Long-term stability at −80°C of oral wash and saliva samples for microbiome analyses

ABSTRACT

Large-scale prospective biological studies necessitate the storage of oral samples for numerous years to accrue adequate sample sizes. However, there is minimal research on the impact of long-term storage of oral samples on the oral microbiome. We investigated the freezer stability over 5 years of the oral microbiome measured from oral wash and saliva samples to provide insight for future microbiome analyses of stored oral samples. Healthy participants provided oral wash and saliva samples using Scope mouthwash and the OMNIgene ORAL collection device, respectively. DNA was extracted from an aliquot of each sample type at baseline, the V4 region of the 16S rRNA gene sequenced, and additional aliquots were then similarly extracted and sequenced after being stored for approximately 1 month, 12 months, and 5 years after collection. Intraclass correlation coefficients (ICC) and 95% confidence intervals (CI) were calculated for 4 alpha-diversity metrics, the first 2 principal coordinates of four beta-diversity matrices, and the 14 most abundant genera. The alpha diversity and beta diversity ICCs for both sample types remained stable over 5 years. For example, the 5-year Shannon index ICCs were 0.94 (95% CI: 0.88, 0.98) and 0.90 (95% CI: 0.72, 0.95) for oral wash and saliva samples, respectively. The ICCs for the relative abundances of the examined genera during the 5 years of freezer storage were also generally stable. Both oral wash and saliva samples were relatively stable for diversity metrics and relative abundance after 5 years when stored at −80°C.

IMPORTANCE

Large, prospective studies will likely need to store biospecimens for many years in the freezer prior to DNA extraction and sequencing for analyses considering the association between the microbiome and specific health conditions. In this study, we demonstrated that oral wash using Scope mouthwash and saliva specimens in the OMNIgene ORAL kit have generally stable microbiome communities for up to 5 years at −80°C.

INTRODUCTION

The oral microbiome is the collection of microorganisms that live in the oral cavity and is one of the largest and most diverse microbial communities in the human body, second only to the gut (1). In recent years, an increasing number of studies have focused on the oral microbiome and its relationship with a multitude of diseases, including cancer, diabetes, rheumatoid arthritis, and Alzheimer’s disease (2). To better understand the impact of the oral microbiome on disease risk, there is a strong need for large-scale prospective studies where oral samples are collected prior to disease diagnosis to demonstrate the temporality of associations (2).

Previous studies have evaluated how different oral collection methods impact the measured microbial communities to support large-scale prospective microbiome studies and concluded that study comparisons should be made using the same collection method (3, 4). However, samples in these prospective studies are typically stored for long periods of time while outcomes accrue, and little research to evaluate the impact of long-term storage of oral samples on the oral microbiome has been conducted. Additionally, acknowledging and understanding the potential impact of sample storage time on microbial composition is crucial to draw accurate conclusions on the relationship between the oral microbiome and health.

Prior research on the long-term stability of the oral microbiome during storage has shown conflicting findings; some studies have found that the oral microbiome is relatively stable for time periods up to 6 months, while others have found that important metrics for examining the oral microbiome, such as relative abundance, are unstable (5). However, most of these studies examined temporal stability for periods well under a year, and only dental plaque and saliva samples were investigated. A more recent study examining the stability of two oral collection methods over 2 years found that most oral microbiome diversity measures were generally stable at −80°C (6). However, longer durations of recruitment and storage during follow-up are likely for many prospective cohort studies. Thus, we investigated the freezer stability of oral wash collected using Scope mouthwash and saliva collected using the OMNIgene ORAL kit over the course of 5 years to provide further details regarding the stability of oral samples in the freezer for future microbiome analyses.

MATERIALS AND METHODS

Study population

We recruited 30 adult National Cancer Institute (NCI) employees from June to August 2017. To be included in the study, individuals had to be at least 18 years of age or older, not taken antibiotics anytime in the past 3 months, and willing to provide oral samples over the course of 2 weeks. Participants who qualified and provided written informed consent were scheduled to provide samples on 6 days over the course of 2 work weeks. At the first visit, the participant filled out a short electronic questionnaire which included questions on potentially important covariates including demographics, behavioral information, and medical factors, such as self-reported height and weight to be able to calculate body mass index (BMI). This study was approved by the Special Studies Institutional Review Board of the NCI.

