The microbial diversity of this type of saline lakes has been studied primarily by focusing on the isolation and characterization of individual organisms with potential industrial applications (15, 16, 18). The phylogenetic diversity of Bacteria and Archaea has also been investigated (8), and variations in the bacterial community composition with the depth of a lake were recently analyzed (17).

Nonsaline alkaline environments are much more rare; their genesis is dependent on complex geological contexts, but they are probably all related to a single geochemical process known as serpentinization. Briefly, this process can be envisaged as the weathering, by CO₂-charged waters, of silicate minerals present in mafic and/or ultramafic rocks. This process is important for the development of the characteristics of a variety of environments, such as groundwater associated with kimberlite formation, as in the Lake Timiskaming and Kirkland Lake kimberlite fields in Canada (30), and groundwater and surface water associated with some ophiolites in northern California (33), the Semail in Oman (2), the Troodos ophiolictic complex in Cyprus (3), and the Maqarin site in Jordan (1). Recently, serpentinization has been implicated in the formation of a new class of seafloor hydrothermal system composed of variable mixtures of calcite, aragonite, and brucite, know as the Lost City Field, which is considered a potential environment for the emergence of life on the Earth's ocean floor (22). Remarkably, water associated with ophiolites and serpentinization activity has also been considered to be a habitat analog of Mars (33).

Most probably, in those environments, alkalinity is generated by the decomposition of the silicate minerals (for example, pyroxene and olivine) by the action of carbon dioxide-charged waters releasing Ca²⁺ and OH⁻ into solution. Magnesium is immobilized as serpentine or precipitated as MgOH (brucite), MgCO₃ (magnesite), CaMg(CO₃)₂ (dolomite), or Mg₃Ca(CO₃)₄ (huntite). Carbonate may be removed as CaCO₃ (calcite), but Ca²⁺ remains in excess, resulting in a Ca(OH)₂-dominated brine in which solid-phase Ca(OH)₂ is in equilibrium with soluble Ca²⁺ and OH⁻, maintaining an alkaline aqueous environment around pH 11 (19).

Studies of microbial diversity in such environments are very scarce. The only known microbiological analysis, which was of a somewhat preliminary nature, was carried out in Oman (2). Bath and colleagues considered the alkaline groundwaters of Oman to be a natural analogue for cement pore waters and aimed to understand and predict the behavior of a cement-based compound for use as a repository for radioactive waste (2).

The groundwater of Cabeço de Vide in the south of Portugal has its origin in an ophiolite-like lithological context and, as its chemical characteristics suggest, serpentinization activity. Our purpose was to phylogenetically characterize the heterotrophic aerobic cultivable populations present in a very rare Ca(OH)₂-enriched oligotrophic groundwater with a pH value of 11.4. Moreover, we tried to understand the pH dependence of the isolated populations.

Samples and sampling site. Groundwater samples were recovered at Cabeço de Vide, Portugal. The water was used in a spa for therapeutic purposes and was extracted from a borehole designated AC3.

Borehole AC3 was an artesian well that was vertically drilled to a 130-m depth (below the land surface); the groundwater was extracted by a pipe system with an associated hydraulic pump. The groundwater was generated in a very complex geological context (Fig. 1), most probably resulting from contact between ultramafic rocks and the carbonate sequence, such as calcareous to dolomite-type rocks, and surely was associated with the main regional NNE-SSW fault, resulting in dilute water with a very high pH value (11.4).

FIG. 1. Geological formations in the vicinity of the sampling area.

Results of a chemical analysis performed three times a year between 1991 and 2001 by a recognized state laboratory according to standard methods (5) are presented here with permission.

Water samples for bacterial analyses were recovered directly from the borehole on two occasions, September 1999 and February 2000. The water samples were transported to the laboratory at 4°C and were assayed within 8 h of sampling.

