Interbacterial Adhesion Networks within Early Oral Biofilms of Single Human Hosts

ABSTRACT

Specific interbacterial adhesion, termed coaggregation, is well established for three early colonizers of the plaque biofilm: streptococci, actinomyces, and veillonellae. However, little is known about interactions of other early colonizers and about the extent of interactions within the bacterial community from a single host. To address these gaps, subject-specific culture collections from two individuals were established using an intraoral biofilm retrieval device. Molecular taxonomy (Human Oral Microbe Identification Microarray [HOMIM]) analysis of biofilm samples confirmed the integrity and completeness of the collections. HOMIM analysis verified the isolation of Streptococcus gordonii and S. anginosus from only one subject, as well as isolation of a previously uncultivated streptococcal phylotype from the other subject. Strains representative of clonal diversity within each collection were further characterized. Greater than 70% of these streptococcal strains from each subject coaggregated with at least one other coisolate. One-third of the strains carry a known coaggregation mediator: receptor polysaccharide (RPS). Almost all nonstreptococcal isolates coaggregated with other coisolates. Importantly, certain Rothia strains demonstrated more coaggregations with their coisolated bacteria than did any Streptococcus or Actinomyces strain, and certain Haemophilus isolates participated in twice as many. Confocal microscopy of undisturbed biofilms showed that Rothia and Haemophilus each occur in small multispecies microcolonies. However, in confluent high-biomass regions, Rothia occurred in islands whereas Haemophilus was distributed throughout. Together, the data demonstrate that coaggregation networks within an individual's oral microflora are extensive and that Rothia and Haemophilus can be important initiators of cell-cell interactions in the early biofilm.

IMPORTANCE Extensive involvement of specific interbacterial adhesion in dental plaque biofilm formation has been postulated based on in vitro coaggregation between oral bacteria from culture collections that are not subject specific. In the present study, subject-specific culture collections were obtained from early plaque biofilm of two volunteers, and coaggregations within each culture collection were assayed. Coaggregations, several of which involved a coaggregation-mediating cell surface molecule known from well-studied streptococci, were widespread. Unexpectedly, the little-studied organisms Haemophilus and Rothia participated in the greatest numbers of interactions with community members; these two organisms showed different distributions within the undisturbed biofilm. The data show that coaggregation networks encompass most organisms within the biofilm community of each individual, and they indicate prominent participation of organisms such as Haemophilus and Rothia in early plaque biofilm formation.

INTRODUCTION

Molecular and cultivation studies have defined a paradigm oral bacterial community. In health, Streptococcus, Corynebacterium, Haemophilus, Actinomyces, Neisseria, Veillonella, and Rothia are prominent genera; streptococci are by far the largest community component (1–3). In a clone library of total oral microfloras pooled from 10 subjects, the most numerous genus was Streptococcus, followed by Haemophilus, Neisseria, Prevotella, Veillonella, and Rothia (4). However, at the level of the individual, Streptococcus dominated in only five subjects. A pyrosequencing study across three individuals (5) showed streptococci to dominate in each subject, but the proportions of other genera varied. Interestingly, 4.4% of reads unique to just one of the subjects belonged to the ubiquitous genus Streptococcus. Thus, overall community compositions differ between individuals, as do the suites of phylotypes within universally represented genera. Most studies report phylotype data only to the genus level (6); species-level analyses are needed, especially within the genus Streptococcus, the members of which are difficult to parse using high-throughput sequencing (HTS). In addition, the microbial populations from different habitats within the oral cavity (e.g., tongue, saliva, tooth surface) are often pooled rather than analyzed as populations of different biogeographies, an approach that masks differences in ecologically relevant parameters such as nutrient availability and oxygen tension.

When high-density pure-culture suspensions of oral bacterial isolates are mixed together pairwise, clumps visible to the eye often result (7, 8). These cell-cell adhesion events observed in vitro are termed coaggregations, and they underpin a model of oral biofilm maturation (9). However, among the initial colonizers of the tooth surface, only interactions between species of streptococci, actinomyces, and veillonellae have been studied in detail (8, 10–12). The relevance of in vitro coaggregation assays to interbacterial adhesion in vivo was demonstrated by immunofluorescence identification of complementary coaggregation-mediating adhesin/receptor molecules on cells intimately juxtaposed within an undisturbed biofilm retrieved from the oral cavity (13). Further, an immunofluorescence-targeted cell cluster micromanipulated from the biofilm yielded isolates of the predicted coaggregating organisms (14). The best-understood coaggregations involve structurally related high-molecular-weight cell surface glycans, collectively referred to as receptor polysaccharides (RPSs). Found on strains of S. oralis, S. sanguinis, and S. gordonii, the polymers are hexa- or heptasaccharide repeats that contain a host-like disaccharide motif, namely, either Galβ1-3GalNAc (the G motif) or GalNAcβ1-3Gal (the Gn motif). The motifs are recognized by protein adhesins on other oral bacteria; particular strains of S. gordonii and S. sanguinis bind the Gn motif, whereas certain Actinomyces strains recognize both motifs (15–17). Other features within the repeat give rise to five currently recognized RPS serotypes (18), but recognition of a particular RPS by a coaggregation partner depends solely on the recognition motif. Thus, putative RPS-bearing strains can be assigned a serotype (numbered 1 to 5) through immunoassay analysis, and the recognition motif (G or Gn) can be identified by coaggregation assays performed with characterized test strains of actinomyces and streptococci. Currently, seven RPS types are known: 1Gn, 2Gn, 2G, 3Gn, 3G, 4Gn, and 5Gn (19, 20). A different polysaccharide isolated from S. oralis strain H1 is known to mediate coaggregation with a strain of Capnocytophaga ochracae (21). Additionally, coaggregations can involve protein-protein recognition (22). Many RPS-mediated coaggregations are reversed upon addition of lactose (representing competitive inhibition of the lectin-like interaction), and adhesin-bearing cells can be identified by protease treatment, which abolishes coaggregation.

