Formation and resuscitation of 'non-culturable' cells of Rhodococcus rhodochrous and Mycobacterium tuberculosis in prolonged stationary phase

It is generally believed that pathogenic, slow-growing mycobacteria, such as Mycobacterium tuberculosis or Mycobacterium leprae, can persist for long periods of time in vivo after initial infection, by passing into a dormant or non-replicating persistent state (Gangadharam, 1995 ; Parrish et al., 1998 ; Wayne & Sohaskey, 2001 ). Such dormant Mycobacterium tuberculosis cells can persist for many years in the host, causing latent tuberculosis before their transition to active, dividing bacteria occurs and disease is activated. It is perhaps surprising that the existence of a dormant state for mycobacteria either in vitro or in vivo has not yet been convincingly established. The most widely accepted in vitro model for the transition of Mycobacterium tuberculosis to an apparently dormant state is Waynes model, in which cells are subjected to gradual oxygen depletion. This results in the formation of an anaerobic, drug-resistant, non-replicating state (Wayne, 1994 ; Wayne & Hayes, 1996 ). Under these conditions, the putatively dormant bacteria shut down protein synthesis, but this restarts after the reintroduction of oxygen (Hu et al., 1998 ). A similar phenomenon has been found for Mycobacterium smegmatis (Dick et al., 1998 ) and Mycobacterium bovis (Lim et al., 1999 ). Mycobacterium tuberculosis, Mycobacterium smegmatis and Mycobacterium bovis revealed similar changes during the transition to anaerobic persistence. For example, the activity of some dehydrogenases was induced to maintain the NAD pool under anaerobiosis, an α-crystallin-like protein was expressed, and cell-wall thickening occurred in Mycobacterium tuberculosis and Mycobacterium bovis BCG (Cunningham & Spreadbury, 1998 ; Hutter & Dick, 1998 ; Lim et al., 1999 ). Another approach to obtain dormant Mycobacterium tuberculosis in vivo is the Cornell mouse model. Mice infected by Mycobacterium tuberculosis are treated with antibiotics, after which the number of culturable cells extractable from organs falls to zero (Wayne, 1994 ). Notably, mycobacterial DNA can still be detected in such apparently sterile tissues and quantitative PCR has revealed the presence of amounts of DNA equivalent to about 10⁵ organisms (g tissue)^-1 (de Wit et al., 1995 ). Subsequently, culturable Mycobacterium tuberculosis cells reappear, either spontaneously or in response to immune suppression (Flynn & Chan, 2001 ; McCune et al., 1966 ). The essential difference between the two models is the maintenance of the high culturability of Mycobacterium tuberculosis cells in Waynes model and the presence of a non-culturable state in the Cornell model. However, the absence of experiments in which the presumably dormant bacteria in the Cornell model can be resuscitated and enumerated in vitro creates some difficulties in data interpretation. On the one hand, PCR may detect a substantial population of non-culturable (dormant) cells that subsequently resuscitate in vivo. On the other hand, it may detect a large proportion of dead cells, in which case the re-emergence of colony-forming units would simply reflect the re-growth of a very small number of viable cells left after treatment with antibiotics (Kell et al., 1998 ). These two possibilities have not yet been distinguished experimentally.

Micrococcus luteus is a member of the Actinomycetales and is therefore related to Mycobacterium tuberculosis. This organism can persist in a dormant state, following growth to stationary phase in batch culture and starvation for several months in the spent growth medium (Kaprelyants et al., 1993 ; Kaprelyants & Kell, 1993a ). Dormant Micrococcus luteus cells have lost the ability to grow on agar plates; however, the addition of supernatant (SN) from growing Micrococcus luteus cultures to dormant cells has been shown to restore the ability to divide freely, thereby resuscitating the cells to normal, colony-forming bacteria (Kaprelyants & Kell, 1996 ; Kaprelyants et al., 1994 ). Micrococcus luteus cells secrete a resuscitation-promoting factor (Rpf), which promotes the resuscitation of dormant cells (Kaprelyants et al., 1996 ). Rpf has subsequently been shown to be a small protein that is active at picomolar concentrations (Kaprelyants et al., 1999 ; Mukamolova et al., 1998 ). Genes encoding Rpf-like proteins are widely distributed throughout the Actinomycetales; database searches have revealed that similar genes are present in mycobacteria, corynebacteria and streptomycetes (Kell & Young, 2000 ). If, as seems likely, the Rpf-like proteins of these organisms have similar functional roles to that of Micrococcus luteus Rpf, we may be able to predict the existence of a non-culturable, dormant state in other members of the Actinomycetales.

The aim of the present study was to find conditions for the in vitro transition of two organisms belonging to the Corynebacterineae (fast-growing R. rhodochrous and slow-growing Mycobacterium tuberculosis) to a non-culturable (possibly dormant) state. We also tested the hypothesis that secreted bacterial proteins may stimulate resuscitation of these two organisms.