Oral sample collections

Participants were instructed to not perform any dental hygiene procedures, eat or drink anything other than water, chew gum, consume throat lozenges or candy, or smoke cigarettes or chew tobacco for at least 12 h before each sample collection.

At each visit, participants first provided up to two saliva samples using the OMNIgene ORAL collection device (DNA Genotek, Canada) following the manufacturer’s protocol. After the saliva collection, participants provided up to two oral wash samples by swishing and gargling Scope mouthwash for 5 s each for a total of 30 s as described previously (7). When two Scope mouthwash samples were provided during the same visit, the collections were separated by approximately 10 min. The number of samples collected on each day is outlined in Table S1. After collection, samples remained at room temperature until transfer to the Cancer Genomics Research (CGR) Laboratory every other day.

Oral sample processing

Upon receipt at CGR, the OMNIgene ORAL samples remained at room temperature and the Scope mouthwash samples were stored at 4°C. Once all samples from one individual were received, the Scope mouthwash and OMNIgene ORAL samples were processed as described in detail previously (8). In brief, the eight Scope mouthwash samples were centrifuged at 1,500 × g (2,650 rpm) for 12 min. The supernatant was removed, and the buccal cell pellets were resuspended with 3 mL of 1× Tris EDTA buffer solution and mixed. Then, the eight resuspended buccal cell samples were pooled, well-mixed, and aliquots of 900 µL were created to generate at least 25 aliquots. For OMNIgene ORAL, the nine samples were mixed and incubated at 50°C for 1 h in a water bath. Then, the OMNIgene ORAL samples from one individual were combined into one vial and mixed. Aliquots of 1,000 µL were created from the pooled sample to generate at least 25 aliquots. The aliquots were stored at −80°C.

DNA extraction, PCR amplification, and 16S rRNA gene sequencing

After all samples were collected from all individuals in the study, DNA was extracted from an aliquot from each individual of the Scope mouthwash and OMNIgene ORAL described in detail below. Additional aliquots were extracted after being stored for approximately 1 month, 12 months, and 5 years after collection. DNA from quality control samples including artificial communities, robogut samples, and extraction blanks were also extracted (9). The quality control samples included the ZymoBIOMICS (D6300: ZymoBIOMICS Microbial Community Standard, D6305: ZymoBIOMICS Microbial Community DNA Standard, D6310: ZymoBIOMICS Microbial Community Standard II [Log distribution]) and artificial gut and oral communities from the MicroBiome Quality Control Project (MBQC) (9). After DNA extraction, the DNA was stored at −80°C until sequencing.

For all DNA extractions, except for year 5, we used the MagMAX Multi-Sample Ultra kit (ThermoFisher Scientific, Waltham, MA), while at year 5, we used the MagMAX Multi-Sample Ultra 2.0 kit (ThermoFisher Scientific, Waltham, MA), both on the KingFisher Flex (ThermoFisher Scientific, Waltham, MA) instrument. The MagMAX Multi-Sample Ultra 2.0 kit is an enhanced, faster, and more versatile version of the other MagMAX DNA kit, but both versions were assumed to yield similar results. For DNA extraction as previously described (10), a 750 µL aliquot was centrifuged for 20 min at 3,500 rpm to pellet and 700 µL of supernatant was removed. To the remaining cell pellet, lysis buffer (300 µL) was added and mixed for 6 min. Then, isopropanol (180 µL) was added and mixed for 5 min, followed by MagMAX DNA Binding Beads (20 µL) and mixed for another 5 min to capture DNA on the bead surfaces. Bound DNA was collected and washed in Wash 1 Buffer (150 µL) for 1 min, High Salt Wash (150 µL) for 1 min, and Wash 2 Buffer (150 µL) for 1 min. The beads were dried for 2 min before eluting purified DNA in the elution buffer (150 µL) for 5 min at 80°C. After extraction, the DNA was stored at 4°C and quantified using the Quant-iT PicoGreen dsDNA assay (ThermoFisher Scientific, Waltham, MA).

