Consistent and reproducible long-term in vitro growth of health and disease-associated oral subgingival biofilms

Background Several in vitro oral biofilm growth systems can reliably construct oral microbiome communities in culture, yet their stability and reproducibility through time has not been well characterized. Long-term in vitro growth of natural biofilms would enable use of these biofilms in both in vitro and in vivo studies that require complex microbial communities with minimal variation over a period of time. Understanding biofilm community dynamics in continuous culture, and whether they maintain distinct signatures of health and disease, is necessary to determine the reliability and applicability of such models to broader studies. To this end, we performed next-generation sequencing on biofilms grown from healthy and disease-site subgingival plaque for 80 days to assess stability and reliability of continuous oral biofilm growth. Results Biofilms were grown from subgingival plaque collected from periodontitis-affected sites and healthy individuals for ten eight-day long generations, using hydroxyapatite disks. The bacterial community in each generation was determined using Human Oral Microbe Identification by Next-Generation Sequencing (HOMINGS) technology, and analyzed in QIIME. Profiles were steady through the ten generations, as determined by species abundance and prevalence, Spearman’s correlation coefficient, and Faith’s phylogenetic distance, with slight variation predominantly in low abundance species. Community profiles were distinct between healthy and disease site-derived biofilms as demonstrated by weighted UniFrac distance throughout the ten generations. Differentially abundant species between healthy and disease site-derived biofilms were consistent throughout the generations. Conclusions Healthy and disease site-derived biofilms can reliably maintain consistent communities through ten generations of in vitro growth. These communities maintain signatures of health and disease and of individual donors despite culture in identical environments. This subgingival oral biofilm growth and perpetuation model may prove useful to studies involving oral infection or cell stimulation, or those measuring microbial interactions, which require the same biofilms over a period of time. Electronic supplementary material The online version of this article (10.1186/s12866-018-1212-x) contains supplementary material, which is available to authorized users.

between individual donors, making it a useful model for long-term study of biofilm each biofilm source to ensure growth (four biofilm lineages). The HA disks were coated for ten generations, HB1 from 60 to 56, DB2 from 116 to 64, DB3 from 83 to 52, while the 281 number of species in HB2 increased from 56 to 75 (Table S1) 66,206 to 31,677,DB2 from 49,189 to 20,239,and DB3 from 28,597 to 28,122. 291 However, some species that disappear in a specific generation were again detected in 292 subsequent remaining generations. The disappearing-one-generation-reappearing-the-next 293 pattern seen for many of the species detected suggests that they are present through the 294 entire study, and are simply not detected, which may be a result of low abundance in the 295 biofilm, DNA extraction and amplification bias, different sequencing depths, or sequencing 296 errors.

297
Our independent QIIME analysis classified a comparable number of reads as 298 HOMINGS (3226031 QIIME, 2383617 HOMINGS), and QIIME had nearly half the 299 number of unmatched reads (808149 QIIME, 1869947 HOMINGS) (Table S2). However, 300 QIIME taxonomic assignment had lower resolution, identifying approximately half the 301 number of species in each biofilm, demonstrating the superior taxonomic assignment 302 provided by the species-specific probes used in HOMINGS. The taxonomic profile 303 identified by QIIME was similar to that of HOMINGS, with the majority of assignments 304 belonging to genera and species known to inhabit the oral biofilm.

306
Community diversity within healthy and diseased site-derived biofilm lineages 307 There are conflicting reports on the differences in microbial diversity at healthy 308 tooth sites and those experiencing periodontal disease [5,44,45]. Therefore, we calculated 309 community diversity scores to assess differences between healthy-and LAP site-derived 310 biofilms using Faith's phylogenetic distance, and observed species. Rarefaction curves show slightly higher diversity in LAP site-derived biofilms than healthy site-derived 312 biofilms for each metric, which were significantly different despite overlapping confidence 313 intervals ( Fig. 2A). Further substantiation of this observation was seen in rarefaction 314 curves using the Chao1 and Shannon metrics (Additional file 2: Figure S1A). Diversity 315 metrics were significantly different between healthy and disease site-derived biofilms at 316 20,000 reads, (Phylogenetic distance p ≤ 0.05; Observed species p ≤ 0.001, Shannon p ≤ 317 0.001) with the exception of Chao1 (Fig. 2B, Additional file 2: Figure S1B). Outliers in the 318 boxplots for each diversity metric are not consistently the same samples, nor are they all 319 early or late generation samples. As diversity in generation 1 is higher than the remaining 320 generations and may be responsible for the significant difference between healthy and 321 disease site-derived biofilms, we repeated our alpha diversity analyses on tables that 322 excluded generation 1 biofilms. We observed that the average alpha diversity scores of 323 healthy and disease site-derived biofilms remain significantly different by observed species 324 and Shannon diversity metrics but not by Faith's phylogenetic distance ( Figure S2).

325
To see whether diversity changes substantially through repeated dispersal and re-326 inoculation of the biofilms, we averaged the alpha diversity scores of both healthy site-327 derived biofilms and both disease site-derived biofilms, and plotted rarefaction curves for 328 individual generations of healthy and disease-site derived biofilms (Fig. 2C, Additional file 329 2: Figure S1C). These curves reveal that for each diversity metric, between the first (black 330 line) and second (red line) generations, diversity increases slightly in healthy site-derived 331 biofilms yet decreases slightly in disease site-derived biofilms, while there is no substantial 332 change in diversity between generations after the second generation in either healthy or 333 disease site-derived biofilms. Additionally, we compared the alpha diversity scores 334 between healthy and disease-site derived biofilms for each generation at 20,000 reads to 335 understand if the overall significant difference in diversity is manifested at each generation 336 (Fig. 3A,C; Additional file 3: Figure S3A,C). There were no significant differences in 337 diversity score between healthy and disease site-derived biofilms for any generation at 338 20000 sequences/sample (Fig. 3B,D; Additional file 3: Figure S3B,D), but having only two 339 samples in each group is likely insufficient power to detect statistical differences.