Organisms and media.
Rhodococcus rhodochrous NCIMB 13805 was grown aerobically at 37 °C in shake flasks (125 ml medium in 750 ml flasks, agitation at 200 r.p.m.) in nutrient broth E (Lab M) or in modified Sautons medium. The latter contained (per litre): MgSO₄.7H₂O, 0·5 g; L-asparagine, 4 g; glycerol, 60 ml; iron ammonium citrate, 0·05 g; sodium citrate, 2 g; ZnSO₄.7 H₂O, 0·1 ml of a 1% solution in H₂O; K₂HPO₄.3 H₂O, 7·75 g; NaH₂PO₄.2 H₂O, 4·25 g. When the culture had reached stationary phase, agitation was continued at 37 °C for up to 2 weeks. The avirulent Academia strain of Mycobacterium tuberculosis (Ogloblina & Ravich-Birger, 1958 ) was obtained from the Phthysiopulmonology Center (Moscow). Cells were grown in loosely capped (plastic screw-caps) 14 ml test tubes containing 2 ml of unmodified Sautons medium supplemented with albumen, glucose and NaCl (ADC), without shaking at 37 °C (Connell, 1994 ). Cultures were held in stationary phase at 37 °C for up to 8 months. To obtain a population of single cells, cultures were passed through two Schott glass filters (50 µm and 20 µm, respectively) followed by a series of five nitrocellulose filters (4 µm, 2 µm, 1·5 µm, 0·8 µm and 0·45 µm, respectively; Whatman). To produce less-aggregated and oxygen-starved populations, Mycobacterium tuberculosis cells were also grown in 200 ml of the same medium in the presence of 0·05% Tween 80, with shaking (150200 r.p.m.) in 750 ml flasks fitted with rubber seals. Needles were employed for sterile sampling without oxygen input.

Scanning electron microscopy.
Cells were fixed for 30 min in 3% glutaraldehyde, followed by treatment with 0·1 M sodium cacodylate buffer (pH 7·2). Samples were air-dried, sputter-coated with 20 nm Au particles and examined using a GSM-840A (Japan) scanning electron microscope.

Respiratory activity, membrane energization and the permeability barrier.
The respiratory activity of the bacterial cells was determined by using the fluorescent redox dye CTC (5-cyano-2,3-ditolyl tetrazolium chloride; Polysciences), as described by Kaprelyants & Kell (1993b ). Briefly, the bacteria were incubated in the presence of 4 mM CTC (freshly prepared in 10 mM sodium phosphate buffer, pH 7·0) for 30 min at 37 °C. Fluorescence was monitored with a Nikon fluorescence microscope with excitation at 530550 nm. Cells were stained with propidium iodide (4 µM in phosphate buffer), to assess the state of the membrane permeability barrier. Rhodamine 123 was used to monitor membrane energization, as previously described (Kaprelyants & Kell, 1992 ). Rhodamine 123 accumulation was sensitive to the uncoupler carbonylcyanide-3-chlorophenyl-hydrazone (CCCP). Cells were studied under the fluorescence microscope (excitation at 510560 nm and emission at 590 nm for propidium iodide; excitation at 450490 nm and emission at 520 nm for rhodamine 123).

Estimation of the redox potential.
The changes in the oxidation/reduction potential of the growth medium during R. rhodochrous growth were measured in undiluted samples (3 ml) taken from the bacterial culture. Measurement of the redox potential was done with a platinum electrode (and an AgCl reference electrode; normal potential at 20 °C was 201±2 mV) by using an I-130.2M ionometer (Soyusanalitpribor) under aerobic conditions at 20 °C, with intensive stirring until a constant value was reached.

Assessment of cell viability.
Bacterial suspensions were serially diluted in growth medium and duplicate 100 µl samples were plated in triplicate onto agar-solidified nutrient broth E (for R. rhodochrous) or onto agar-solidified, ADC-supplemented Sautons medium (for Mycobacterium tuberculosis). Plates were incubated at 37 °C. After 5 days of growth for R. rhodochrous and after 2 months of growth for Mycobacterium tuberculosis, c.f.u. values were enumerated. The detection limit was 5x10⁰ c.f.u. ml^-1. The same serially diluted samples of Mycobacterium tuberculosis were also employed for resuscitation and most probable number (MPN) assays. Using a fresh pipette tip each time, 100 µl samples were taken from each dilution. Ten of these samples were added to 10 replicate 10 ml Pyrex screw-cap tubes, each containing 2 ml Sautons medium supplemented with ADC. Rpf was added to five of the tubes to give a final concentration of 100 pM. Five additional samples were added to five replicate 10 ml Pyrex screw-cap tubes containing 2 ml filter-sterilized SN taken from cultures of Mycobacterium tuberculosis in the late-exponential phase of growth (3040 days post-inoculation). The 15 tubes for each dilution were incubated at 37 °C without shaking for 2 months.