Little data exist on species-level community composition and on variation in coaggregation phenotypes within oral biofilms from single human hosts. The goals of the present study were to establish broad-based culture collections representative of the oral biofilm for two individuals, to use results of complementary molecular analysis of the biofilm as indicators of culture collection composition and completeness, and to assess the occurrence of RPS and the extent of interbacterial adhesion among the coisolated community members. Extensive coaggregation networks were identified, and RPS was found on one-third of the streptococci. Unexpectedly, Haemophilus and Rothia isolates demonstrated the greatest number of coaggregations, and these genera were found to be integrated differently within undisturbed biofilm. Together, the results show coaggregation to be ubiquitous within the human oral ecosystem of individuals, and they demonstrate the poorly studied genera Rothia and Haemophilus to be initiators of interspecies interactions with early oral biofilms.

RESULTS

HOMIM analysis of community composition.

Of 379 probes used in microarray analysis (see Data Set S1 in the supplemental material), only 39 had a score of ≥2 on any chip (Fig. 1; see Materials and Methods for a description of the scoring method and of the HOT [Human Oral Taxon] numbering system). Twenty-three probes (black text) that failed to reach a score of ≥2 on at least one chip in each visit were deemed inconsistently reactive. Six probes (red text) reached a score of ≥2 on every chip in both subjects: those representing Gemella spp., Rothia spp., and several Streptococcus species, including S. salivarius and S. australis. Ten probes reached a score of ≥2 for at least one chip in each visit. Of the 10, those for Haemophilus parainfluenzae, S. infantis, and S. parasanguinis were reactive in both subjects (blue text). Four of the 10 (including a probe for S. anginosus/gordonii) (green text) were reactive only in subject 1, whereas 3 of the 10 (including a probe for S. dentisani) (brown text) were reactive only in subject 2. Check marks in Fig. 1 indicate HOMIM probes for which at least one corresponding isolate was obtained.

FIG 1.

HOMIM analysis of biofilm DNA samples (heat map) and existence of isolates corresponding to particular HOMIM probes (check marks). Red text: probes with a score of ≥2 on every chip in both subjects. Blue text: probes with a score of ≥2 on at least one chip in every visit for both subjects. Green text: probes with a score of ≥2 on at least one chip in every visit for subject 1 only. Brown text: probes with a score of ≥2 on at least one chip in every visit for subject 2 only. The probe identifier (ID) presents the species name followed by the HOT number_probe number.

Subject-specific culture collections and coaggregation profiles: streptococci.

A total of 115 isolates were obtained from subject 1 and 129 from subject 2. After Human Oral Microbiome Database (HOMD) identification and repetitive extragenic palindromic-PCR (REP-PCR) fingerprinting (partitioning of strains within species), 48 representative strains (RSs) were obtained from subject 1 and 71 from subject 2. Within each subject, streptococci made up roughly 70% of all isolates and RSs (81 of 115 isolates and 34 of 48 RSs in subject 1, 87 of 129 isolates and 49 of 71 RSs in subject 2). Table 1 summarizes data on the streptococcal isolates. S. mitis made up roughly 23% of each subject's streptococcal isolates, and isolates of S. salivarius and S. vestibularis (including those designated S. salivarius/vestibularis; see Materials and Methods for the definition of this designation) made up roughly 30%. S. anginosus, S. gordonii, and S. tigurinus were isolated only from subject 1, as predicted by HOMIM analysis (Fig. 1), but HOMIM data neither supported nor contradicted the subject-specific recovery of most other streptococcal isolates. Two common plaque streptococci, S. mitis HOT 677 and S. sanguinis HOT 758, were isolated from both subjects; no microarray probes exist for these species (Fig. 1, bottom). Similarly, probes are also lacking for two unnamed streptococcal isolates collected from subject 2, one of which (HOT 431) was previously uncultivated. The sole clear inconsistency between the HOMIM results and streptococcal isolate recovery was the lack of S. australis isolates from subject 1. Of note, slightly more than one-third of streptococcal RSs in each subject (12 of 34 in subject 1, 18 of 49 in subject 2) were reactive with an anti-RPS antibody.

TABLE 1.

Characteristics of streptococcal isolates^a

HOMD ID	Subject 1						Subject 2
	No. of clones	RS ID no.	RPS type	No. of coaggregation partners			No. of clones	RS ID no.	RPS type	No. of coaggregation partners
				Streptococci	Others	Total				Streptococci	Others	Total
Streptococcus anginosus HOT 543	4	1		0	0	0	No isolates
Streptococcus australis HOT 073	No isolates						1	1		3	10	13
							1	2	3G(w)	1	11	12
							1	3	3G	3	12	15
							1	4	3G	2	14	16
							1	5	3	10	11	21
							1	6	3G	4	11	15
Streptococcus dentisani HOT 398	No isolates						1	20		0	1	1
							1	21		12	11	23
							3	22		0	0	0
							1	23	3(w)	0	0	0
Streptococcus gordonii HOT 622	1	2	1Gn	0	6	6	No isolates
Streptococcus infantis HOT 638	2	3		5	2	7	1	7		2	12	14
	2	4	4(w)	3	2	5	1	8		6	11	17
							1	9		5	11	16
							1	10	4	4	7	11
Streptococcus mitis HOT 677	1	5	3	0	0	0	7	11	4(w)	1	3	4
	3	6	3	3	3	6	2	12		0	2	2
	1	7	4	0	1	1	3	13	3Gn	1	11	12
	1	8		2	2	4	3	14		0	8	8
	3	9	1Gn	1	6	7	1	15		0	1	1
	4	10	1Gn	1	6	7	1	16		0	2	2
	1	11		0	0	0	1	17		0	6	6
	1	12	1Gn	1	6	7	1	18		0	0	0
	2	13		1	2	3	1	19	3Gn	1	8	9
	1	14		1	0	1
Streptococcus oralis HOT 707	No isolates						1	24		6	14	20
							1	25		7	15	22
Streptococcus parasanguinis II HOT 411	4	15		2	2	4	1	26		0	5	5
	2	16		0	1	1	1	27		0	1	1
	1	17		nt	nt	nt	1	28		2	4	6
	1	18		nt	nt	nt	1	29		10	3	13
	1	19		2	2	4
Streptococcus parasanguinis I HOT 721	No isolates						1	30		0	1	1
Streptococcus salivarius HOT 755	4	20		0	1	1	9	31		1	0	1
	2	21		0	0	0
Streptococcus salivarius/vestibularis	3	22	3	5	4	9	1	32	3	5	8	13
	2	23		0	2	2	5	33	3	8	8	16
	7	24		6	5	11	2	34	3	6	8	14
	1	25		0	2	2	2	35	3(w)	8	8	16
							1	36		1	0	1
							2	37		1	1	2
Streptococcus sanguinis HOT 758	2	26	3G	5	8	13	5	38		0	3	3
							2	39		0	2	2
							2	40	1Gn	0	7	7
							2	41		2	5	7
Streptococcus sp. strain HOT 061	No isolates						1	42	3(w)	2	10	12
							1	43	3(w)	2	9	11
Streptococcus sp. strain HOT 064	11	27	4Gn	5	6	11	No isolates
	2	28		0	2	2
Streptococcus sp. strain HOT 431 (previously uncultivated)	No isolates						4	44		7	2	9
							1	45	4	0	2	2
							1	46		0	1	1
							1	47		9	2	11
Streptococcus tigurinus HOT 071	1	29	H1	0	0	0	No isolates
	3	30		0	0	0
Streptococcus vestibularis HOT 021	2	31		1	2	3	1	48		nt	nt	nt
	1	32		0	0	0	1	49		nt	nt	nt
	2	33		nt	nt	nt
	2	34		0	0	0
Avg/median				1.3/1	2.4/2	3.8/3				2.8/1	5.3/5	8.6/9