The V4 region of the 16S rRNA gene was PCR amplified and sequenced at baseline, 1 month, 12 months, and 5 years using primer designs and custom sequence primers as previously described (11). Twenty nanograms of DNA was amplified in two PCRs per sample, consisting of 10 ng (10 µL) of DNA, 10 µL of 2.5 × 5Prime HotMasterMix (VWR, Radnor, PA), 3 µL of MBG Water, and 2 µL of the 5 µM 16S rRNA v4 (515f-806r) barcoded primer mix, comprised of equimolar forward and reverse primer pairs targeting the V4 region of the 16S rRNA gene, and amplified using the following conditions: 94°C hold for 3 min, denature at 94°C for 45 s, anneal at 50°C for 1 min, extend at 72°C for 1 min 30 s for 25 cycles, followed by a 72°C hold for 10 min. Replicate PCRs per sample were pooled and purified using a 1:1 AMPure XP (Beckman Coulter Genomics, Danvers, MA) ratio. Amplified sample libraries were quantified using Quant-iT PicoGreen dsDNA Reagent (ThermoFisher Scientific, Waltham, MA) and up to 192 samples, with unique barcoded adapters, were combined in equal amounts (100 ng each) and pools normalized to 10 nM with Buffer EB for sequencing on the MiSeq (Illumina) with v2 chemistry, using paired-end 2 × 250 bp reads.

Bioinformatics

Our bioinformatic workflow is available on GitHub (https://github.com/NCI-CGR/QIIME_pipeline/tree/QIIME_pipeline_dev). In brief, sequence data were bioinformatically processed using QIIME 2 2024.5 (12). We first demultiplexed the sequences using bcl2fastq from Illumina and then quality-filtered the sequences using DADA2 (13) with a Phred quality score of 33 to generate amplicon sequence variants (ASV). After error correction and removal of chimera and phiX sequences, the remaining reads were used for taxonomic classification. We performed taxonomic classification with a naive Bayes classifier trained on the SILVAv138.1 99% database for the V4 region (14). We removed reads which did not align to bacteria at least to the phylum-level. The relative abundance and presence of bacterial taxa were then calculated from the phylum to genus level.

Using the generated ASVs, we calculated alpha and beta diversity metrics after rarefaction to 20,000 reads per sample. Two samples in this study were excluded due to having fewer than 20,000 reads. Alpha diversity measures included observed ASVs, Shannon index, Pielou’s evenness, and Faith’s phylogenetic diversity. For beta diversity, we generated Bray-Curtis, Jaccard, and unweighted, generalized, and weighted UniFrac distance matrices. Principal coordinate analysis (PCoA) vectors were calculated for each distance matrix. We also estimated the genus-level relative abundance without rarefaction, but restricted to individuals with at least 20,000 reads. All genus-level relative abundance testing and comparisons were limited to the 14 genera with average relative abundances greater than 0.01 (i.e., 1%) for the oral wash or the saliva sample.

Quality control analysis

Quality control samples in this study included 17 artificial communities, 11 robogut samples, 15 extraction blanks, 13 PCR water blanks, and 13 PCR no template control blanks. After rarefaction, all of the artificial community and robogut samples remained as well as two extraction blanks (13% of the extraction blanks). Prior to rarefaction, the median number of reads for the extraction blanks was 364 with a range of 0 to 83,256. Upon further examination of the PCoA plots for the beta diversity metrics, the remaining two extraction blanks appeared to group with the study samples, so these likely were related to well-to-well contamination as both the artificial communities and robogut samples remained clearly distinct from the study samples (Fig. S1). For the MBQC artificial community, the coefficients of variation for alpha diversity across batches were 0.15, 0.02, 0.05, and 0.12 for observed ASVs, Shannon Index, Pielou’s evenness, and Faith’s phylogenetic diversity, respectively, representing low inter-batch variability.

Statistical analysis

We presented the demographic characteristics of the participants and calculated means and standard deviations (SD) for continuous variables and counts and proportions for categorical variables. We estimated the percent variability explained by the individual, the oral sample collection method, and the duration of freezing on the beta diversity matrices using the adonis2 function in the R vegan package. We also created a plot of all estimated alpha diversity estimates by individual over time to visualize changes based on time in the freezer.