Resuscitation and MPN assays for R. rhodochrous were performed in 48-well plastic plates (Corning) containing either 0·5 ml modified Sautons medium (see above) or 0·5 ml filter-sterilized SN taken from R. rhodochrous cultures. Some wells contained 125 pM Rpf. All wells were supplemented with 0·5% yeast extract (LabM). Appropriate serial dilutions of R. rhodochrous cells (50 µl) were added to each well. Plates were incubated at 37 °C, with agitation at 150 r.p.m. for 5 days. Tubes or wells with visible bacterial growth were counted as positive, and MPN values were calculated using standard statistical methods (de Man, 1975 ).

Spent medium preparation.
SNs were obtained from R. rhodochrous and Mycobacterium tuberculosis cultures harvested at the times indicated in Results. The R. rhodochrous SN was subjected to centrifugation (12000 g, 20 min). The SNs were sterilized by passage through a 0·22 µm filter (Whatman) before use. Proteinase treatment of the SNs was performed at 37 °C by adding trypsin (200 µg ml^-1) or proteinase K immobilized on acrylic beads (0·04 U ml^-1; Sigma), followed by the addition of trypsin inhibitor (400 µg ml^-1) to samples containing trypsin. The time of incubation was 30 min for trypsin-treated SNs and 3 h for proteinase-K-treated SNs.

Bacterial counts.
The total number of Mycobacterium tuberculosis cells was determined microscopically using a Helbers chamber. A minimum of 100 bacteria was counted and the SD of the total counts was not in excess of 20%.

Preparation of recombinant Micrococcus luteus Rpf.
The Rpf protein of Micrococcus luteus (histidine-tagged recombinant form) was obtained as described by Mukamolova et al. (1998) . Mono Q ion-exchange purification was omitted in some experiments. The purified protein was stored in 10 mM Tris/HCl (pH 7·4) containing 50% (v/v) glycerol at -20 °C for up to 2 weeks, and the protein concentration was determined spectrophotometrically. Before use, all preparations were screened for growth-promoting activity using a small inoculum of Micrococcus luteus, as described by Mukamolova et al. (1998) . Some preparations had poor activity; only those with substantial activity were employed for these experiments.

Preparation of antibody columns.
Rabbits were immunized three times at 3-week intervals by subcutaneous injection with 1 ml of a 50% (v/v) mixture of Rpf (1 mg ml^-1 in water) and incomplete Freunds adjuvant (Sigma). Serum was collected 10 days after the last immunization, and the immunoglobulin fraction was purified by a standard protocol using polyethylene glycol. Similar methods were used by Micropharm (Newcastle Emlyn, Carmarthenshire, UK) to raise and purify antibodies to a truncated version of Rpf (residues A₃₉L₁₁₅) in sheep. Immunoglobulins were conjugated to CNBr-activated Sepharose 4B (Sigma) (Osterman, 1985 ).

Formation of non-culturable cells of R. rhodochrous
R. rhodochrous cells were initially grown in the rich medium, nutrient broth E, for different times and then employed as an inoculum (10⁶ cells ml^-1) for growth in modified Sautons minimal medium (the elevated concentrations of phosphates in this medium stabilize the culture pH during stationary phase). Stationary phase was established within about 24 h, and further incubation resulted in a gradual decrease of the c.f.u. value, starting about 30 h post-inoculation (Fig. 1a). A minimum value was reached between 80 and 100 h post-inoculation, after which time the c.f.u. value gradually recovered until it finally exceeded the initial value. After the initial growth phase, the total count (estimated microscopically; data not shown) and the optical density remained almost constant, except that the optical density fell slightly, coincident with the period of non-culturability (Fig. 1a). The precise time at which culturability (c.f.u. counts) reached its minimum was variable from one experiment to another, but it correlated with a transient increase in the redox potential of the culture (Fig. 1b). The extent of the transient loss of culturability depended on the age of the inoculum. The largest effect was observed with an inoculum taken from a culture that had itself been inoculated some 1819 h previously (Fig. 1c). Optimum conditions for observing the loss of culturability (i.e. the biggest effect) were as indicated in Methods. However, loss of culturability was observed over a range of agitation speeds, between 100 and 250 r.p.m., and at culture volumes corresponding to between 12% and 25% of the volume of the culture flask.

(22K): Fig. 1. R. rhodochrous cells grown in batch culture in modified Sautons medium show a transient loss of culturability. A culture in nutrient broth E (inoculated 18 h previously) was employed as inoculum (106 cells ml-1) in (a) and (b) and the number of colony-forming units were monitored over time. The OD600 (a) and Eh (b) were also monitored. In (c), the effect of inoculum age on the maximum loss of culturability is shown. This experiment was repeated 15 times; the results of a typical experiment are shown. For further details see Methods. The SD values for c.f.u. determinations were generally within the limits almost covered by the symbols in the Figure; for example, they were 38%, 3% and 11% for the points at 24 h, 48 h and 96 h, respectively, in part (a). The SD values on measurements of Eh and OD600 in stationary phase were about 3% and 5%. (a) •, OD600 value; , c.f.u. value. (b) •, Eh value; , c.f.u. value.