^aThe number of clones in each REP-PCR pattern for the isolates of each HOT, the identifier (ID) number of the representative strain (RS), the type of RPS (if present), and the number of coaggregations for each RS with coisolated streptococcal and nonstreptococcal RSs are listed. For example, 4 isolates of S. infantis HOT 638 were obtained from subject 2 which displayed 2 distinct REP-PCR patterns, with 2 isolates falling into each pattern. One isolate from each pattern was designated a representative strain, and those isolates were given strain-representative strain numbers 3 and 4. Strain 4 reacted weakly (w) with anti-serotype-4 antibody, but a recognition motif could not be assigned. This strain coaggregated with 3 other coisolated streptococcal RSs and 2 nonstreptococcal RSs. nt, not tested.

Table 1 also summarizes the numbers of intrageneric (streptococcal partner) coaggregations, intergeneric (nonstreptococcal partner) coaggregations, and total coaggregations (sum of intra- and intergeneric coaggregations) for each streptococcal RS (a similar table that provides RS numbers for each coaggregation partner and tables showing numerical coaggregation scores are available in Data Set S1). Three of the 34 streptococcal RSs from subject 1 autoaggregated and were not tested. Of the remaining 31 RSs, 23 (74%) coaggregated with at least one coisolate (see “Total” columns in Table 1). The highest number of coaggregations seen in any streptococcal RS from subject 1 was 13 (S. sanguinis RS 26). In subject 2, after removal of the two autoaggregating RSs, 94% (44 of 47) of the isolates coaggregated with at least one coisolate, and several streptococcal RSs had more than 13 partners—S. dentisani RS 21 had 23 interactions. Importantly, almost all streptococcal RSs participated in more intergeneric than intrageneric coaggregations. Even those with no intrageneric partners often coaggregated with other genera. Thus, streptococci from both individuals overwhelming possess an innate potential to initiate mixed-species communities within single human hosts during initial colonization of the tooth surface.

Interestingly, adhesin-bearing S. infantis RS 3 from subject 1 recognized its sole G-RPS-bearing coisolate but only one of the five Gn-RPS-bearing coisolates (Data Set S1). When screened against seven structurally verified RPS-bearing test strains, it coaggregated with all four G-bearing strains but with none of the three Gn-bearing strains (data not shown). To the best of our knowledge, this is the first adhesin-bearing streptococcus characterized to bind the G recognition motif. More importantly, although the average numbers of intrageneric coaggregations were not greatly different between subjects, the mechanisms were clearly subject dependent. Protein-protein interactions accounted for 56% (38 of 68) of Streptococcus-Streptococcus coaggregations in subject 2 but for only 9% (2 of 22) in subject 1 (Data Set S1).

Coaggregations of RPS-bearing isolates.

Table 2 summarizes coaggregations of anti-RPS-reactive strains in which the partner bears an adhesin (i.e., interactions in which the RPS-bearing cell bears an adhesion for the partner are not included). Of the 12 anti-RPS-reactive RSs from subject 1, 7 could be assigned a recognition motif. Of the 18 antibody-reactive RSs from subject 2, again, 7 had a defined recognition motif. RSs with a defined recognition motif coaggregated with at least two coisolates; as expected, Actinomyces strains were frequent partners (Data Set S1). The sole exception was the anti-H1-reactive S. tigurinus RS from subject 1. H1 glycan seems to mediate interactions only with Capnocytophaga ochracae ATCC 33596 (24); Capnocytophaga spp. were neither isolated nor detected by HOMIM analysis. For strains that lacked a defined recognition motif (i.e., that were antibody reactive but failed to coaggregate with characterized Actinomyces and Streptococcus test strains), 3 of the 5 from subject 1 had a coisolated partner, as did 5 of the 11 from subject 2. Interestingly, the 4 antibody-reactive S. salivarius/vestibularis RSs from subject 2 bore serotype-3-reactive polysaccharides with undefined recognition motifs; each of these RSs displayed a high number of intrageneric interactions (i.e., with their coisolated streptococci) but showed no RPS-mediated intergeneric coaggregations. Likewise, the antibody-reactive S. salivarius/vestibularis RS isolated from subject 1 bore a serotype 3 polysaccharide and had a high number of intrageneric coaggregations. Overall, RPS and antigenically similar cell wall glycans are well represented within each subject's community, and they are recognized by protein adhesins borne on their coisolated community members.

TABLE 2.