Enumeration and isolation of heterotrophic populations. Determinations of the total numbers of microorganisms present were performed by ethidium bromide staining (11), with counts being performed with a Leitz Laborlux epifluorescent microscope equipped with an I3 fluorescence filter block (blue excitation filter BP, 450 to 490 nm; dichromatic mirror, 510 nm; emission filter LP, 515 nm).

For the isolation and enumeration of cultivable bacteria, aliquots of 1, 10, and 100 ml of the water samples were filtered through 47-mm-diameter filters (Gelman Supor 200 sterile membrane; 0.2-μm pore size), the filters were plated on the surfaces of different isolation media, and the plates were then incubated at three different temperatures, 22, 30, and 37°C, for up to 10 days. The cultures were observed daily for enumeration, and all different morphological colony types were isolated. Isolates were purified by subculturing on corresponding isolation media (adjusted to the pH of isolation) and were stored at −70°C in 5% isolation media with 15% (vol/vol) glycerol.

Isolation media. As a result of preliminary attempts to isolate bacteria from the water under study, we decided to use the following basic medium formulations: alkaline nutrient agar (ANA); a medium devised by us, called alkaline buffered medium 2 (ABM2); and R2A (Difco Laboratories, Detroit, Mich.). Since we were dealing with a very oligotrophic environment, we decided to use 10×-diluted variations of the ANA and ABM2 media. The five resulting media were modified by adjusting the pH of each of the basic compositions to five different pH values, 6, 7, 8.5, 9.5, and 11, by using buffer solutions at a final concentration of 100 mM. Since each medium variant was incubated at the refereed three different temperatures, we actually attempted to isolate strains from seventy-five culture variations.

(i) Medium composition and preparation. ANA contained the following components per liter of medium: 5 g of peptone (Difco), 5 g of meat extract (Difco), 15 g of agar (Difco), and 100 ml of a specific buffer solution at a concentration of 1 M (Deutsche Sammlung von Mikroorganismen und Zellkulturen; www.dsmz.de/media/med031.htm).

ABM2 contained the following components per liter of medium: 5 g of yeast extract (Difco), 5 g of tryptone (Difco), 1 g of α-ketoglutaric acid monopotassium salt (Sigma), 15 g of agar (Difco), 100 ml of a macronutrient solution concentrated 10×, 10 ml of a micronutrient solution concentrated 100×; and 100 ml of a defined buffer solution (listed below) at a concentration of 1 M. The 10× concentrated macronutrient solution contained the following per liter: 1 g of nitrilotriacetic acid (Sigma), 0.6 g of CaSO₄·2H₂O (Merck), 1 g of MgSO₄·7H₂O (Merck), 0.8 g of NaCl (Merck), 1.03 g of KNO₃ (Merck), 6.89 g of NaNO₃ (Merck), and 1.11 g of NaHPO₄ (Merck). The 100× concentrated micronutrient solution contained the following per liter: 0.22 g of MnSO₄·H₂O (Merck), 0.05 g of ZnSO₄·7H₂O (Merck), 0.05 g of H₃BO₃ (BDH Analak), 0.0025 g of CuSO₄·5H₂O (Merck), 0.0025 g of Na₂MoO₄·2H₂O (BDH Analak), and 0.0046 g of CoCl₂·6H₂O (Merck).

The R2A medium (Difco) was used according to the manufacturer's instructions (Difco manual, Difco Laboratories), but 100 ml of a defined buffer solution (listed below) at a concentration of 1 M was added in a final volume of 1 liter of medium.

The buffer solutions were autoclaved separately from the other components of the different media and were mixed after cooling to 50°C.

The 10× diluted variants of the ANA and ABM2 media were prepared exactly in the same way, but only decimal parts of the components were used, with the exception of the buffer solution required to adjust the pH of the media, which was maintained at the same concentration.