Microscopic examination of the cells during the period of non-culturability revealed that the majority were ovoid or coccoid forms; this morphology contrasted with the characteristic elongated curved rod shape of exponentially growing cells (Fig. 2a). Only about 10% of the cells in the population were metabolically active [according to fluorescence microscopy with Rhodamine 123 as an indicator of membrane energization (Kaprelyants & Kell, 1992 ) and CTC as an indicator of respiratory activity (Kaprelyants & Kell, 1993b )]. Between 80% and 90 % of the cells failed to stain with propidium iodide, indicating that their membrane permeability barrier was intact (Davey et al., 1999 ).

(141K): Fig. 2. Scanning electron micrographs of populations of non-culturable cells of R. rhodochrous and Mycobacterium tuberculosis. R. rhodochrous cells (a) were grown in modified Sautons medium and sampled when the c.f.u. value had attained its minimum (see Fig. 1). Mycobacterium tuberculosis cells (b) were grown for 4 months in ADC-supplemented, unmodified Sautons medium, as described in Methods, and cells that were not retained by a 2 µm filter were prepared for electron microscopy. The insets in (a) and (b) are bacteria from exponentially growing cultures of R. rhodochrous (20 h post-inoculation) and Mycobacterium tuberculosis (1 month post-inoculation), respectively.

Resuscitation of non-culturable cells of R. rhodochrous
Resuscitation of cells taken during the period of non-culturability (Fig. 1a) was performed in liquid medium, using procedures previously developed for resuscitating non-culturable cells of Micrococcus luteus (Kaprelyants et al., 1994 ). Fig. 3(a) shows that dilution of the culture into fresh Sautons medium containing 0·5% yeast extract followed by incubation for up to 40 h did not give any evidence of resuscitation. Addition of catalase (3 µg ml^-1) or pyruvate (0·5 mM) to the medium did not promote resuscitation either in liquid or on solid medium. Resuscitation was observed when the cells were incubated in yeast-extract-supplemented (0·5%), filter-sterilized SN taken from a late-exponential phase culture (20 h post-inoculation) of bacteria grown in Sautons medium (Fig. 3a). The resuscitation activity of the SN depended on its concentration; a maximal effect was observed when the proportion of SN to cells was 50% (v/v) and above (Fig. 3b). For a numerical estimation of the resuscitation of non-culturable cells, MPN values were determined. This procedure (Kaprelyants et al., 1994 ) permits true resuscitation to be distinguished from the re-growth of a small number of initially viable cells, since it occurs at high dilutions where initially viable cells are absent. Incubation of the non-culturable cells in Sautons medium resulted in a 10-fold increase in viability measured by the MPN method when compared with c.f.u. counts (Fig. 4a). Different variations in medium composition did not result in any further increase of the MPN value. Values were similar in Sautons medium and in a rich medium (nutrient broth E) (Fig. 4a), and the addition of ADC supplement or yeast extract at concentrations greater than 0·5% (v/v) was inhibitory (data not shown). However, a dramatic increase in MPN values was observed when non-culturable cells were incubated in SN taken from bacterial cultures (Fig. 4a). The age of the SN was critically important. SN taken from cultures 1820 h and 118132 h post-inoculation was active, whereas that taken from cultures 4570 h post-inoculation was inactive (Fig. 4b). Moreover, the SN taken between 45 and 70 h contained some inhibitory material(s) that resulted in a decrease of the MPN value, compared with the value obtained in the absence of SN (Fig. 4b). The inhibitory SN also suppressed the growth of exponential cells of R. rhodochrous (data not shown). The speed of agitation of the 48-well plates used to measure MPN value was also critical for maximum resuscitation. The optimal speed was 150 r.p.m.; above or below this speed, the MPN value was significantly decreased.

(19K): Fig. 3. Effect of culture SN on the viability of non-culturable cells of R. rhodochrous. A culture of R. rhodochrous was harvested during the period of minimum culturability. (a) Bacteria were inoculated (106 cells ml-1 total count) in 10 ml of fresh modified Sautons medium (), or in filter-sterilized SN (•) taken from an exponentially growing culture. The cultures were incubated at 37 °C (150 r.p.m.) in 30 ml flasks; samples were then removed periodically for the determination of c.f.u. values. In (b), the MPN of the culture at the point of minimum culturability was determined in modified Sautons medium mixed with various proportions of SN. The SD values for c.f.u. determinations were <20%. For MPN assays±SD values, expressed as a percentage of the mean, are shown. These experiments were repeated several times and representative results are shown.

