0% of B. cereus spores and 35.3% ± 11% of A. suci spores from surfaces. While significantly (P < 0.05) higher than the two vegetative cells tested, these numbers were also significantly different from each other. Similarly, the recovery percentages of the two fungi were significantly (P < 0.05) higher than those of vegetative cells and differed from each other. On average, 80% ± 19% of E. phaeomuriformis conidia and 36% ± 12% of A. fischeri ascospores were recovered from surfaces. Differences in recovery between cell morphologies could be due to variations in cell size, which can affect the attachment of cells to surfaces, thus impacting their removal by swabs (30). Previous studies have identified cell size as a physiological property affecting cell removal (31, 32). For instance, Cai et al. (31) demonstrated that fluid shear stress had a greater impact on cell removal when cell size exceeded the width of surface roughness topology (33). Specifically, the study showed that shear stress from rotational fluid had a greater impact on the removal of E. phaeomuriformis, a spore exceeding the roughness measurements of the surface, than on the removal of L. monocytogenes from stainless steel surfaces (33). This finding was consistent with our observation that sponge swabs recovered a lower percentage of L. monocytogenes cells from stainless steel surfaces than E. phaeomurifomis (Table 1). In fact, cell size may have also contributed to the recovery differences between some of the tested vegetative cells (e.g., L. monocytogenes) and bacterial spores (e.g., A. suci) since A. suci spores (1.5–1.8 by 0.9–1.0 µm) are larger than L. monocytogenes (0.5 by 2 µm) cells (34–36). In addition to cell size, other physio-chemical factors, such as chemical composition and hydrophobicity, had also been identified as factors that could affect cell removal (10–12). Importantly, these physiological properties can also affect cell adhesion to swabs, leading to different release capacities for different organisms (10, 37). Differences in release can further contribute to the swab recovery difference (13). Consequently, swab recovery difference can affect analyses linking organism abundances to environmental characteristics. Understanding factors impacting cell recovery using swabs can inform strategies to improve recovery. For instance, recognizing that hydrophobicity of cell membranes can affect removal, some methods have added surfactants to swabs to improve cell recovery (37). Enhanced recovery could provide a more accurate understanding of the association between microbiota characteristics and environmental attributes. Extending the duration of bead beating improved DNA extractions for spores Because of the variability in starting spore concentration among trials, the results in Fig. 2 were shown individually for each trial. Extending beating time (PS + 10) increased DNA concentrations in lysates across all three trials (Trial 1: +4.8 ng/mL, Trial 2: +0.6 ng/mL, and Trial 3: +2.4 ng/mL) compared to the standard method (Fig. 2A). A significant (P < 0.05) increase in lysate DNA concentrations using the PS + 10 method was confirmed by a paired-end t-test. This result aligned with the previous finding that extending bead-beating time by 10 min increased extraction yield from Campylobacter jejuni by 1.6–1.8 folds (38). On the other hand, the method with 95°C heat pre-treatment (PS95), 65°C heat pre-treatment (PS65), and lysozyme pretreatment (PSlys) did not consistently improve DNA concentrations in lysates although some trials produced higher concentrations than the standard method. qPCR was conducted to evaluate the amplification rate of each extraction product in downstream PCR (Fig. 2B). None of the modified methods (PS + 10, PS65, PS95, and PSLys) consistently lowered Ct values, suggesting that these modifications did not enhance amplification efficiency. Interestingly, while the PS + 10 method increased DNA yields, it did not improve amplification rates. The Ct value lowered slightly in two trials (Trial 1: −0.46, Trial 3: −0.39) but increased in one trial (Trial 2: +1.01) when using PS + 10. This outcome could be attributed to DNA shearing caused by extended bead beating. Excessive mechanical force produced from a longer bead beating could fragment DNA molecules released from spores (39, 40). The fragmented DNA may not be amplified in PCR as