RPS-mediated coaggregations of streptococci^a

Subject	HOMD ID	Defined recognition motifb				Undefined recognition motifc
		RS ID no.	RPS type	No. of coaggregations		RS ID no.	RPS type	No. of coaggregations
				Intrageneric	Intergeneric			Intrageneric	Intergeneric
1	Streptococcus gordonii HOT 622	2	1G	0	6		None
	Streptococcus infantis HOT 638		None			4	4(w)	0	0
	Streptococcus mitis HOT 677	9	1Gn	1	4	5	3	0	0
		10	1Gn	1	4	6	3	0	1
		12	1Gn	1	4	7	4	0	1
	Streptococcus salivarius/vestibularis		None			22	3	5	2
	Streptococcus sanguinis HOT 758	26	3G	1	5		None
	Streptococcus sp. strain HOT 064	27	4Gn	3	5		None
	Streptococcus tigurinus HOT 071	29	H1	0	0		None
2	Streptococcus australis HOT 073	2	3G(w)	1	2	5	3(w)	0	0
		3	3G	1	2
		4	3G	0	2
		6	3G	1	2
	Streptococcus dentisani HOT 398		None			23	3(w)	0	0
	Streptococcus infantis HOT 638		None			10	4	0	0
	Streptococcus mitis HOT 677	13	3Gn	1	2	11	4(w)	0	1
		19	3Gn	0	2
	Streptococcus salivarius/vestibularis		None			32	3	5	0
						33	3	6	0
						34	3	4	0
						35	3(w)	5	0
	Streptococcus sanguinis HOT 758	40	1Gn	0	7
	Streptococcus sp. strain HOT 061		None			42	3(w)	0	0
						43	3(w)	0	0
	Streptococcus sp. strain HOT 431		None			45	4	0	0

^aEach anti-RPS-reactive representative strain (RS; w, weak reactivity) was characterized for a recognition motif using coaggregation test strains of A. naeslundii and S. gordonii and then screened for coaggregation with streptococcal and nonstreptococcal coisolates. Only RPS-dependent coaggregations (the partner bears the protein adhesin) are listed.

^bAntibody-reactive strains that coaggregate with defined test strains and can therefore be assigned a recognition motif. In the RPS designations, the number indicates the serotype and G or Gn indicates the receptor motif.

^cAntibody-reactive strains that do not coaggregate with test strains and therefore cannot be assigned a recognition motif. In the RPS designations, the number indicates the serotype.

Subject-specific culture collections and coaggregation profiles: nonstreptococci.

Nonstreptococcal isolates from each subject are listed in Table 3. Consistent with HOMIM probe reactivity, isolates of Haemophilus parainfluenzae, Gemella sanguinis, Gemella haemolysans, Rothia mucilaginosa, and Rothia dentocariosa were obtained. In contrast, Actinomyces isolates were obtained despite the lack of corresponding HOMIM probe reactivity (Fig. 1, bottom). Additionally, an isolate of Neisseria flava was obtained from subject 1 but no reactivity was seen with a potentially reactive probe (Neisseria cluster). Neither Pseudomonas nor Fusobacterium spp. were isolated from subject 2 despite consistent probe reactivity; a selective medium for Fusobacterium yielded no isolates, and no medium appropriate for Pseudomonas was employed. Aside from that for Fusobacterium, the only notably reactive HOMIM probes indicative of anaerobes was that for Veillonella, an organism not specifically targeted with selective medium in this study. Thus, the lack of obligate anaerobe isolates would be predicted from the HOMIM data.

TABLE 3.

Characteristics of nonstreptococcal isolates^a

HOMD ID	Subject 1					Subject 2
	No. of clones	RS ID no.	No. of coaggregation partners			No. of clones	RS ID no.	No. of coaggregation partners
			Streptococci	Others	Total			Streptococci	Others	Total
Actinomyces odontolyticus HOT 701	No isolates					1	50	10	5	15
Actinomyces sp. strain HOT 169	1	35	6	0	6	No isolates
	1	36	6	0	6
	1	37	6	0	6
Actinomyces sp. strain HOT 170	No isolates					2	51	9	4	13
Actinomyces sp. strain HOT 175	No isolates					1	52	9	3	12
Gemella haemoysans HOT 626	2	38	0	0	0	No isolates
	3	39	0	0	0
	5	40	0	0	0
	1	41	0	0	0
Gemella sanguinis HOT 757	No isolates					2	53	0	1	1
Haemophilus parainfluenzae HOT 718	3	42	17	2	19	3	54	20	1	21
	1	43	0	0	0	2	55	3	1	4
	1	44	3	1	4	3	56	24	5	29
	6	45	11	1	12	4	57	21	7	28
						1	58	22	7	29
						1	59	nt	nt	nt
						1	60	8	2	10
Neisseria flava HOT 609	2	46	6	0	6	2	61	7	4	11
						1	62	20	4	24
						1	63	12	0	12
Neisseria flavescens HOT 610	No isolates					6	64	19	4	23
Rothia dentocariosa HOT 587	No isolates					2	69	4	4	8
						1	70	13	6	19
						2	71	5	4	9
Rothia mucilaginosa HOT 681	3	47	8	3	11	3	65	23	4	27
	4	48	10	1	11	1	66	13	4	17
						1	67	18	3	21
						1	68	23	6	29
Avg/median			5.2/6	0.6/0	5.8/6			13.5/13	3.8/4	17.2/15

^aColumn headings are as defined in footnote a of Table 1. nt, not tested.

Subject-specific coaggregations of nonstreptococcal RSs are also summarized in Table 3. While the average and median numbers of total coaggregations for all nonstreptococci (column “total”) are higher than those for all streptococci (Table 1, “Total” columns), the results of this comparison are statistically significant only for subject 2 (U test; P = 0.0019) and are related to the high numbers of coaggregations displayed by Neisseria RSs; this genus was poorly represented in subject 1. Similarly, the lower average and median numbers of total coaggregations seen in subject 1 are associated with higher representation of Gemella isolates, organisms that have almost no coaggregation partners. However, in both subjects, strains of Haemophilus, Neisseria, and Rothia were promiscuous coaggregators, and the number of interactions for several of these strains exceeded that for any coisolated Streptococcus and Actinomyces strains by 30 to 50% (cf. Tables 1 and 3). In all cases, the Haemophilus strains carry a protein adhesion (Data Set S1). In subject 2, their coaggregation mechanism was protein-protein (both partners were protease sensitive). However, in subject 1, the Haemophilus coaggregations were in most cases confirmed as protein-carbohydrate in nature (lactose reversible, and only the Haemophilus strains were protease sensitive). In both subjects, coaggregations of R. mucilaginosa, R. dentocariosa, and N. flava involved a protease-sensitive adhesin on partner streptococci (Data Set S1); thus, the Rothia and N. flava strains likely bear an as-yet-to-be-described cell surface carbohydrate receptor. In summary, subject-dependent differences exist in the mechanisms of Haemophilus-Streptococcus coaggregations, and three genera that are common in the early biofilm (Rothia, Haemophilus, and Neisseria) undergo myriad coaggregations.