(24K): Fig. 4. Resuscitation of non-culturable cells of R. rhodochrous. (a) An R. rhodochrous culture was harvested during the period of minimum culturability and the total and c.f.u. counts were determined. MPN assays (columns with bold outline) were performed in nutrient broth E, modified Sautons medium containing 0·5% yeast extract with and without recombinant Rpf (125 pM), and SN taken from exponential (SN1) and stationary phase (SN2) cultures of R. rhodochrous (1820 h and 118132 h, respectively). The results of 35 different experiments are summarized, ±SD. Resuscitation was also measured using SN from cultures of different ages (b). The c.f.u. and MPN values for the cells used in part (b) in the absence of SN were 1·5x105 and 1·3x106 cells ml-1, respectively. The results of a typical experiment are shown.

Some properties of the resuscitation-promoting activity present in the culture SN are shown in Table 1. Autoclaving or boiling the SN for 20 min resulted in a loss of activity. Activity was lost during storage at 4 °C over several days. The resuscitation-promoting activity was sensitive to proteolytic enzymes, removed by ultra-filtration (molecular mass >10 kDa) and partially retained by dialysis (molecular mass >10 kDa). The passage of the SN through affinity columns containing immobilized anti-Rpf (or anti-truncated-Rpf) antibodies resulted in a substantial loss of activity, suggesting that part of the observed resuscitation activity was due to the presence of Rpf-like proteins. This conclusion was reinforced by the finding that picomolar concentrations of recombinant Rpf also stimulated resuscitation in Sautons medium supplemented with 0·5% yeast extract, although the effect was less pronounced than that of culture SN (Fig. 4a).

Table 1. Influence of different treatments on resuscitation activity of SN taken from exponentially growing cultures of R. rhodochrous

Formation of non-culturable cells of Mycobacterium tuberculosis
Mycobacterium tuberculosis (Academia strain) was grown in Sautons medium supplemented with ADC, without agitation in capped test tubes. After growth for 1 month (early-stationary phase), samples were taken periodically (one tube per time point) and the total count and c.f.u. count were estimated. Since the cells were grown in medium lacking detergent, large aggregates (and sometimes also a surface pellicle) were produced. To facilitate the manipulation of individual cells, cultures were passed through a series of filters of decreasing pore size (200·45 µm). Moreover, we expected that cells with a non-culturable phenotype would have a reduced size, as has previously been observed with Micrococcus luteus (Mukamolova et al., 1995 ), Mycobacterium tuberculosis isolated from murine macrophages (Biketov et al., 2000 ) and R. rhodochrous (see above). The culturability of the cells (estimated as the ratio c.f.u. count/total count) decreased with decreasing filter pore size. The 1·5 µm filtrate was unable to form colonies on agar (Fig. 5), despite the fact that microscopy revealed the presence of large numbers of cells (total count 4x10⁷ cells ml^-1).

(21K): Fig. 5. Viability of Mycobacterium tuberculosis in relation to cell size and culture age. (a) A culture of Mycobacterium tuberculosis (4 months post-inoculation) was filtered consecutively through 50, 4, 2, 1·5 and 0·45 µm filters and the c.f.u. count/total count for each of the filtrates was determined. In part (b), actively growing cells of Mycobacterium tuberculosis were inoculated (105 cells ml-1) into Sautons medium supplemented with ADC, and replicate cultures were incubated at 37 °C without agitation for 8 months in capped test tubes. Periodically, samples were withdrawn and filtered through a 2 µm filter. The c.f.u. (; SD between 10% and 20%) and total (; SD about 17%) counts of the filtrates were determined. The results of a typical experiment are shown.

Non-culturable cells in the 1·5 µm filtrate were a mixture of ovoid cells, with a length of about 1·1 µm, and small individual cocci, with a diameter of 0·50·7 µm (Fig. 2b). Both types of cells had an intact membrane barrier (according to propidium-iodide staining) and did not reveal any respiratory activity (not stained by CTC). As seen with R. rhodochrous, the degree of non-culturability of the cells in the 1·5 µm filtrate was time dependent (Fig. 5b). Culturability was minimal about 4 months post-inoculation, after which time the c.f.u. count gradually increased and approached the value of the total count after 8 months of cultivation. The total count of bacteria in this fraction did not change significantly during the entire period of the experiment (Fig. 5b). Oxygen availability was certainly restricted in the tube cultures used for this experiment, because incubation of Mycobacterium tuberculosis in stationary phase in oxygenated shake flasks resulted in the maintenance of significant viability (about 10⁷ c.f.u. ml^-1) of cells in the 1·5 µm filtrate over a period of 8 months.

To avoid problems associated with heterogeneity, we also determined whether similar non-culturable cells could be formed in the more homogeneous cultures that are produced when Mycobacterium tuberculosis is grown in the presence of a suitable detergent. To this end, cultures were grown in flasks containing ADC-supplemented Sautons medium with 0·05% Tween 80, which were sealed with a rubber cap and agitated at 200 r.p.m. Sampling was done using a syringe needle to avoid introducing oxygen from the laboratory atmosphere. Under these conditions, we also found a decrease in the c.f.u. count in the total cell population, with kinetics similar to those shown in Fig. 5(b) for filtered cells. Moreover, the culture contained many of the typical ovoid and coccoid cells after 4 months of cultivation (not shown). When such cells were plated onto agar medium c.f.u. counts of zero were obtained, even after the plates were incubated for a 2 month period (Fig. 6a). However, in contrast to cells grown in static culture without Tween 80, this state of non-culturability was maintained for the remainder of the experimental period (48 months post-inoculation).