readily, and therefore, the Ct value did not improve (41). PCR is a key step that enables downstream sequencing and analysis to determine identity, taxonomic classifications, and relative abundance calculations (42). Hence, while the PS + 10 method increased DNA concentrations in lysates, it may provide limited improvements to the accuracy of relative abundance assessments for spores. Future studies should investigate strategies to reduce DNA shearing due to bead beating (e.g., increase bead beating intensity but shorten the duration). These strategies could further improve the accuracy of relative abundance assessments in microbial community analysis (43). Changes in DNA concentrations in lysates (A) and Ct values generated by qPCR (B) compared to the standard method (PS) when using the modified extraction methods (PS + 10, PS65, PS95, or PSLys). Lysate DNA concentrations and Ct values from the modified extraction methods were compared with those of the standard method for each trial. Paired-end t-tests were conducted to identify modified methods with significant improvements in lysate DNA concentrations and Ct values across the three extraction trials. Asterisk (*) indicates the modified method with a significant increase in DNA concentrations compared to the standard method. Gray color grids indicate that results were below the detection limit of Qubit (<10 ng/mL). Blue grids indicate an increased lysate DNA concentration or Ct value compared to PS. Orange grids indicate a decrease in lysate DNA extraction or Ct value compared to PS + 10. The PS+10 method did not improve extractions uniformly across different bacteria and fungi The PS+10 method increased DNA concentrations in lysates from the extraction of some organisms (Fig. 3A and B). The extended beads beating method (PS+10) significantly (P < 0.05) increased DNA concentration in lysates from the extraction of A. suci, B. cereus, E. coli, and E. phaeomriformis. Compared to extractions with PS, PS+10 increased the DNA concentrations in lysates of E. coli by 14.0 to 24.8 ng/mL, which was the greatest increase among all tested organisms. Besides E. coli, PS+10 also increased DNA concentrations in lysates of A. suci by 6.2 to >8.4 ng/mL, and B. cereus by 5.2 to 6.4 ng/mL. Among fungal organisms, PS+10 only significantly (P < 0.05) improved DNA concentrations in lysates of E. phaeomuriformis, with increases ranging from 125 to >557 ng/mL. In contrast, PS+10 did not improve DNA extraction from L. monocytogenes and A. fischeri. For L. monocytogenes, PS+10 slightly increased lysate DNA concentrations in trial 3 (+6.0 ng/mL) but reduced concentrations in trial 1 (−6.0 ng/mL) and trial 2 (−3.2 ng/mL) compared with extractions using PS. Similarly, PS+10 increased lysate DNA concentrations of A. fischeri in trial 2 (+2.0 ng/mL) and 3 (+254 ng/mL) but reduced concentrations in trial 1 (−4.0 ng/mL). L. monocytogenes and A. fischeri have previously been identified as organisms with low DNA extraction efficiencies (44, 45). The lack of improvement in lysate DNA concentrations for these organisms suggests that their relative abundances may be underestimated when using the PS+10 method. DNA concentrations in lysates as a result of extractions using the PS and PS+10 methods. DNA concentration in lysates was measured by Qubit, with a limit of detection of 10 ng/mL. Lysates with less than the Qubit detection limit are indicated as <10 on the figure. Lysate DNA concentrations from extractions using PS and PS+10 were compared for each bacterial (A) and fungal (B) organism. Comparisons of DNA concentrations were performed for each trial. The difference in lysate DNA concentrations between using the PS+10 and the standard method (PS) for each trial is indicated above the bars. Negative values indicate a decrease in DNA concentrations when using the PS+10 method. Paired t-tests were conducted to identify organisms with a significant increase in lysate DNA concentration when using the PS+10 method. Asterisk (*) indicates a significant (P < 0.05) increase in (DNA concentration) extraction yield when using the modified method (PS+10). The efficacy of the PS+10 method was likely affected by cell rigidity. This was evident by the high extraction efficiency of E. coli cell, which contains a permeable cell wall that facilitates lysis and DNA release (45, 46). In addition to rigidity, cell size could also