Bacterial counts were performed on each maple sap sample. Total aerobic counts were obtained by diluting the samples in 0.1% peptone water (Difco Laboratories, Detroit, Mich.) and spread plating in duplicate on plate count agar (Difco Laboratories) supplemented with 0.5% of sucrose. Incubation was performed at 30°C for 48 h followed by 7°C for 10 days to increase the differences in colony appearance. Bacterial isolates were obtained from plate count agar plates presenting between 30 and 150 colonies (from one to eight colonies per sample). The isolates were picked to represent the widest variation of colony appearance (form, color, and texture) and then purified by three successive cultivations on tryptic soy agar (Difco) at 30°C for 24 to 48 h. Stock cultures were prepared in brain heart infusion medium (Difco) supplemented with glycerol and kept frozen at −80°C prior to use.

The DNA of bacterial isolates was prepared according to the procedures of Vincent et al. (27) with the exception that for gram-negative bacteria, no mutanolysin was used. The concentration of purified DNA was measured using Hoechst 33258 dye and a fluorometer (Hoefer DyNA Quant 200; Pharmacia Biotech, Piscataway, N.J.).

The amplified 16S rRNA gene was obtained from each isolate and reference strain by PCR with the universal primers F27 (5′-AGAGTTTGATCMTGGCTCAG-3′) (3) and R1492 (5′-TACGGYTACCTTGTTACGACTT-3′) (28), which are targeted to universally conserved regions and permit the amplification of an approximately 1,500-bp fragment. PCR amplification was carried out in a GeneAmp PCR System 9600 thermocycler (Perkin-Elmer Corporation, Norwalk, Conn.). Reaction tubes contained 25 ng (5 μl) of DNA extract, 1 U of Taq DNA polymerase (Pharmacia Biotech), 1× buffer (10 mM Tris-HCl [pH 9.0], 1.5 mM MgCl₂, 500 mM KCl), 10 mM deoxynucleoside triphosphate, and 20 pmol of each primer/μl. Initial DNA denaturation and enzyme activation steps were performed at 95°C for 4 min, followed by 35 cycles of denaturation at 94°C for 30 s, annealing at 56°C for 1 min and extension at 72°C for 1 min, and a final extension at 72°C for 10 min. The presence and yield of specific PCR product (16S rRNA gene) was monitored by 1% agarose (wt/vol) (BioShop Canada) gel electrophoresis at 200 V for 1 h in 1× Tris-acetate-EDTA buffer and made visible by ethidium bromide staining and UV transillumination.

Three restriction enzymes were selected on the basis of simulated digests of the complete 16S rRNA gene sequences of chosen reference strains retrieved from GenBank. These strains are listed along with the accession numbers in Table ​Table1.1. The selection of the reference strains was based on a preliminary study performed in our laboratory on bacterial population of maple sap from which different isolates were identified (data not shown) using the API 20E and 20NE strips (Biomerieux, Marcy-l'Etoile, France). The enzymatic reactions were simulated via the NEB cutter software, version 1.0, of the New England BioLabs website, using all commercially available enzymes. For digest experiments, cost-effective combinations of three enzymes were chosen, revealing clearly different (and thus by agarose gel electrophoresis clearly resolvable) restriction site polymorphisms in the reference strain sequences. Unpurified PCR products were each digested with the three selected restriction enzymes in separated reactions. The chosen enzymes were RsaI, HaeIII, and AluI. The digestions were performed for 2 h at 37°C in 20-μl reaction volumes containing 10 μl of the PCR product solution, 2 μl of the commercially supplied incubation buffer, 7.5 μl of water, and 0.5 μl (10 U/μl) of the restriction enzyme. Reaction products (10 μl) were run on a 2% agarose 1000 gel (GibcoBRL, Life Technologies Inc., Burlington, Ontario, Canada) in 1× Tris-borate-EDTA buffer for 4 h at 250 V under refrigeration. Gels were stained with ethidium bromide, made visible by UV transillumination, and digitalized with the gel print 2000i system (Bio/Can Scientific Inc., Mississauga, Ontario, Canada). The images were analyzed with the GelCompar software (Molecular Analyst Software Fingerprinting Plus; Bio-Rad Laboratories, Hercules, Calif.). The background was subtracted by the rolling disk method, and the normalized patterns obtained with each enzyme were combined to obtain a single pattern for each isolate or reference strain. The patterns were used to construct a dendrogram by using the unweighted pair group method using arithmetic averages (UPGMA) clustering algorithm using the Pearson correlation coefficient along with the fine optimization option. Bacterial isolates showing the same banding pattern were assembled to form a group. After distribution of the isolates according to their group and the sampling period (Table ​(Table2),2), the bacterial diversity was calculated for each sampling period by the Shannon diversity index (H) (17). For the diversity index calculation (see Fig. ​Fig.5),5), the following equation was used: H = 1/N Σ − (logNi − logN) Ni, in which N represents the total number of isolates in a sampling period and Ni is the number of isolates for a single group.