(21K): Fig. 6. Production and resuscitation of non-culturable cells of Mycobacterium tuberculosis in Sautons medium. Growing cells of Mycobacterium tuberculosis were inoculated (105 cells ml-1) in ADC-supplemented Sautons medium containing 0·05% Tween 80 (a) or no Tween (b) and were grown for 45 months, as indicated in Methods. (a) Samples of comparatively homogeneous cultures were withdrawn periodically and the c.f.u. values were determined without prior filtration (SD between 10% and 20%). This experiment was repeated twice with similar results. In part (b), the bacterial population at 4 months (post-inoculation) was passed though a 2 µm filter and the total and c.f.u. counts of the filtrate were determined. MPN assays (columns with bold outline) were performed in ADC-supplemented Sautons medium with and without recombinant Rpf (125 pM) and also in SN taken from an actively growing culture of Mycobacterium tuberculosis (34 weeks post-inoculation). The results of 20 independent experiments are summarized, ±SD.

Resuscitation of non-culturable cells of Mycobacterium tuberculosis
Resuscitation of the ovoid and coccoid cells in the 1·5 µm filtrate of 4-month-old cultures was performed in capped test tubes containing liquid Sautons medium supplemented with ADC, using an MPN assay. In contrast to Micrococcus luteus or R. rhodochrous, estimation of viability in liquid medium (MPN) led to a substantial (5 log) increase in the number of viable cells (Fig. 6b). The presence of picomolar concentrations of recombinant Rpf occasioned an additional 12 log increase in cell viability. SN taken from a culture of Mycobacterium tuberculosis cells growing exponentially in ADC-supplemented Sautons medium produced an equivalent increase in the viable count. In some experiments, such as that shown in Fig. 6(b), viability estimated using the SN was equivalent to the total bacterial count. We were unable to obtain resuscitation of cells that passed through the 0·45 µm filters under any of the above conditions (data not shown).

Mycobacterium tuberculosis cells grown in the presence of Tween 80 were also able to resuscitate in liquid medium. Table 2 shows that 5-month-old and 6-month-old cells, which had completely lost the ability to form colonies on solid medium, were resuscitated in liquid medium. An additional 12 log increase in resuscitation was obtained in the presence of SN. These results are very similar to those observed with cells obtained by filtration of heterogeneous cultures.

Table 2. Resuscitation of Mycobacterium tuberculosis grown with agitation and in the presence of Tween 80

The phenomenon of non-culturability of bacterial cells is a matter of intensive debate [for reviews see Barer (1997) , Barer & Harwood (1999) , Barer et al. (1998) and Kell et al. (1998) ]. There are many publications claiming the existence of viable but non-culturable (VBNC) bacteria, including several pathogenic species of bacteria. However, only in relatively few cases has the transition of such bacteria to a normal, viable (culturable) state been shown. The great majority of the experiments published to date fail to provide an unequivocal demonstration of the resuscitation of non-culturable bacteria (Kell et al., 1998 ). Unless such VBNC cells can be resuscitated, their microbiological and medical significance must be considered highly questionable. Cells with non-culturable phenotypes may either maintain some detectable activity or persist in a dormant, inactive state. Previously we found that stationary phase cultures of Micrococcus luteus contain a large proportion of non-culturable cells that remained in a dormant state for several months (Kaprelyants & Kell, 1993a ). In this study, we used similar methods to obtain non-culturable cells of R. rhodochrous and Mycobacterium tuberculosis, two members of the Corynebacterineae related to Micrococcus luteus.

As we found for Micrococcus luteus, rather strict culture conditions must be satisfied to observe the transition of the majority of the cell population to a non-culturable state (Mukamolova et al., 1998 ). For R. rhodochrous, inoculum age (Fig. 1b), medium composition, speed of flask agitation and culture volume in relation to the flask capacity significantly influenced the formation of non-culturable bacteria. However, in contrast to Micrococcus luteus, non-culturability was a transient phenomenon for R. rhodochrous cells. For each culture, there was a window of several hours during which about 99·9% of the bacteria in the culture had lost culturability. Some previously published results also show a transient decrease in the c.f.u. count when a fast-growing organism, Mycobacterium smegmatis, is held in stationary phase (Dick et al., 1998 ; Smeulders et al., 1999 ). The authors did not draw this specific conclusion from their data and considered this as a death phase. Recently, Keer et al. (2000) have found that several mutants of Mycobacterium smegmatis, defective in stationary phase survival under aerobic or anaerobic conditions, revealed a transient decrease in viability followed by an increase in culturability. The authors suggest at least two possible explanations for this effect: (a) re-growth (cryptic growth) of viable cells, and (b) formation of non-culturable (and possibly dormant) bacteria in stationary phase and their subsequent resuscitation (Keer et al., 2000 , 2001 ). Since the authors did not measure culturability by MPN counts, direct evidence was neither sought nor obtained for resuscitation of non-culturable cells.