affect extractions. Smaller cells are more difficult to recover by centrifugation during DNA extraction, resulting in lower DNA yields (47). This may explain why PS+10 exhibited lower extraction efficiency for L. monocytogenes (0.5 by 2 µm) than for A. suci (1.5–1.8 by 0.9–1.0 µm) and B. cereus (1.6 by 0.9 µm) spores despite all having rigid cell structures (34–36). Differences in the extraction efficiency of PS+10 could skew the relative abundance of bacterial and fungal organisms. Therefore, better protocols are needed to enhance extraction efficiency for cells with small sizes or rigid structures, such as L. monocytogenes. The PS+10 method skewed the relative abundance of organisms The observed relative abundances of organisms were significantly (P < 0.05) different from the theoretical values (Fig. 4A). The only exception was the relative abundance of A. suci in the microbiota with a high spore low vegetative cell ratio, in which 38.9%, 39.3%, and 43.8% of A. suci was observed in trial 1, 2, and 3, respectively. These percentages were similar (P > 0.05) to the theoretical value (45.5%). Differences between the theoretical and observed relative abundances indicated that using PS+10 in DNA extraction did not produce a result that accurately reflected the proportion of organisms in the microbiota. Furthermore, we consistently recorded higher observed relative abundances of E. coli and E. phaeomuriformis than their theoretical relative abundances across all three ratios tested. For instance, in trial 1, the relative abundance of E. coli was 70.6%, 37.3%, and 70.2% in microbiota with equal, high, and low spore-to-vegetative cell ratios, and the relative abundance of E. phaeomuriformis was 82.1%, 94.2%, and 86.7% in the respective ratios. These percentages were all significantly (P < 0.05) greater than their respective theoretical relative abundances (E. coli: 25.0%, 4.55%, and 45.0%, E. phaeomuriformis: 50% for all three microbiota). The overestimation of E. coli and E. phaeomuriformis relative abundances aligned with the elevated extraction efficiency of these species as shown in Fig. 3. This suggested that the discrepancy in extraction efficiency may have contributed to the bias of relative abundance. In fact, a similar trend has been observed in organisms with lower relative abundances. For instance, the underestimation of B. cereus across microbiota corresponds to its lower extraction efficiency shown in Fig. 3. The species level relative abundances of bacterial (A) and fungal (B) organisms as a result of extracting DNA from three ratios of microbiota using the PS+10 method. These ratios include an equal ratio of bacterial spores and vegetative cells, a high ratio of bacterial spore low ratio of vegetative cell, and a low ratio of bacterial spore high ratio of vegetative cell. The theoretical relative abundances were calculated according to the cell counts through hemacytometer of each organism in the microbiota. Fungal species had equal theoretical relative abundances across all microbiota since equal concentrations of each fungal organism were included in the mixtures. Relative abundances obtained from DNA extraction and amplicon sequencing were compared with the theoretical relative abundances for each trial. Paired-end t-tests were conducted to evaluate the statistical differences between the sequence-derived relative abundances and the theoretical values. Asterisk (*) indicates the organism with a similar (P > 0.05) relative abundance as the theoretical value. qPCR was conducted to assess the PCR amplification efficiency for each organism, targeting the 16S rRNA V3-V4 or ITS 2 region using equal concentrations of DNA from each organism (C). Asterisk (*) indicates a significant (P < 0.05) difference in Ct values between organisms. The qPCR results further suggested that DNA extraction, rather than amplification via PCR, was the primary contributor to the skewed relative abundances. The Ct values generated by qPCR were used to assess the relative amplification efficiency targeting the 16S V3-V4 region or the ITS2 region for each organism. A lower Ct value indicates a higher amplification rate, which can lead to an over-estimated relative abundance for the corresponding organism (48). However, the Ct values of bacterial organisms were at a similar (P > 0.05) level ranging from 20.12 ± 0.34 to 21.65 ± 1.60, indicating that