We calculated interclass correlation coefficients (ICC) using a linear mixed effects model without adjustment for covariates to estimate the freezer stability of the microbiome determined from the saliva and oral wash samples. The ICC was calculated using the following formula: let 𝑦_𝑖j be a microbiome metric for observation j and subject 𝑖. We fit a linear mixed model , $y_{ij}=a+b_i+e_{ij}$, where $b_i\sim N(0,{\sigma }_b^2)$ and $e_{ij}\sim N(0,{\sigma }_w^2)$. The ICC is then calculated as $\frac{{\sigma }_b^2}{{\sigma }_b^2+{\sigma }_w^2}$.

We calculated ICCs for the alpha and beta diversity metrics and genus-level relative abundances for the 14 genera with an average relative abundance greater than 0.01 (i.e., 1%) for at least one of the sample types. For relative abundance, we also calculated the ICCs for the centered log-ratio (CLR) transformed abundances. We used 1,000 bootstrap samples to generate 95% confidence intervals. We considered an ICC value less than 0.50 as poor, 0.50–0.75 as moderate, 0.75–0.90 as good, and greater than 0.90 as excellent reliability (15). Analyses were conducted using R v4.4.3, and the code used to generate the tables and figures is available on GitHub (https://github.com/EVogtmann/oral-microbiome-freezer-stability).

RESULTS

A total of 30 participants completed the sample collection for this study. The average age of the participants was 40.8 years (SD = 9.89) and about two-thirds were female (n = 19; 63.3%); the participants were highly educated with 83.3% (n = 25) having a master’s or doctoral degree (Table S2). Out of 237 study samples included for DNA extraction and sequencing, 235 samples had at least 20,000 reads and were retained for downstream analyses. The included samples had a minimum of 20,293 reads and a maximum of 101,179 reads with an average of 56,723 reads per sample.

The average genus-level relative abundances at each DNA extraction timepoint for the 14 genera that met the average relative abundance criteria of greater than 0.01 (i.e., 1%) are shown in Fig. 1. As expected, the most common genera were Streptococcus and Prevotella for both sample types. The relative abundance for each participant at each timepoint is visually shown in Fig. S2 and S3 for oral wash and saliva, respectively. Each participant had similar genus-level composition over the different DNA extraction timepoints for both collection methods. The average relative abundances for the 14 genera are shown in Table S3 for buccal cells and Table S4 for saliva.

Fig 1.

Average genus-level relative abundance for DNA extracted at baseline, after 1 month, 12 months, and 5 years from collection of the 14 genera with at least a relative abundance of 0.01 (i.e., 1%) grouped by sample collection type (i.e., oral wash/buccal cells or saliva).

When we estimated the percent variability explained in beta diversity by individual, oral sample type, and duration of freezing, the majority of the variability was explained by inter-individual variability ranging from 79.3% for Bray-Curtis to 62.5% for generalized UniFrac. Duration of freezing explained a maximum of 3.1% of the variability in the weighted UniFrac matrix and was less for all other beta diversity indices (Fig. 2).

Fig 2.

Percent variability explained in beta diversity (i.e., Bray Curtis, Jaccard, unweighted Unifrac, generalized Unifrac, and weighted Unifrac) by specimen storage time (i.e., time in the freezer), source material (i.e., oral wash or saliva), and subject.

For alpha and beta diversity metrics, the microbial communities from both the oral wash and the saliva samples remained consistent over 5 years (Fig. 3 and Table S3 and S4). The four alpha diversity metrics were particularly stable showing minimal to no change when compared with the baseline extracted samples for both oral wash and saliva samples demonstrated by good and excellent reliability of ICC values. For example, the 5-year Shannon index ICCs were 0.94 (95% CI: 0.88, 0.98) and 0.90 (95% CI: 0.72, 0.95) for oral wash and saliva samples, respectively. Individual trends for alpha diversity for the oral wash and saliva samples are presented in Fig. S4. For beta diversity, the ICCs for the first PCoA vectors from generalized UniFrac and weighted UniFrac had moderate reliability for oral wash at 5 years (generalized UniFrac PCoA1 ICC: 0.73 [95% CI: 0.40, 0.89]; weighted UniFrac PCoA1 ICC: 0.76; [95% CI: 0.43, 0.91]), but all of the remaining beta diversity ICCs had good or excellent reliability.

Fig 3.