(ii) Buffer solutions. Different buffer solutions were used to adjust the medium variants to the different pH values. All buffer solutions were prepared as stock solutions at 1 M according to the method of Gomori (12) and were autoclaved separately. For the preparation of each medium, 100 ml of each stock solution was used to reach a final concentration of 100 mM. The following buffer solutions were used to adjust the media to the different pH values: citrate buffer to adjust the different media to pH 6, phosphate buffer to adjust the different media to pH 7, Tris buffer to adjust the different media to pH 8.5, carbonate-bicarbonate buffer to adjust the different media to pH 9.5, and a single-carbonate KOH solution to adjust the different media to pH 11.

The pH value of each lot of medium was verified on the medium surface with a surface-testing Sentix Sur pH electrode (WTW, Hoskin Scientific) prior to use.

RAPD analysis of isolates. We used random amplified polymorphic DNA (RAPD) analysis as a primary method to group the isolates. Crude cell lysates were used as DNA templates, as described by Wiedmann-al-Ahmad et al. (38). Amplification reactions were performed in a total volume of 50 μl containing 1.5 U of Taq polymerase, 1.5 mM MgCl₂ (Pharmacia Biotech), a 0.2 mM concentration of each deoxynucleoside triphosphate, 0.6 μM primer OPA3 (5′-AGTCAGCCAC-3′), and 2.0 μl of crude cell lysate. Samples were subjected to 45 cycles of amplification (Perkin-Elmer model 240 thermocycler) as follows: 1 min at 94°C, 1 min at 45°C, and 2 min at 72°C, followed by a final extension step of 7 min at 72°C. The fragments were analyzed by electrophoresis in a 2% agarose gel in Tris-acetate-EDTA buffer.

Fatty acid analysis of isolates. The isolates were also grouped and characterized by fatty acid analysis. For that purpose, cultures were grown on plates of AMB2 buffered at the pH of isolation of each strain and were incubated in sealed plastic bags at 28°C in a water bath for 24 h. Fatty acid methyl esters (FAMEs) were obtained from fresh wet biomass by saponification, methylation, and extraction as described previously by Kuikendall et al. (24) and were then separated, identified, and quantified with MIS library generation software (Microbial ID Inc., Newark, N.J.). A dendrogram based on the fatty acid patterns was generated by clustering the Euclidean distances by the unweighted pair group method with an arithmetic average algorithm provided by the MIDI software package.

Morphological and biochemical characteristics and determination of G+C contents of DNAs of representative strains of the isolated populations. Cell morphology was examined by phase-contrast microscopy after cultivation on ABM2. Gram staining and catalase and cytochrome oxidase activities were determined as described by Smibert and Krieg (34).

DNAs for determinations of guanine-plus-cytosine (G+C) contents were isolated as described by Nielsen et al. (28). The G+C content of each DNA was determined by high-performance liquid chromatography as described by Mesbah et al. (27).

Phylogenetic analysis of representative strains of the populations. The extraction of genomic DNAs for 16S rRNA gene sequence determination, PCR amplification of the 16S rRNA gene, and sequencing of the purified PCR products were performed as described by Rainey et al. (29). Purified reaction mixtures were electrophoresed in a model 310 genetic analyzer (Applied Biosystems, Foster City, Calif.).

The quality of the 16S rRNA gene sequences was checked manually by the use of BioEdit sequence editor software (14), and the sequences were aligned against representative reference sequences of the most closely related members, obtained from the Ribosomal Database Project (6) and the European Molecular Biology Laboratory (EMBL), by use of the multiple-alignment CLUSTAL X software package (37). The method of Jukes and Cantor (20) was used to calculate evolutionary distances, phylogenetic dendrograms were constructed by the neighbor-joining method (31), and tree topologies were evaluated by performing bootstrap analysis (10) of 1,000 data sets by use of the MEGA2 package (25).

Determination of pH tolerance of representative strains of the isolated populations. The pH range for growth was determined on solid medium, as is generally done for alkaliphilic organisms. For comparison purposes, we chose the only medium on which all of the representative strains of the isolated populations could grow. For that reason, we used ABM2 buffered to pH values varying by 1 unit from 5 to 12.