Characterization of immunofluorescence antibodies for identification of organisms within the in situ community.

The discovery that Haemophilus, Rothia, and Neisseria isolates display extensive coaggregation interactions and that Gemella isolates had almost no interactions spurred attempts to identify these organisms within the biofilm in situ. Antibodies for identification of these organisms were prepared against a selected RS, after which all RSs of that genus were screened for antibody reactivity. Both R. mucilaginosa RSs from subject 1 were reactive with the “anti-Rm” antibody produced against R. mucilaginosa RS 47 from subject 1, as were three of the four R. mucilaginosa RSs from subject 2. However, no R. dentocariosa RSs were reactive. All Neisseria RSs from both subjects were reactive with the “anti-Nf” antibody prepared against N. flava RS 63 from subject 2. All four Haemophilus RSs from subject 1 reacted with the “anti-Hp” antibody prepared against H. parainfluenzae RS 42 from that subject, but only 3 of the 7 RSs from subject 2 were reactive. The “anti-Gem” antibody prepared against the G. sanguinis RS from subject 2 also reacted with one of the four G. haemolysans RSs from subject 1. Together, these results suggest that immunofluorescence identification of R. mucilaginosa and Neisseria sp. cells within the biofilms of both subjects might be efficient. In contrast, H. parainfluenzae biomass might not label efficiently in subject 2, and Gemella sp. biomass might be poorly reactive in subject 1.

Localization of Haemophilus, Rothia, and Neisseria within undisturbed biofilm.

Immunofluorescence microscopy of 4-h biofilms from subject 1 revealed nascent communities consisting of a few anti-Hp-reactive cells in direct contact with small numbers of other cells, including those reactive with anti-RPS4; in 8-h biofilms, confluent regions of such composition were seen (Fig. 2a). One of the three RPS4-bearing RSs from subject 1 (Streptococcus sp. HOT 064, strain 27) coaggregated with Haemophilus immunogen strain RS 1 (Data Set S1). Similar topologies were seen with anti-Hp-stained cells and anti-RPS3-stained cells (see Fig. S2a in the supplemental material); of the four RPS3-bearing RSs, three (S. mitis RS 6, S. salivarius RS 22, and S. sanguinis RS 26) coaggregated with Haemophilus strains (Data Set S1). In subject 2, seven 4-h chips and four 8-h chips were stained with anti-Hp, but antibody-reactive cells were seen only on a single 4-h chip (Fig. S2b). While anti-Hp was reactive with all Haemophilus RSs from subject 1, only two of the four Haemophilus RSs from subject 2 that coaggregated widely with coisolated streptococci were reactive; i.e., only two of the four could have been identified in situ.

FIG 2.

(a) Immunofluorescence localization of Haemophilus and associated cells in undisturbed biofilms from subject 1. Top panels, 4-h biofilm. Bottom panels, 8-h biofilm. Right panels, zoom of central region in left panels. Scale bar = 10 μm. (b) Immunofluorescence localization of Rothia and associated cells in undisturbed biofilms from subject 1. All images represent the same field of view. Grayscale images show cells stained with DAPI, anti-Rm antibodies, or anti-RPS4 antibodies. The lower right panel shows a red-green-blue (RGB) overlay of grayscale images. Circles mark cells in intimate interaction. (c) Rothia cells had reduced antibody reactivity in 8-h biofilms. All images represent the same field of view. Grayscale images show cells stained with DAPI, anti-Rm antibodies, or anti-RPS4 antibodies, plus an image with the anti-Rm signal enhanced. RGB overlays include an image showing the enhanced anti-Rm channel. Scale bar = 10 μm.

In 4-h biofilms from subject 1, cells reactive with anti-Rm were seen in tight association with other cells and in clusters containing as few as three cells, including those reactive with anti-RPS4 (Fig. 2b). Six streptococcal RSs from subject 1, one of which (S. infantis RS 4) was anti-RPS4 reactive, coaggregated strongly with R. mucilaginosa strains; a weak coaggregation was seen with the anti-RPS4-reactive strain S. mitis RS 7. In subject 2, anti-Rm-reactive cells were likewise seen in small clusters containing cells reactive with anti-RPS3 (Fig. S2c); 9 of the 18 coisolated streptococcal RSs that coaggregated strongly with R. mucilaginosa RSs were anti-RPS3 reactive. Thus, in both subjects, R. mucilaginosa-Streptococcus coaggregations can explain the topology of nascent biofilms. In 8-h biofilms, the fluorescence intensity of anti-Rm-reactive cells in the midst of confluent biofilm was typically low. Figure 2c shows anti-Rm-reactive cells at the edges of the biofilm, and it also shows inner regions of the biofilm that are poorly stained with DAPI (4′,6-diamidino-2-phenylindole). However, these inner regions contained weakly anti-Rm-reactive cells that could be revealed through postacquisition enhancement of the anti-Rm signal. Thus, the condition of Rothia cells within high-biomass regions appears to be different from that of isolated cells and of cells associated with small multispecies aggregates.