In our study, we applied MPN assays to measure viability. This method permits numerical estimation of the number of resuscitatable cells in liquid medium at high dilutions, effectively circumventing the problem of re-growth of residual viable cells during resuscitation (Kaprelyants et al., 1994 ; Kell et al., 1998 ). This assay revealed the presence of non-culturable cells of R. rhodochrous during a rather short time interval in stationary phase. The transient character of the loss of cell viability during stationary phase suggests that there are sequential processes occurring. First, there is a transition to a non-culturable state. This then seems to be followed by the (cryptic?) growth of residual viable bacteria, possibly accompanied by the growth of resuscitated bacteria after 8696 h of incubation. The individual contributions of (a) residual viable cells and (b) resuscitated cells to the observed growth could vary, depending on the particular set of culture conditions (inoculum age, aeration conditions, growth medium, etc.). Non-culturable cells may be formed under many different conditions, but we have found conditions where resuscitatable cells make up the majority of the bacterial population (and are therefore experimentally accessible). Several questions remain. For example, why do R. rhodochrous cultures start to lose their culturability? Why do the viable bacteria fail to grow during the period of transition to the non-culturable state, and why do they start to grow subsequently? During the transition to non-culturability, we found evidence for the presence of an inhibitory component(s) in the SN, which is no longer present once the c.f.u. counts start to increase again. Further studies are needed to elucidate the possible role of accumulation/destruction of this inhibitor in culture behaviour.

The extent of resuscitation often exceeded the proportion of active cells in the population (about 10% in all experiments, according to Rhodamine-123 and CTC staining) and in some experiments approached 100%. This fact indicates that inactive cells, with a morphology characteristic of dormancy (small coccoidal cells with enhanced phase-contrast; Mukamolova et al., 1995 ), can resuscitate. Therefore, we suggest that incubation of R. rhodochrous cells under the conditions described resulted in the formation of a population of dormant bacteria, with a characteristic non-culturable phenotype.

Mycobacterium tuberculosis held in stationary phase in capped tubes also revealed the transient formation of non-culturable cells (on a different time scale). However, because of the marked non-homogeneity of Mycobacterium tuberculosis cultures grown without detergent, filtration was required to isolate the fraction of non-culturable cells. The filtered fraction containing non-culturable bacteria may not include all such bacteria in the population; aggregates, unable to pass through the filter, may also contain such bacteria. The increase in the c.f.u. count after 4 months of incubation in stationary phase (Fig. 5b) could be due to the growth of initially viable cells, to the resuscitation of non-culturable cells, or to a mixture of both of these factors in the highly heterogeneous population. The substantially homogeneous populations grown in the presence of Tween 80 were unable to form colonies for a long period of time (Fig. 6a). We suspect that it may also be possible to establish stable populations of non-culturable cells for R. rhodochrous in stationary phase, but the appropriate conditions for this were not identified here.

There is a substantial difference between the non-culturable cells of the two species studied, in respect of their ability to resuscitate. For R. rhodochrous, SN was needed for resuscitation (Fig. 4a), whereas significant resuscitation (up to 5 logs) of Mycobacterium tuberculosis cells occurred in liquid Sautons medium without any additions (Fig. 6b). SN or Rpf further increased resuscitation of Mycobacterium tuberculosis cells by up to 2 logs (Fig. 6b). These data clearly demonstrate the fact that the term non-culturability has an operational meaning only (Barer, 1997 ; Kell et al., 1998 ); Mycobacterium tuberculosis cells that were unable to form colonies on plates were able to grow in liquid medium, as was reported by Biketov et al. (2000) .

As previously seen with Micrococcus luteus (Kaprelyants et al., 1994 ), SN taken from exponentially growing cells of both species studied possessed a resuscitation-promoting activity. In R. rhodochrous SN, activity was associated with a small protein, which probably belongs to the Rpf family as SN passed through an affinity column containing anti-Rpf antibodies had significantly reduced activity. Indeed, immunoblotting of this SN after SDS-PAGE revealed several protein bands that interacted with specific anti-Rpf antibodies (M.O. Shleeva, unpublished results). Moreover, Southern hybridization of BamHI-, PstI-, PvuII- or XhoI-digested R. rhodochrous DNA using a rpf-specific probe revealed the presence of four hybridizing bands (data not shown). Recombinant Micrococcus luteus Rpf stimulated resuscitation of R. rhodochrous, but it was less active than SN.