Interclass correlation coefficients and 95% confidence intervals for alpha and beta diversity metrics from oral wash/buccal cells and saliva samples with DNA extraction at 1 month, 12 months, and 5 years compared to the baseline DNA extraction. ICC = interclass correlation coefficient; ASV = amplicon sequence variant; PC = principal coordinate analysis vector.

At the genus level, we observed relatively high ICCs for the relative abundances of many of the genera during the 5 years of freezer storage with only a couple notable exceptions (Fig. 4). Considering the 5-year ICCs for the oral wash samples, three genera (Neisseria, Actinomyces, and Haemophilus) had excellent reliability, six genera (Megasphaera, Streptococcus, Leptotrichia, Alloprevotella, Rothia, and Gemella) had good reliability, and five genera (Prevotella, Fusobacterium, Porphyromonas, Veillonella, Granulicatella) had moderate reliability. The highest 5-year ICC for oral wash was 0.95 (95% CI: 0.89, 0.98) for Neisseria and the lowest was 0.57 (95% CI: 0.17, 0.82) for Granulicatella. For the saliva samples, two genera (Neisseria and Megasphaera) had excellent reliability, three genera (Prevotella, Alloprevotella, and Haemophilus) had good reliability, six genera (Streptococcus, Fusobacterium, Leptotrichia, Porphyromonas, Rothia, and Veillonella) had moderate reliability, and three genera (Gemella, Actinomyces, and Granulicatella) had poor reliability after 5 years in the freezer. Similar to the oral wash samples, the highest 5-year ICC for saliva was 0.97 (95% CI: 0.93, 0.99) for Neisseria and the lowest was 0.37 (95% CI: 0.01, 0.75) for Granulicatella (Table S3 and S4). After CLR transformation of the relative abundances, the ICCs were generally slightly increased compared to the untransformed ICCs and were also largely stable over the 5 years (Table S3 and S4).

Fig 4.

Interclass correlation coefficients and 95% confidence intervals for genus-level relative abundances from oral wash/buccal cells and saliva samples with DNA extraction at 1 month, 12 months, and 5 years compared to the baseline DNA extraction. The genera presented represent the 14 taxa that met the average relative abundance criteria of greater than 0.01 (i.e., 1%). ICC = interclass correlation coefficient

DISCUSSION

In this study of samples collected from 30 healthy individuals, we found that the oral microbiome measured from oral wash samples collected using Scope mouthwash and saliva samples stored in the OMNIgene ORAL preservative are generally stable in a −80°C freezer for at least 5 years. The four alpha diversity metrics showed minimal changes after freezing and the ICC estimates for most measures of alpha and beta diversity were high. Similarly, the ICCs for the relative abundances of the 14 most abundant genera also generally had good or excellent reliability, but a handful demonstrated moderate to poor ICC reliability after 5 years, particularly for the saliva samples. After CLR transformation of the relative abundances, there was marginal improvement of the ICC estimates.

The findings from this study align with the limited previous research examining freezer stability of samples collected to measure the oral microbiome. A previous study on saliva of four healthy adults found that DNA concentration and the quantities of most bacterial species remained stable for 6 months (16). Another study of 51 individuals similarly found that the oral microbiome measured from the same collection methods we tested and stored over the course of 2 years had relatively high freezer stability with generally high ICCs for diversity metrics and abundant genera (6). We did find some specific differences, such as the previous study calculated low ICCs for Faith’s phylogenetic diversity estimated from the oral wash samples (ICC = 0.23) over 2 years, but we found high ICCs for this measure up to 5 years (ICC = 0.94) (6). For the relative abundance of genera, all oral wash ICCs showed at least good reliability, but three genera, Actinomyces, Gemella, and Granulicatella, had poor ICCs for untransformed relative abundances, but good reliability after CLR transformation. Notably, these three taxa with the lowest relative abundance ICCs all are gram-positive facultative anaerobes although Rothia and Streptococcus, also gram-positive facultative anaerobes, exhibited higher reliability ICCs for both untransformed and CLR transformed relative abundances (17–21).