For determinations of the quality of growth, 24-h cultures (representing the isolated populations) were resuspended in sterile buffer to a turbidity equivalent to a McFarland no. 2 standard (34). Portions of 0.1 ml were spread onto ABM2 agar adjusted to different pH values and then incubated at 30°C for 72 h. Growth was then quantitatively verified as follows: −, no growth; +, poor growth; ++, good growth; and +++, exuberant growth. After growth, the pHs of the cultures were confirmed with a surface-testing Sentix Sur pH electrode (WTW, Hoskin Scientific).

Nucleotide sequence accession numbers. The 16S rRNA gene sequences determined in this study were deposited in the EMBL data library under accession numbers AJ635359 and AJ717350 to AJ717393.

Chemical characterization of groundwater and enumeration of bacterial populations. The groundwater recovered over a period of 10 years at borehole AC3 in Cabeço de Vide had a very stable chemical composition dominated by ions such as hydroxide, chloride, sodium, and calcium; carbonate was also present, but at a lower concentration (Table 1). The resulting aqueous solution was highly alkaline (pH 11.4) but very dilute.

TABLE 1. Chemical composition and pH of the groundwater of Cabeço de Videa

During the sampling periods and on other occasions, bacteriological parameters that are usually determined to assess the bacteriological quality of the water, such as total and fecal coliforms, fecal streptococci, Pseudomonas aeruginosa, and clostridium spores, were never detected (results not shown).

The total cell concentrations determined by the ethidium bromide staining method were 6.0 × 10⁴ ± 3.4 × 10⁴ cells 100 ml⁻¹ and 7.2 × 10⁴ ± 4.4 × 10⁴ cells 100 ml⁻¹ for the first and second sampling periods, respectively.

As expected, the number of heterotrophic bacteria recovered was dependent on the medium composition, the pH of the medium used, and the incubation temperature. The numbers varied between 3.0 × 10⁰ CFU 100 ml⁻¹ (for ANA medium buffered at pH 11 and incubated at 37°C) and 8.0 × 10³ ± 2.0 × 10³ CFU 100 ml⁻¹ (for R2A medium buffered at pH 9.5 and incubated at 30°C). These particular values were verified for the first isolation period, and during the second isolation period only a little variation was observed.

Nevertheless, in general, the highest levels of recovery, reaching 10³ CFU 100 ml⁻¹, were verified in 10× diluted ANA (pH 9.5) and R2A (pH 8.5 and 9.5) media. Furthermore, independent of the medium or pH used, the highest recovery rates were obtained at incubation temperatures of 22 and 30°C.

In the first sampling period, 89 strains (labeled CV) were recovered from all isolation possibilities, while the second sampling period resulted in the isolation of 96 strains (labeled AC). Interestingly, the only medium capable of supporting the growth of the totality of the isolates was the ABM2 formulation, so for comparative purposes we used this medium to grow the isolates for further characterization.

Grouping the isolates. Primary grouping of the 185 isolates was performed by RAPD analysis in an attempt to find similar strains that may have been isolated several times. With this analysis, we recognized 45 different RAPD profiles (Fig. 2). This procedure, together with an analysis of the fatty acid profiles (Fig. 3) of the isolates, made it possible to identify strains with the same RAPD patterns and very similar FAME profiles; on the other hand, we also found strains with very similar FAME profiles but with distinct RAPD patterns. In order to choose representative strains of the populations recovered and to ensure the preservation of the maximum amount of diversity possible, we decided to further characterize all of the types defined by particular RAPD profiles.

FIG. 2. RAPD profiles of representative strains of the isolated populations.

FIG. 3. Dendrogram based on unweighted pair group average linkage of FAME profiles of the isolates.

Characterization of the recovered populations. Despite our attempt to isolate diverse morphotypes, the diversity found either in cell morphology or in colony morphology and pigmentation was rather discrete. All of the populations recovered were catalase positive, half of the strains were oxidase positive, and the vast majority of the populations were gram positive (Table 2). Only 10 isolates had a molar G+C content below 50%, and accordingly, the majority of the populations detected clearly belonged to high-G+C bacterial groups (Table 2).