Fluorescence in situ hybridization (FISH) (Fig. 3) complemented and extended the immunofluorescence data. The overwhelming majority of the biomass consisted of cells labeled with the Streptococcus STR probe, and cells reactive solely with the Eubacteria (EUB) probe (Fig. 3b, panel b₂) were rare. Cells labeled with the Pasteurellaceae (PAS) probe (reactive across the Pasteurellaceae, including Haemophilus) were numerous and were distributed throughout the confluent biofilm but were also seen in small multispecies clusters consisting of a few cells. In particular, FISH unambiguously documented distinctive streptococcus-enveloped islands of Rothia similar in topology to those revealed by enhancement of immunofluorescence signal (cf. Fig. 2c). Regions labeled with Neisseria FISH probe NEI (Fig. 3a, panel a₂) had a topology similar to that seen with Rothia and were likewise seen in antibody-labeled biofilms (Fig. S3a and b). Cells labeled with the Gemella probe GEM were distributed throughout the biofilm (Fig. 3b, panels b₁ and b₂). In contrast, gemellae localized by immunofluorescence (anti-Gem reactive) appeared to be less numerous than those labeled by FISH, and immunofluorescently labeled gemellae appeared to have less contact with other biofilm cells than did those labeled by FISH (Fig. S3c and d).

FIG 3.

FISH localization of genera in undisturbed 8-h biofilms of subject 1. Asterisks in low-magnification images (upper panels) mark regions shown in high-magnification fields of view (bottom panels). The vast majority of cells were STR-stained streptococci. PAS-stained cells (Haemophilus) were distributed throughout the biofilm (upper panels), as well as in small multispecies clusters (a₂). Rothia (ROT probe) and Neisseria (NEI probe) occurred in distinctive packets. GEM-labeled cells (Gemella) were distributed throughout the biofilm (b, b₁, and b₂) and as single cells (upper right of panel b). Few cells labeled solely with EUB were seen (b₂).

DISCUSSION

Previous coaggregation surveys employed bacterial strains obtained from disparate culture collections. Sometimes these organisms were chosen precisely for their high coaggregation specificity (few partners). Results of previous coaggregation studies have been extrapolated to encompass oral bacterial communities of all human hosts despite recognition of subject-specific differences in community composition and the strain-specific nature of many coaggregations. One important discovery in earlier studies was that organisms bearing particular complementary coaggregation mediators were intimately juxtaposed in vivo (13, 14), and yet those data likewise fail to address the overall potential for interbacterial adhesion within the community of a single host. The present study employed broad-based subject-specific culture collections as the starting point for pairwise coaggregation assays of all isolates (roughly 500 assays, including lactose reversibility and protease sensitivity assays), an approach impractical for study of many subjects. However, even for this set of 2 subjects, investigation of subject-specific clinical isolates yielded important ecologically relevant information, not only for isolates of little-studied organisms but also for genera (e.g., streptococci) central to the earlier studies. In light of the present data, it can be stated that, within the oral microflora of single individuals, coaggregations are ubiquitous and encompass numerous representatives of nearly every genus investigated. Organisms that comprise the early plaque biofilm of single human hosts are predisposed to form diverse assemblages through cell-cell adhesion.

The present study was the first to comprehensively examine the subject-specific occurrence of RPS and RPS-mediated coaggregation. Together, five of the seven described RPS types were found in these two individuals, and, within each subject, multiple coaggregation partners for each coisolated RPS type were isolated. A streptococcus species bearing the unique H1 polysaccharide occurred in one host, but its very specific coaggregation partners (Capnocytophaga strains that bear a particular adhesin) (24) were not isolated; this polysaccharide may be of less ecological consequence than is RPS. Furthermore, RPS was found on streptococcal species not previously known to bear the molecule (S. infantis, S. australis) as well as on two unnamed streptococcal species, one of which was first isolated in the course of the present study. More importantly, several streptococcal isolates bore anti-RPS-reactive cell surface components that have no classical recognition motif and that appear to mediate primarily intrageneric coaggregations. In particular, S. salivarius/vestibularis strain 22 from subject 1 bears a polysaccharide that, according to nuclear magnetic resonance (NMR) spectroscopy, is structurally different from any currently characterized RPS (C. A. Bush, personal communication). Taken together, the results indicate not only that RPSs and RPS-mediated coaggregations are important components of oral communities at the level of the individual host but also that the diversity of these coaggregation-mediating polysaccharides is higher than previously thought.

Ribosomal gene sequence microarray (HOMIM) analysis showed the oral microfloras of the subjects to be similar: streptococci dominated, with substantial representation of Haemophilus and Rothia. These data agree with those of high-throughput sequencing studies (3, 5); however, species-level HOMIM data were important in assessing the completeness of the streptococcal collection. Likewise, isolation of S. sanguinis and S. mitis was important because no HOMIM probe exists for these common species. The consistent reactivity of the Pseudomonas cluster probe (O96) in subject 2 is puzzling. Probes specific for the common PCR contaminant P. aeruginosa were unreactive (see Data Set S1 in the supplemental material). The remaining Pseudomonas species ascribed to the cluster probe occur at miniscule abundance in the HOMD clone library, i.e., are minor components of oral communities. One clear discrepancy between HOMIM data and culture collection composition is the isolation of S. australis from only one subject despite consistent detection by HOMIM analysis in both subjects. S. australis is closely related to S. infantis (25, 26). The 500-base ribosomal gene sequences of the present S. australis isolates had as few as 2 bases of difference from those of the HOMD infantis reference sequences and vice versa. However, isolates of these species participate in high numbers of mechanistically distinct coaggregations: the S. australis isolates bear RPS, whereas the S. infantis strains bear adhesins. Furthermore, two S. infantis isolates bind the G recognition motif, a specificity that was first observed in the present study. The prominence of these species and their propensity for coaggregation suggest an important role for these organisms in oral biofilm initiation.