Amplified PCR products of representatives of each ARDRA group were randomly selected and ligated into the pCR2.1-TOPO cloning vector (Invitrogen, Burlington, Ontario, Canada). Ligated DNA was transformed into competent Escherichia coli TOP10 cells (Invitrogen) by electroporation (Gene Pulser, 25 μF, 2.5 kV; Bio-Rad). To determine positive clones, plasmid inserts were extracted by the method of Holmes and Quigley (11) except that isopropanol precipitation of DNA was replaced by two successive precipitations with ammonium acetate (7.5 M) and ethanol (100%). The extracts were digested with the restriction endonuclease EcoRI. Two microliters of the extracts were digested with 1 U of the restriction endonuclease in a total volume of 20 μl. The digestions were carried out as recommended by the manufacturer, with the appropriate restriction buffer at 37°C for 2 h. DNA fragments were resolved in 1% (wt/vol) agarose gels (BioShop Canada) with 1× Tris-acetate-EDTA buffer. DNA was visualized by transillumination with UV light after the gels were stained with ethidium bromide. Plasmid DNA of positive clones was then purified with the QIAprep Spin Miniprep kit (Qiagen, Mississauga, Ontario, Canada) according to the recommendations of the manufacturer.

The ARDRA method was first applied on reference strains obtained from American Type Culture Collection (Table ​(Table1)1) to verify whether the method was able to provide enough discrimination between the banding patterns of different genera and species of bacteria. The results of ARDRA analysis on reference strains (Fig. ​(Fig.2)2) show that the method was able to differentiate between reference strains at the genus level, with a similarity lower than 50% between strains of different genera. The method was also able to differentiate between species of the same genus, like Pseudomonas, where similarity was 70% or less. Species of the genera Burkholderia and Chryseobacterium seemed to be harder to differentiate, with similarities ranging between 70 and 80%.

As described for the reference strains, 190 maple sap isolates obtained from each of the sampling periods of the 2001 and 2002 seasons were subjected to ARDRA analysis by digestion of the amplified 16S rRNA gene with RsaI, HaeIII, and AluI. Similar banding patterns obtained after combination of the three independent digestions were grouped (Fig. ​(Fig.3)3) to finally obtain a total of 32 groups. Representatives of each group were selected for cloning and partial 16S rRNA gene sequencing so as to retrieve sequence similarity and bacterial identity from sequence databases (Table ​(Table2).2). As shown in Fig. ​Fig.3,3, each group presented a specific banding pattern, and groups further known to be of different genera (Table ​(Table2)2) were clearly differentiated, with a similarity lower than 70%. However, exceptions were found for groups 24, 25, and 31, where the similarity was between 70 and 80%, and for groups 21 and 22, where similarity was higher than 90%. These last groups were further known to be of different genera after 16S rRNA gene partial sequencing (Table ​(Table2)2) and nevertheless gave relatively similar banding patterns. Some groups formed according to ARDRA analysis were more easily distinguishable and have been differentiated at the species level. For instance, groups among the Pseudomonas genus (groups 13 to 19) had a similarity lower than 65% and were clearly differentiated. After partial sequencing of the 16S rRNA gene corresponding to a 525-bp fragment coding for the hypervariable V1, V2, and V6 regions (8) of ARDRA group representatives, closest identities of isolates were retrieved from sequence databases. Identity results showed a strong similarity with library sequences, close to 100% for each isolate (Table ​(Table2).2). Some isolates, like those of group 9, were strongly related to library sequences, with 100% similarity. Relatedness between maple sap isolate sequences is demonstrated in Fig. ​Fig.4,4, where isolates were clustered according to the organism class. Isolates were therefore clustered in six different classes: Actinobacteria, bacilli, β- and γ-Proteobacteria, Flavobacteria, and Sphingobacteria. Actinobacteria and γ-Proteobacteria encountered the highest number of different ARDRA groups, with 12 and 9, respectively.