differences in amplification rates did not contribute substantially to the biases of relative abundance (Fig. 4C). Although amplification rates differed significantly between E. phaeomuriformis (20.59 ± 1.24) and A. fischeri (17.59 ± 1.25), these differences did not align with the observed differences in their relative abundances. Specifically, A. fischeri had a higher amplification rate than E. phaeomuriformis, yet the observed relative abundance of A. fischeri was consistently lower than its theoretical value. This suggested that discrepancies in PCR amplification rates may not be the major cause of relative abundance biases for fungal organisms. Together, these results implied that biases in relative abundances were mostly attributable to DNA extraction. Incomplete extractions of cells with rigid structures have been identified as the major factor in skewing relative abundance calculations (49). Previous studies have suggested that increasing the extraction efficiency of hard-to-lyse cells is key to enhancing the accuracy of relative abundances (46, 50). Although the PS+10 method increased DNA yields from bacterial spores, the improvement was not sufficient to enhance the accuracy of relative abundance assessments. These findings further highlight the need to optimize the extraction efficiency for hard-to-lyse cells. In addition, the reduced quality of DNA due to the extended bead-beating treatment may be another factor negatively affecting relative abundances. As a result, developing methods to minimize DNA shearing during extractions may improve the accuracy of relative abundance assessments. Recovery with sponge swabs followed by DNA extraction with the PS+10 method caused changes to relative abundances and microbiota composition Sample collection, processing, DNA extraction, and analysis affected the relative abundance of bacterial organisms (Fig. 5A). After following these processes, the sequencing- derived relative abundances of L. monocytogenes (trial 1: 12%, trial 2: 9%, trial 3: 17%) were significantly (P < 0.05) lower than the theoretical value (25%), whereas those of B. cereus (trial 1: 37%, trial 2: 44%, trial 3: 36%) were significantly (P < 0.05) higher than the theoretical value (25%). On the other hand, the relative abundances of A. suci (trial 1: 22%, trial 2: 24%, trial 3: 21%) and E. coli (trial 1: 29%, trial 2: 23%, trial 3: 26%) were similar to theoretical values. Interestingly, the process of sample collection, processing, DNA extraction, and sequence analysis resulted in a different relative abundance variation between the sequencing-derived and theoretical values, compared to extraction alone (Section The PS+10 method skewed the relative abundance of organisms, Fig. 4). For example, in Fig. 4, the relative abundances of B. cereus in the sequencing result were lower than the theoretical value, whereas in Fig. 5A, it was higher. Differences among sequencing results suggest that both swab recovery and DNA extraction contributed to the overestimation of B. cereus and the underestimation of L. monocytogenes relative abundances. Relative abundances (A) changes of bacteria organisms following cell recovery, DNA extraction, and amplicon sequencing of microbiota inoculated onto stainless steel surfaces at equal cell ratios. The theoretical relative abundances were calculated according to the cell counts using a hemacytometer. Relative abundances calculated from each sequencing trial were compared with the theoretical relative abundances. Paired-end t-test was conducted to evaluate the statistical differences between the observed and theoretical values. “+” indicates organisms with significantly higher relative abundance and “−” indicates organisms with significantly lower relative abundance compared to the theoretical value. Surface microbiota was also plated on selective media and enumerated. The relative abundance of each bacterium in the inoculated and recovered microbiota was determined from colony counts (B). Relative abundances of organisms in the recovered microbiota were compared with those in the microbiota inoculated on surfaces. Because the relative abundances in the inoculated microbiota varied between trials, the results are presented separately for each trial. Paired-end t-tests were conducted to evaluate the statistical differences in relative abundances. “+” indicates organisms with significantly