Rothia, Haemophilus, and Neisseria are common in early plaque but have been poorly represented in previous coaggregation studies. Isolates of these genera from the present study participated in a large number of coaggregations, in some cases more than any single streptococcus or actinomyces species. This observation provides a context for an interesting result from a recent study in which Rothia strains were directly isolated by cell-cell adhesion from dispersed plaque of all three participants (27). Based solely on their extensive coaggregations, a greater role in models of oral biofilm development is warranted for Rothia, Haemophilus, and Neisseria. Should these organisms be shown to coaggregate with those associated with periodontal disease, they could play a role similar to that of fusobacteria in coaggregation-based models of oral biofilm maturation (9). In the present study, complementary use of immunofluorescence and spectral FISH demonstrated that Rothia and Neisseria appear as islands within the confluent biofilm. The type strain of R. dentocariosa bears an antigenic cell wall fructan (28), and some Rothia strains from the present study have a prominent capsule (Fig. S3d). Fructans of cariogenic S. mutans have ecological relevance as extracellular carbon storage products (29, 30), and the commensal streptococcal species S. gordonii and S. sanguinis are predicted by the CAZy database (www.cazy.org) (31) to have cell surface fructosidases. Other commensals, including S. salivarius, produce soluble fructan-hydrolyzing enzymes and grow on fructan (32, 33). Accordingly, it is possible that the streptococci intimately associated with Rothia cells degrade that organism's capsular components as an in situ carbon source, a process that could explain the observed reduction in anti-Rm binding in situ (Fig. 2). Thus, while Rothia cells are a nucleation point for cell-cell interactions, their role in high-biomass regions of the biofilm may be limited. In contrast, anti-Haemophilus-reactive cells, as well as those reactive with the PAS FISH probe, were seen not only in intimate contact with other cells in nascent multispecies colonies but also dispersed throughout high-biomass regions (Fig. 2 and 3). In a FISH study of older (h 24 to 48) plaque scraped from the teeth and gingival margin of healthy subjects, PAS-reactive cells were likewise prominent (34). The cells were integrated into the distal regions of large structures (“hedgehogs”) composed to a great extent of corynebacteria. Furthermore, PAS-reactive cells, together with Lautropia and streptococci, were a major component of a unique “cauliflower” structure. The present in situ model generates an analogue of the thin biofilms typical for early smooth-surface plaque. Neither the “hedgehog” or “cauliflower” structures, nor the organisms central to those structures (corynebacteria and Lautropia), have been found in the chip model. The extent to which haemophili are incorporated into nascent as well as highly developed plaque suggests these organisms to be important in biofilm maturation.

MATERIALS AND METHODS

Subjects and biofilm model.

Orally and systemically healthy volunteers were recruited through NIDCR Clinical Protocol 13-D-0014. Human enamel chips (35) were affixed to a full-mandibular stent using dental wax (see Fig. S1 in the supplemental material). Chips were carried intraorally for 4 h and 8 h and were then stained with DAPI and fluorescently labeled antibodies against RPS serotypes 1, 2, 3, and H1 (13). Fourteen subjects participated in an initial screening visit. Two subjects (males, ages 45 and 30; referred to here as subject 1 and subject 2) were selected for further study based on an increase in biofilm biomass between h 4 and h 8 and on the number of antibody-reactivity cells relative to those in other subjects.

HOMIM analyses.

On three separate visits, a chip was removed after 4 h and after 8 h, dipped three times in chilled phosphate-buffered saline (PBS), placed in 54 μl of chilled 10× lysis buffer (200 mM Tris HCl [pH 8.0], 20 mM EDTA [pH 8.0], 12% [vol/vol] Triton X-100), and sonicated (Bransonic 1510; Emerson Industrial, North Olmsted, OH) for 15 min. The chip was removed, and the DNA was extracted from the sonicate using a DNA-Easy blood and tissue kit (Qiagen, Valencia, CA) following the procedure for Gram-positive (Gram⁺) organisms and then sent to the Forsyth Institute (Cambridge, MA) for HOMIM analysis (36). The analysis is based on reverse capture of fluorescently labeled 16S rRNA gene amplicons onto a microarray of 379 probes correlated to taxa in the highly curated Human Oral Microbiome Database (HOMD), and it yields heat maps of amplicon abundances with values 1 to 5, representing a nonselective semiquantitative inventory of the major oral taxa. Each taxon is defined by Linnaean nomenclature as well as by a human oral taxon (HOT) number important for delineation of unnamed or uncultivated organisms. In the present report, all HOTs with names that include “sp.” are cultivated but unnamed organisms. For example, Streptococcus sp. strain HOT 431, previously designated “uncultivated phylotype” within the HOMD, was cultivated during the present study and is now designated “cultivated unnamed.” Details of HOTs and the HOMD are available (37) (http://www.homd.org).

Subject-specific culture collections.

Two rounds of isolation were performed. An 8-h chip was dipped three times in reduced transport fluid (RTF), placed in 250 μl chilled RTF, and sonicated as described above. The sonicate was plated at dilutions from 10² to 10⁴ onto aerobically (5% CO₂) and anaerobically (N₂/H₂/CO₂; 90/5/5) incubated TSA blood agar (tryptic soy agar with 5% sheep blood; Remel, Lenexa, KS), Columbia blood agar (Remel), and laked blood agar (Anaerobe Systems, Morgan Hill, CA). Campylobacter agar (Anaerobe Systems) and Fusobacterium agar (Anaerobe Systems) were used but yielded no isolates. Cadmium sulfate fluoride acridine Trypticase (CFAT) agar (Anaerobe Systems; used for Actinomyces spp.) yielded only a single Actinomyces isolate from one individual. Plates were examined after 24 and 48 h. All colonies of unique morphology were picked, and multiple picks were made of colonies with similar morphologies. In additional rounds of isolation, the following nonstreptococcal genera prominent in the HOMIM analyses but not obtained with the previously listed media were targeted: Haemophilus spp. by using chocolate agar (Remel) and chocolate agar plus bacitracin (Hardy Diagnostics, Santa Maria, CA) and Gemella spp. by using Gemella agar plus colisitin (38). Colonies were transferred to brain heart infusion (BHI; Oxoid, Hampshire, United Kingdom) broth, BHI broth supplemented with 10 μg/ml NAD and 5 μg/ml hemin (isolates from chocolate agar), or BHI broth with 10% horse serum (isolates from Gemella agar). Transfers with visible growth were examined microscopically and streaked onto agar, and a single colony was regrown in broth. In rare cases, more than one colony type was present in the streak, and each was picked for regrowth. Cultures were concentrated 2-fold into broth containing 20% glycerol and frozen at −20°C.

A genomic DNA template was obtained for species identification either by freeze/thaw of cells or, when the yield from the freeze/thaw was low, by using a DNA-Easy kit (Qiagen). Amplification of the 16S rRNA gene from genomic template was performed using previously published primers and PCR programs (37). The presence of a 1,500-base PCR product was confirmed on an agarose gel, and then the product was purified from the reaction mixture with ExoSAP-IT (Affymetrix, Cleveland, OH). When multiple bands were present on the gel, the 1,500-base band was excised from a preparatory gel and the product released using a QIAEX II gel purification kit (Qiagen). Sequencing of the initial 500 bases was performed at the NIDCR Combined Technical Research Core, and sequences were identified by BLAST on the HOMD website.