Understanding the impact of freezer storage on microbial composition is critical for large epidemiological studies that require storage for potentially decades. For example, prospective studies on diseases, like cancer, can span multiple years of recruitment and many years of follow-up while sufficient cases accrue, so accounting for potential changes in microbial composition due to freezer storage duration is necessary to make accurate conclusions. We investigated two common collection methods since the sample type and composition of the fixative could impact freezer stability. Scope mouthwash is a predominately water-based mixture which contains ethanol (ethanol accounts for 5%–10% of total weight). OMNIgene ORAL is a proprietary mixture that contains water and sodium dodecyl sulfate (SDS) (SDS accounts for 1%–10% of total weight) and is considered a non-ethanol based preservative method. As ethanol-based methods (Scope) may allow for the continued growth of ethanol-tolerant bacteria after collection, it is possible that non-ethanol preservation buffers (OMNIgene ORAL) may better limit growth after collection and preserve original microbial composition. However, we previously observed no large differences between ethanol and non-ethanol mouthwash (10), and the oral wash samples and saliva samples had generally similar stability in our study. Based on our findings, future prospective analyses should consider matching cases and controls on time in the freezer or stratifying analyses on time in freezer to help improve comparability.

This study has multiple strengths. To our knowledge, this is the first study of freezer stability longer than 2 years. We included multiple DNA extraction timepoints to better ascertain timing of changes due to freezer storage. In addition, we used a consistent protocol for most of the sample processing (e.g., PCR amplification, sequencing) over time which reduces potential deviations in microbiome composition due to laboratory-derived factors. This study also had limitations. We used a consistent DNA extraction kit, the MagMAX DNA Multi-Sample Ultra kit for all extractions except for year 5 where we used the updated MagMAX DNA Multi-Sample Ultra 2.0, so it is difficult to determine if any differences after 5 years are related to the change in kit or due to storage time. However, few differences were observed after 5 years, so the updated kit was likely similar to the original kit. In addition, these DNA extraction kits use enzymatic lysis without bead beating. Therefore, it is possible that freezer stability differences may exist based on DNA extraction method. Another limitation was the use of 16S rRNA V4 gene sequencing as we were unable to accurately identify species-, strain-, or gene-level information. Thus, this study was unable to determine if specific species or strains within a genus were impacted by duration of freezer storage. Finally, the study only included employees working at one research institution with likely better oral health than the general population, so genera that might be more prevalent in other populations, younger individuals, or associated with poor oral health were not examined.

To conclude, this study provides evidence that oral wash samples collected using Scope mouthwash and saliva collected using the OMNIgene ORAL kit have high stability in a −80°C freezer for at least 5 years for diversity metrics and for genera with an average relative abundance above 1%. Regardless, matching cases and controls on freezer storage length or stratifying analyses on freezer storage length should be considered to help improve comparability. Future studies should investigate the freezer stability of the oral microbiome for even longer periods of time (e.g., 10 years) as well as use metagenomic sequencing to evaluate species-, strain-, and gene-level freezer stability.

DATA AVAILABILITY

Data are available in SRA under project number PRJNA1338219. Our bioinformatic workflow and statistical analysis code are available on GitHub (Bioinformatics: https://github.com/NCI-CGR/QIIME_pipeline/tree/QIIME_pipeline_dev; Statistical analysis: https://github.com/EVogtmann/oral-microbiome-freezer-stability).

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/spectrum.00573-26.

Supplemental figure legends. spectrum.00573-26-s0001.docx.

Figure legends for Fig. S1 to S4.

Fig. S1. spectrum.00573-26-s0002.png.

Principal coordinates analysis plots of the study samples and quality controls for the 5 beta diversity metrics. The points are colored by type to indicate whether it is a study sample or quality control. For the quality controls, the artificial communities are further identified by shape to indicate the different communities.

Fig. S2. spectrum.00573-26-s0003.png.

Genus-level relative abundances from oral wash samples of the 14 most common genera grouped by participant ID at four timepoints (baseline, one month, 12 months, and 5 years).

Fig. S3. spectrum.00573-26-s0004.png.

Genus-level relative abundances from saliva samples of the 14 most common genera grouped by participant ID at four timepoints (baseline, one month, 12 months, and 5 years).

Fig. S4. spectrum.00573-26-s0005.png.

Individual trends in alpha diversity estimated from oral wash/buccal cell samples and saliva samples extracted at four timepoints (baseline, one month, 12 months, and 5 years). ASV = amplicon sequence variant.

Supplemental tables. spectrum.00573-26-s0006.xlsx.

Tables S1 to S4.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.