340
The consistency of biofilm community composition between generations was tested 341 by Spearman's correlation for each lineage. The correlation coefficients demonstrated significant positive and strong correlation between generations (correlation coefficient 0.5-343 0.7, p < 0.001), which was strongly supported by statistical tests (Table 1). Because low-344 abundance taxa were inconsistently detected between generations, we repeated the tests 345 using a species table containing only species present at >0.1% abundance (Table 1). After 346 removing low abundance species, the correlation coefficients between generation rose in 347 each lineage (0.6-0.9, p < 0.001), except between the first and tenth generations, however, 348 still maintaining significant correlations, as observed before. Community diversity between healthy and diseased site-derived biofilm lineages 358 We also assessed community diversity differences between LAP site-and healthy 359 site-derived biofilms (beta-diversity) to determine if the biofilms maintained a distinct 360 community structure throughout growth in identical media. Principal coordinates analysis 361 was used to determine separation of healthy and disease site-derived biofilms based on the 362 weighted UniFrac metric. A clear separation of healthy and disease site-derived biofilms is 363 seen based on the weighted UniFrac distance (Fig. 4), and the separation is significant as 364 determined by adonis (p ≤ 0.001), indicating that there is significant community diversity 365 between the two groups. Additionally, the Sorensen metric separated healthy and disease 366 site-derived biofilms (p ≤ 0.001) as well as the two LAP site-derived biofilm lineages 367 (Additional file 4: Figure S4), which suggests the initial plaque inocula from the two 368 distinct disease patients had different microbial communities, which in fact remained 369 distinct despite culture in the same media, while the initial healthy site plaque inocula from 370 individual donors contained similar communities.

371
To test the strength of these results we also performed jack-knifed beta diversity 372 using weighted UniFrac and Sorensen metrics, and performed PCoA on the matrices 373 (Additional file 5: Figure S5A  culture, we did not have any remaining inocula from 3 of our samples to sequence.

488
However, a comparison with initial plaque inocula was previously done for our model 489 [28,33], and our goal was instead to demonstrate consistent and long-term reproducible 490 biofilm growth. There are no in vitro oral biofilm growth models that support greater than 491 80% of the species present in the initial plaque or saliva inoculum, and based on found that supplementing growth media with high amounts of serum was necessary to 501 support growth of Treponema species, and further studies of plaque biofilm ecology are 502 clearly needed to improve in vitro models. Despite this limitation, we were able to 503 demonstrate here that serially-grown biofilms could consistently reproduce similar, stable 504 communities, through several generations.

505
It was recently shown in E. coli that cell lineage history was a stronger predictor of 506 overall cell metabolic activity than spatial proximity [48]. If this holds true for other 507 bacterial species, especially when growing in complex biofilm communities such as these, Therefore, dental plaque inocula may be more likely develop in vitro biofilms that maintain 520 metabolic characteristics of healthy and disease site plaque, and thus, this warrants further 521 investigation.

522
Although it is uncommon for in vitro oral biofilm studies to report biofilm 523 community composition beyond 72 hours of growth [29,31], allowing the biofilms to grow 524 to maturity (seven to ten days for this model [33]) may be critical to maintaining 525 community profiles for propagating the biofilms, as longer growth periods allow slow-526 growing, low-abundance and "late colonizer" species to reach sufficient levels that they are 527 not out-competed by rapid-growing species, or lost by chance from the small aliquots used 528 to inoculate subsequent generations. Saliva-and plaque-derived biofilms as well as subgingival plaque develop continuously over at least seven days [28,30,49], which 530 supports using more mature biofilms (older than 72 hours) in studies of biofilm-host 531 interactions, as older biofilms are more representative of plaque collected from subgingival 532 sites, which themselves require days to develop familiar community profiles.

533
While community changes between generations due to inconsistent detection of low 534 abundance species may appear to make the system unreliable, it is similar to the recently here. This may be related to the production of short chain volatile fatty acids and volatile 582 sulfur compounds, which are attributed to proteolytic, disease-associated species [54].

583
Recent reports on the transcriptional activity of biofilm plaque from healthy and 584 periodontitis-affected sites suggest that community activity is more uniform across disease 585 sites than is microbial profile [6,7,55], and further investigation of protein expression and 586 metabolic output in healthy and disease-site derived biofilms is warranted [56].

587
In conclusion, we have demonstrated that our subgingival oral biofilm growth and 588 perpetuation model can maintain reliable communities and differentiated health/disease 589 microbial profiles for at least 80 days, and mimics dynamics of natural plaque. This method of biofilm perpetuation may be useful for growing biofilms to use in in vivo oral infection     Figure S7. Differential abundance of species present at >0.1% abundance between healthy 964 and disease site-derived biofilms. Species statistically different (q ≤ 0.05) and with an 965 effect size (DP) >1 by analysis in STAMP are considered differentially abundant.

966
Healthy plaque collection site 95% confidence intervals Diseased plaque collection site