An isolate was assigned to a HOT when the level of identity to a HOMD reference sequence (RefSeq v13.2) was ≥98.5%; most were ≥99%. Identity to other HOT organisms was typically <98%. Certain isolates (including some streptococcus and Neisseria isolates) had >98.5% identity to more than one reference sequence. In such cases, the HOT of highest identity was used; these identifications were often supported by HOMIM data and phenotype. Levels of identity to S. salivarius and to S. vestibularis were sometimes equal; i.e., no best match existed—these isolates were designated S. salivarius/vestibularis. After taxonomic assignment, strain-level relatedness was assessed using REP-PCR fingerprinting with primers REP2-D and Rep1R-D (39). A randomly selected isolate within each REP-PCR fingerprint group was designated a representative strain (RS) for further analysis.

Coaggregation assay.

Overnight cultures were washed twice with coaggregation buffer (40) and resuspended to 260 Klett units (Klett-Summerson turbidometer), and then 100 μl of each suspension was added pairwise to a 10-mm-by-75-mm glass tube and subjected to vortex mixing. Coaggregations were scored as +++ (large clumps, clearing of the suspension), ++ (aggregates easily visible, limited clearing of the suspension), + (small clumps), or − (no interaction). Some pure-culture suspensions were slighty clumpy. These were tested after the suspension had been subjected to vortex mixing and allowed to settle for 1 to 2 min; only rapidly apparent coaggregations with scores of ++ and +++ were recorded for these strains. A few cultures autoaggregated and were not tested. Reversibility of coaggregation by addition of lactose (60 mM final concentration) served as a simple test for certain carbohydrate receptors. The sensitivity of a coaggregation to protease (which reveals protein adhesins) was assessed by adding Streptomyces griseus protease (Sigma; 2 mg/ml final concentration) to a pure suspension of each cell type of the coaggregating pair, heating at 50°C for 1 h, and then mixing each treated cell type with its untreated partner. If coaggregates did not form in one of the two mixtures, the treated cell type was scored as having a protein adhesion. If coaggregates did not form in either combination of treated cells with untreated cells, the mechanism was defined as a protein-protein mechanism. In the results, coaggregation data are reported only by the criterion “yes/no”; coaggregation scores are provided in Data Set S1.

Antibody reactivity.

Affinity-purified anti-RPS antibodies have been described previously (13). Other antibodies were produced for this study by washing whole cells from an overnight culture twice with sterile PBS, concentrating them 2-fold into PBS, adding sodium azide (final concentration of 0.02%), and storing the suspension overnight at 4°C. The suspension was centrifuged, resuspended in sterile PBS, and shipped on ice for preparation of rabbit antisera (Covance Inc., Denver, PA). IgG was purified (Nab Protein A Plus spin columns and Gentle buffers; Thermo/Pierce Scientific, Rockford, IL) and labeled with Alexa Fluor kits (ThermoFisher, Grand Island, NY). To assess reactivity, PBS-washed cells were subjected to reactions with 5 μg/ml of labeled primary antibody and examined using a 100× 1.3 numerical aperture (NA) Fluorotar oil immersion lens on a Leica DM epifluorescence microscope with a 100-W mercury lamp. A single examiner assessed fluorescence. Dim fluorescence was scored “w” (weak).

Immunofluorescence microscopy.

Chips were washed with PBS, stained with 5 μg/ml Alexa Fluor-labeled antibody plus 1 μg/ml DAPI for 15 min, and then washed again (13). Stained chips were mounted on dental wax in a petri dish filled with PBS and then examined with a Zeiss 710 laser confocal microscope using a 63× (0.9 NA) water immersion lens (Zeiss, Thornwood NY). Histograms were adjusted using the min/max function in Zen software (Zeiss).

Fluorescence in situ hybridization (FISH).

Protocols were adapted from Pernthaler et al. (41). Chips were fixed overnight in 4% paraformaldehyde at 4°C, washed 3 times in water, dehydrated in an ethanol series, placed in Eppendorf tubes containing 100 μl of hybridization buffer (0.9 M NaCl, 0.02 M Tris [pH 7.4], 0.01% SDS, 20% “Hi-Di”-grade formamide) containing FISH probes (Table 4) (each at 2 μM), and then incubated at 46°C for 18 h. Chips were again washed in hybridization buffer and then in hybridization buffer lacking formamide (both washes for 15 min at 48°C). Chips were again dehydrated, mounted on coverslips in ProLong Gold antifade mounting medium (Life Technologies), and allowed to cure for 24 to 72 h. Spectral images were acquired using a Zeiss 710 laser confocal microscope with a 32-channel multianode spectral detector, a 63×/1.4 NA objective, and laser lines at 488, 561, and 633 nm. Linear unmixing was performed with Zen software using reference spectra acquired on cultured cells using the same fluors and the same acquisition settings as were used for imaging of the chips.

TABLE 4.

FISH probes used in this study

Probe name	Target organism(s)	Probe sequence	Fluorophore	Reference or source
EUB 338	Most bacteria	GCTGCCTCCCGTAGGAGT	Rhodamine Red-X	42
STR 405	Streptococcus	TAGCCGTCCCTTTCTGGT	ATTO 590	43
PAS 111	Pasteurellaceae	TCCCAAGCATTACTCACC	ATTO 630	44
ROT 491	Rothia	TAGCCGGCGCTTTCTCTG	ATTO 532	44
NEI 1030	Neisseriaceae	CCTGTGTTACGGCTCCCG	ATTO 550	44
GEM 844	Gemella	GCTGCAGCACTGATCTCT	Alexa Fluor 488	This study
GEM 992	Gemella	GTGTCCTCACAGTATGTC	Alexa Fluor 488	This study

Accession number(s).

Complete 16S sequences from four isolates of the previously uncultivated Streptococcus sp. strain HOT 431 were determined by the Forsyth Institute and are available in GenBank (accession numbers KU351674 to KU351677).