TABLE 2. Characteristics of representative strains of the isolated populations

Curiously, a large part of the representative strains were isolated at a neutral or slightly alkaline pH, with only two strains isolated at pH 11 (AC6t and AC84) and, surprisingly, one strain (AC76) isolated at pH 6 (Table 2).

Phylogenetic analysis of the recovered populations. The majority of the isolates recovered were deeply related to the phylum Actinobacteria, some were related to Firmicutes (Fig. 4), and only five strains, isolated during the second sampling period, were related to gram-negative bacterial groups (Fig. 5).

FIG. 4. Phylogenetic dendrogram based on a comparison of the 16S ribosomal DNA sequences of the gram-positive representative isolates and some of their closest phylogenetic relatives. The tree was created by the neighbor-joining method. The numbers on the tree (more ...)

FIG. 5. Phylogenetic dendrogram based on a comparison of the 16S ribosomal DNA sequences of the gram-negative representative isolates and some of their closest phylogenetic relatives. The tree was created by the neighbor-joining method. The numbers on the tree (more ...)

The majority of the isolates were related to Microbacteriaceae family members. These strains were associated with the genera Microbacterium, Agrococcus, Frigoribacterium, Clavibacter, and Leifsonia. Isolates designated as belonging to group A and group C, which represented two of the three most frequently isolated strains, belonged to this category. While strains clustered in group C were phylogenetically related to the species Microbacterium kitamiense, the isolates clustered in group A were most closely related to the lineage containing Frigoribacterium and Clavibacter species. However, the values of 16S rRNA gene sequence similarity between group A representatives and the described taxa of these two genera were only about 95% (Fig. 4).

Another set of isolates represented populations related to different species in the lineage of the Micrococcaceae, namely, Micrococcus luteus, Citrococcus muralis, and Rothia dentocariosa, and others were related to various species of the genera Kocuria and Nesterenkonia (Fig. 4).

Other isolates, although belonging to the Actinobacteria, presented a more disperse relationship with several species of the genera Kytococcus, Janibacter, Actinomyces, Rhodococcus, and Dietzia. Moreover, a large number of isolates that constituted the designated group B were mostly found to be related to the species Dietzia natronolimnaea (Fig. 4).

Ten representative strains of the isolated populations were related to low-G+C gram-positive bacteria, namely, different species of the genera Staphylococcus and Bacillus, although some of the determined 16S rRNA gene sequence similarities with described taxa of these genera were only 95 and 96%, as for the isolates AC6t and CV53, respectively (Fig. 4).

Only five isolates were gram negative. Four were α-Proteobacteria related to some genera within the families Caulobacteriaceae and Sphingomonadaceae, namely, Brevundimonas, Phenylobacterium, and Sphingomonas, although the calculated similarity of the 16S rRNA gene sequence for the strain AC49 with the only described species of the genus Phenylobacterium, P. immobile, was only 95% (Fig. 5).

Finally, the other gram-negative isolate (AC74) was related to the lineage of the Flexibacteriaceae of the class Sphingobacteria (Fig. 5). The 16S rRNA gene sequence showing the highest pairwise similarity (93%) to the sequence of isolate AC74 was the type strain of a recently described species, Algoriphagus ratkowskyi, isolated from Antarctic sea ice (4).

pH tolerance for growth. Despite the pH determined in the environment, when we tested the tolerance of the various representative strains, only two (AC6t and AC84) were able to grow at pH 11 (Table 2). However, five isolates could grow at pH 6, and one of these (AC44) had the largest pH range, between 6 and 10 (Table 2).

All of the isolates grew at a pH of 8; in contrast, none grew at pH 5 or 12. Since most of the isolated strains belonged to groups A, B, and C, the majority of the populations recovered were able to grow in laboratory media at about pH 9 (Table 2).