A similar conclusion could be made with respect to the active compound in Mycobacterium tuberculosis SN, as Rpf also showed stimulatory activity (although it was less active than SN; Fig. 6b). Moreover, the Rpf-like proteins of Mycobacterium tuberculosis do indeed have growth factor activity and are able to stimulate the growth of non-culturable cells of Mycobacterium bovis BCG held in prolonged stationary phase (G.V. Mukamolova, unpublished results). In addition, non-culturable cells have also been found after persistence of Mycobacterium tuberculosis in murine macrophages ex vivo. The viability of these cells was restored by incubation in the presence of Rpf in liquid medium (Biketov et al., 2000 ).

Zhang and co-workers have recently reported that non-culturable cells accumulate during prolonged incubation of Mycobacterium tuberculosis H37Ra cells in stationary phase in the presence of Tween 80 (Sun & Zhang, 1999 ; Zhang et al., 2001 ). In contrast to our study, their cell populations contained a significant number of colony-forming units. Slight differences in culture conditions may account for this (e.g. we maintained bacteria in capped tubes without oxygen input). They showed that SN from growing Mycobacterium tuberculosis cultures contained phospholipids, which increased the number of colony-forming units in aged cultures when added to plates and permitted growth from low inocula in liquid medium. Another active compound isolated from the SN was an 8 kDa protein, Rv1174c. Chemically synthesized peptides corresponding to three different segments of Rv1174c were added at micromolar concentrations (i.e. much higher than those at which Rpf is active) to liquid medium inoculated with old cells. After 510 days incubation, the number of recoverable colony-forming units was enhanced. We suggest that the phospholipids and micromolar concentrations of peptides employed by the authors may have promoted repair of injured cells (this may be especially true for the dried cells with zero c.f.u. counts used in this study) (Ray & Speck, 1973 ). Since the authors did not measure viability by MPN, re-growth of a residual portion of viable cells cannot be excluded in some experiments and the extent of the proposed resuscitation activity cannot be quantified.

The transition of mycobacterial cells to dormancy in the Wayne model in vitro is based on the formation of non-replicating bacteria during adaptation to anaerobic conditions (Wayne, 1994 ; Wayne & Hayes, 1996 ; Wayne & Sohaskey, 2001 ). Protein synthesis in this state is generally reduced (Hu et al., 1998 ), although certain genes, such as acr, are up-regulated (Desjardin et al., 2001 ; Hu et al., 1999 ; Yuan et al., 1996 ). The model has been established in several different species, viz. Mycobacterium tuberculosis, Mycobacterium smegmatis and Mycobacterium bovis BCG (Dick et al., 1998 ; Lim et al., 1999 ; Wayne, 1994 ). However, in contrast to our study, mycobacterial cells in Waynes model retain high viability and develop sensitivity to metronidazole as they become anaerobic (Wayne & Sramek, 1994 ), indicating that they remain metabolically active. Therefore, they may not be considered truly dormant, but rather in a state of (O₂) starvationsurvival (Kaprelyants et al., 1993 ). We suggest that the difference in culturability may reflect the different times of persistence of Mycobacterium tuberculosis cells in a non-growing state (days in Waynes model and months in our study). Restoration of culturability requires resuscitation in liquid medium, which may either occur spontaneously or require the provision of compounds (e.g. growth factors) present in the SN of growing cells. The proportion of cells able to resuscitate spontaneously and those requiring growth factors may vary from one experiment to another, and resuscitation may depend on the particular conditions employed for the establishment of non-culturability.

Prolonged incubation of Mycobacterium tuberculosis cells in stationary phase may reflect more accurately the situation in vivo. In this connection we may stress that the essence of the Cornell model of dormant Mycobacterium tuberculosis is the creation of a period of sterility in vivo, induced by the administration of antibiotics after infection, when Mycobacterium tuberculosis cannot be detected by c.f.u. counts (Wayne, 1994 ) (in contrast to PCR; de Wit et al., 1995 ). As shown in Fig. 7, Waynes model may possibly represent a step towards the establishment of true dormancy in Mycobacterium tuberculosis, which is characterized by the emergence of non-culturable cells that need to be resuscitated before they can resume active growth. The necessity for anaerobic conditions to produce non-culturable cells of Mycobacterium tuberculosis in our experiments may reflect a mechanistic link between Waynes model and the model employed here. More work is required to establish whether the proposed sequence of events does indeed take place during the transition to, and reactivation from, the latent state in vivo. The observed activity of SN or Rpf with non-culturable bacterial populations suggests that one or more of the cognate Mycobacterium tuberculosis proteins could be involved in mechanisms of latency and reactivation of tuberculosis in vivo.

(19K): Fig. 7. A model depicting the transition of actively growing, viable bacteria to dormant forms during prolonged stationary phase. The intensity of shading of the arrows reflects the intensity of metabolic activity. The term non-culturable is operational (Barer et al., 1998 ) and refers, in this figure, to the ability of the bacteria to form colonies.

We thank the Wellcome Trust (grant no. 055840), the UK Biotechnology and Biological Sciences Research Council and The Russian Foundation for Basic Research (grants 00-04-48691 and 01-04-06180mas) for their financial support.