higher relative abundance, and “−” indicates organisms with significantly lower relative abundance in the recovered microbiota. The same analyses and comparisons were made between the fungal communities (C and D). To specifically assess the impact of swab recovery, we quantified viable cells of each bacterial species using selective media and compared counts between the inoculated and recovered microbiota. The results indicated that bacterial relative abundances differed between the inoculated and recovered microbiota (Fig. 5B). These changes also corresponded with the recovery efficiency of individual organisms (Table 1). For instance, E. coli, an organism with low recovery efficiency (2.9% ± 3.0%), showed a decrease from the inoculated level of ~26% to 10% in trial 1 and trial 2 and 5% in trial three based on the recovered microbiota. Conversely, B. cereus, an organism with high recovery efficiency (94.9% ± 5.0%), showed an increase from the inoculated level of 7% to 25% in trial 1, 11% to 35% in trial 2, and 5% to 44% in trial 3. These trends confirmed that recovery partially contributed to the skewed relative abundance observed in sequencing results. Notably, L. monocytogenes was underestimated due to the impact of DNA extraction (Fig. 3 and 4) and swab recovery (Fig. 5A; Table 1). Consequently, food safety and quality concerns associated with these organisms may be challenging to address when evaluating surface microbiota with amplicon sequencing (51, 52). Therefore, research aimed at improving swab recovery and DNA extraction should place emphasis on pathogens such as L. monocytogenes. The relative abundance of the fungi also differed from theoretical values (Fig. 5C). The sequencing-derived relative abundances of fungal organisms were significantly (P < 0.05) different from those based on theoretical composition (Fig. 5C). The relative abundances of E. phaeomuriformis (trial 1: 39%, trial 2: 38%, trial 3: 40%) in the sequencing results were significantly lower (P < 0.05) than the theoretical value (50%), while those of A. fischeri (trial 1: 60%, trial 2: 63%, trial 3: 60%) were significantly higher (P < 0.05) (Fig. 5C). When comparing viable cell concentrations in the inoculated and recovered microbiota, we observed a significantly higher (P < 0.05) relative abundance of E. phaeomurifomis and a significantly lower (P < 0.05) relative abundance of A. fischeri in the recovered microbiota (Fig. 5D). The relative abundance of E. phaeomuriformis increased from 98.0% to 99.9% in trial 1, 98.0 to 99.5% in trial 2, and 99.0% to 99.4% in trial 3 between the inoculated and recovered microbiota. In contrast, A. fischeri decreased from 2.00% to 1.00% in trial 1, 2.00% to 0.50% in trial 2, and 1.00% to 0.60% in trial 3 between the inoculated and recovered microbiota. It remains unclear why the calculated relative abundance of E. phaeomuriformis from amplicon sequencing was lower than the theoretical value even though biases introduced by both recovery (Fig. 5D) and DNA extraction (Fig. 4B) were expected to overestimate its abundance. One possible explanation is that current bioinformatic analysis pipelines (e.g., DADA2), which are primarily designed for 16S rRNA gene analysis, can generate errors when clustering fungal ITS sequences into taxonomic units (53, 54). These errors may have caused the sequencing-derived values to be different from the theoretical value (55–57). Potential errors in ITS sequence analysis highlight the need to improve bioinformatic methods for analyzing fungal ITS data. Conclusion Amplicon sequencing analysis distorted the relative abundance of spoilage and pathogenic organisms. This bias was attributed to differences in variability in cell recovery and DNA extraction among microbes. Variations in physiological properties and cell integrity likely contributed to these differences. To improve the accuracy of relative abundance estimates, future studies should investigate cellular mechanisms affecting recovery and DNA extraction and develop strategies to mitigate their impact, particularly for the pathogen Listeria monocytogenes. Mechanisms associated with Gram-positive organisms and fungi need greater attention due to their lower recovery and DNA extraction efficiency. Such efforts will improve relative abundance estimates and enable more accurate identification of key microbiota using amplicon sequencing.