Calcium-rich alkaline waters are extremely rare, and chemical and microbiological analyses of these environments are very scarce. However, because they were considered analogous to the effluents from cement manufacture, they were considered interesting as a model system for investigating the chemical and biological interactions in cement pore waters (2). Curiously, at present, they are also considered a habitat analog of Mars (33), and somehow the geological process that generates these type of environments is implicated in a new class of seafloor hydrothermal system (21, 22).

The groundwater found in Cabeço de Vide is generated by its intricate geological context, but a typical ophiolitic intrusion of mafic and ultramafic rocks in calcareous formations can be surmised, and in this circumstance, it is likely that serpentinization processes occur. Nevertheless, we found a chemically stable environment with a high pH where the most abundant ions are Cl⁻, OH⁻, Na⁺, and Ca²⁺. This peculiar water is in someway similar to those found in the alkaline springs in Oman by Bath et al. (2), but the concentrations of ions present in Cabeço de Vide groundwater are much lower than those determined for the more dilute springs in Oman. Another interesting feature is that carbonate was found in the water from Cabeço de Vide, which is unlikely to occur in this type of environment. Carbonate at high pH values in this type of chemical context is generally removed from solution by precipitation in the form of calcite (CaCO₃), magnesite (MgCO₃), dolomite [CaMg(CO₃)], or huntite [Mg₃Ca(CO₃)₄]. The presence of carbonate coexisting with calcium is probably an indicator of the complexity and uniqueness of this groundwater, or perhaps it can be explained by a continuous introduction of carbon dioxide into the liquid phase due to turbulence caused by the action of the hydraulic pump used to extract the groundwater, which at high pH values forms carbonate.

Nevertheless, independent of the particular genesis of the groundwater at the Cabeço de Vide spring, we have found an interesting alkaline water environment that is very dilute, highly oligotrophic, and extremely rare.

The maximum number of cultivable bacteria determined for this particular groundwater was very low, and even the total bacterial counts were low. The differences between the two counts were usually verified with water samples (11) and with other types of environment such as soil (9). In our particular case, the percentage of populations recovered was slightly higher than those in other studies (9). Nevertheless, the total numbers obtained in this water (≈10⁴) were approximately 100 times lower than those determined for the cultivable heterotrophic bacteria in Oman (2). However, while we recovered water samples directly from the borehole, in Oman the samples were recovered from small ponds and runoffs but not directly from springs. The presence of bacteria identified as Enterobacteriaceae may be an indication of contamination of the Oman water samples with organic materials. This may explain the relatively large differences between the bacterial counts for Oman and those that we determined. The presence of such organisms in the Oman samples may also indicate, as Bath et al. pointed out, that the bacterial diversity found there may not be the most characteristic of these types of environment (2). For Cabeço de Vide, we never found bacterial indicators of organic material, so there was probably no contamination of the aquifer or the groundwater with surface water and we may infer that the borehole is properly protected.

Our primary goal was to determine the bacterial diversity in the alkaline groundwater at Cabeço de Vide, namely, to measure the cultivable aerobic populations. Therefore, using different culture medium variants, we isolated a total of 185 bacterial strains, on two occasions, that surely represent an important fraction of the populations present in this particular environment. Although we have characterized 38 populations, probably belonging to 31 different species, the majority of the isolates were clustered into three distinct groups which may be considered the most frequent and characteristic populations present in this bacterial community. The assemblage of these three groups of isolates accounted for 81% of the recovered strains.

Group B included isolates that were closely related to the type strain of the species D. natronolimnaea (7), which was isolated from a soda lake in Africa. Curiously, only the strains belonging to group B, which represented the most frequently isolated populations, were phylogenetically related to a bacterial lineage isolated from another alkaline environment.

The strains of group A, accounting for 29% of the total isolates, were related to the Frigoribacterium and Clavibacter lineages. However, the determined 16S rRNA gene sequence similarities were low, and therefore the group A isolates constitute a previously undescribed new species of Actinobacteria.

The strains of group C, representing 20% of the isolates, were closely related to the type strains of the species Microbacterium kitamiense and Microbacterium aurantiacum, which were isolated from the wastewater of a sugar beet factory and from sewage, respectively (26, 36).

Another large fraction of the strains, which was isolated less frequently, was phylogenetically related to very distinct lineages of the high-G+C gram-positive bacteria, including species of the genera Agrococcus, Leifsonia, Microbacterium, Kytococcus, Janibacter, Kocuria, Rothia, Nesterenkonia, Citrococcus, Micrococcus, Actinomyces, and Rhodococcus.

Another set of isolates was related to low-G+C gram-positive bacteria of the genera Streptococcus and Bacillus. Strains associated with the latter were frequently isolated from soda lakes (35). In that type of alkaline environment, strains from the genera Streptomyces (18), Bogoriella (13), Nesterenkonia, Arthrobacter, Terrabacter (8), and the already referred-to Dietzia have also been found, and these probably constituted the populations of Actinobacteria that were most frequently isolated.

We only isolated five strains related to gram-negative lines of descent, and none were closely related to strains found in other alkaline environments.

A comparison of the diversity observed in this study with that observed in other studies is rather difficult because of the lack of studies of similar environments. Nevertheless, we isolated many strains that were related to a relatively large number of different species compared, for instance, with the number of species isolated from the alkaline springs in Oman (2). The number of species isolated in this study was also high, even compared with similar studies performed in other alkaline environments such as soda lakes or deserts (18, 19). Adding to the difficulty (if not the impossibility) of a comparison is the fact that in those environments the salinity as well as the alkalinity is an important limiting factor that has to be considered.

The evidence that the majority of the populations found in Cabeço de Vide groundwater are not related to other species isolated from alkaline environments is apparently paradoxical. Nevertheless, alkaliphilic bacteria are not confined to a single group or to a defined set of phylogenetic lineages; instead, they are distributed in various evolutionary branches (18, 19, 23). The real paradox is that strains isolated from an environment with a pH value of 11.4 did not grow under laboratory conditions in culture media adjusted to pHs higher than 11. However, this effect is relatively common, and various isolates from extreme alkaline environments only proliferate under laboratory conditions at more moderate pH values (7, 18, 19). Even with different types of organisms, such as Legionella and Aquicella, this effect has been noted, but for these it is hypothesized that intracellular growth in protozoa protects these organisms from pH variations in the environment (32). The inability of our isolates to grow at pH values similar to those found in the environment probably reinforces the general idea that the composition of the media and other conditions of incubation were decisive for the determination of the pH range and the optimal pH for growth (16, 23). Therefore, further studies are needed for more accurate determinations of the pH tolerance of strains isolated from alkaline environments such as the isolates from Cabeço de Vide.

Nevertheless, since all of the organisms were isolated from groundwater with a pH of 11.4, they should have particular adaptations that enable, if not proliferation, at least survival under these conditions. Moreover, the bacterial populations present in the alkaline water at Cabeço de Vide may be independent of the pH value of the environment. The buffering capacity of these waters upon exposure to CO₂ or as a result of microbial activity may be very limited; therefore, the presence of determined bacterial strains that are not adapted to live in high-pH environments may be supported or induced by the activities of other microbial strains. Probably, in this type of extreme environment, as in others, the most important factor that explains the presence or absence of a particular population is the relationships established by the whole community. This crucial factor is poorly understood and critically underestimated when we try to determine the particular characteristics of a single population.

This research was funded by FCT/FEDER project POCTI/BSE/42732/2001.

We are indebted to Milton Costa (University of Coimbra, Coimbra, Portugal) for advice and critical discussions. We thank Fernanda Nobre for help with the FAME analysis and M. Fontainhas and the Junta de Freguesia de Cabeço de Vide for permission to collect samples and to use the chemical analysis data presented here. We also thank Judite Fernandes for permission to include the geological map of the Cabeço de Vide area.