Sampling was carried out during the February-March 2025 tomato cropping season, and all laboratory and screen-house experiment was done at Tanzania Agricultural Research Institute, Tengeru in Arusha, Tanzania. Samples were collected from three tomato-producing wards of Arusha District Council (Table 1), an area with bimodal rainfall and loamy to sandy clay soils under year-round irrigation.

Three fields were selected purposively per ward, giving nine; a field was eligible if tomato was grown intensively, if the farmer reported wilt-related losses, and if symptomatic and asymptomatic plants were both present.

Within each field, three composite rhizosphere soil samples were taken, giving 27 composites, each pooling the rhizosphere soil of several asymptomatic plants along a diagonal transect; plants were classed as asymptomatic if free of wilting, leaf yellowing, stunting and stem discoloration. Plants were uprooted carefully, loosely adhering soil shaken off, and soil remaining firmly attached to the roots taken as rhizosphere soil [21]. One symptomatic plant was also collected from one field in each ward. The 30 samples were bagged individually, transported in a cool box and processed on arrival.

The field, not the composite sample, is the independent sampling unit: the three composites within a field are subsamples, so the 27 composites represent nine fields rather than 27 independent sites. They were processed separately to maximize recovery of distinct colony types, not as replicates, and no inference about variation among fields or wards is drawn. Isolate counts per ward (Table 1) are descriptive.

Casamino acid-peptone-glucose (CPG) medium was used as the basal medium, and triphenyl tetrazolium chloride (TZC) medium was prepared from the same base by amending it after autoclaving with filter-sterilized triphenyl tetrazolium chloride to 0.005% (w/v), that is 50 mg·l⁻¹ [22].

Symptomatic plants were subjected to the stem-streaming test, in which a slimy bacterial exudate from cut stem sections is characteristic of bacterial wilt [3]; vascular tissue was then surface-sterilized, macerated and dilution-streaked onto CPG and TZC plates. For soil, 1 g of each composite was suspended in 9 ml of sterile distilled water and a tenfold dilution series prepared to 10⁻⁶ [21]. Aliquots of 0.1 ml from the 10⁻³ and 10⁻⁵ dilutions were spread on TZC plates, three per dilution per sample, and incubated at 28˚C for 48 - 72 h; lower dilutions gave confluent growth and 10⁻⁶ too few colonies. Colonies were purified by three successive single-colony transfers, and colony form, margin, elevation, texture, colour and fluidity recorded on three TZC plates per isolate after 48 h at 28˚C. Isolates were coded by ward (BWN, MLN, SMS), then A for the non-fluidal soil phenotype or P for the fluidal phenotype from plants, and a serial number.

Colony morphology on tetrazolium medium supports only preliminary phenotypic grouping; it does not establish species identity or phylotype within the species complex, which requires sequence-based molecular identification, nor virulence status, which is assessed separately by plant pathogenicity assays under specified conditions. The groups are therefore designated presumptive virulent and presumptive avirulent throughout, and the isolates described as Ralstonia-like rather than assigned to a species.

Inoculum was prepared from 24 h nutrient agar cultures and adjusted to an optical density of 0.1 at 600 nm, approximately 10⁸ colony-forming units (CFU)·ml⁻¹. Seedlings of the tomato varieties Tanya and Tengeru 97 were raised in sterilized soil to the four-true-leaf stage at three weeks, root-tip wounded, transplanted singly into pots of sterilized soil and inoculated by drenching the root zone with 30 ml of suspension. Plants were held in a screen house at 28˚C and irrigated at two-day intervals.

The experiment comprised 14 treatments: the 12 presumptive avirulent isolates, a presumptive virulent isolate BWN P1 as positive control and sterile distilled water as negative control, each applied to both varieties, giving 28 treatments by variety combinations. It was a randomized complete block design with four blocks, the blocking factor being position along the screen-house bench, which accounts for the gradient in light and temperature. Each experimental unit was one pot with one seedling, giving four plants per combination and 112 plants in total.

Severity was assessed at 7, 14, 21, 28 and 35 days after inoculation on a 0 - 5 scale, where 0 = no symptoms; 1 = 1% - 25% of leaves wilted; 2 = 26% - 49%; 3 = 50% - 74%; 4 = 75% - 100%; and 5 = plant dead [23]. Final severity at 35 days was analyzed. Plants were then uprooted, stems sectioned longitudinally to record vascular discoloration, and re-isolation carried out on TZC medium from stem tissue and pot soil: for the positive control, to satisfy Koch’s postulates, and for the symptomless plants inoculated with the candidate isolates, to determine whether the inoculated isolates had persisted in the plant and soil medium.

Antagonism was assessed by the agar well diffusion method, with modifications [24]; this widely used screening method measures inhibition mediated by diffusible compounds and is not a dual culture assay [25]. A lawn of the presumptive virulent test isolate BWN P1 was established by spreading 0.1 ml of a suspension of approximately 10⁸ CFU·ml⁻¹ over a CPG plate, and four wells of 6 mm diameter were punched at equidistant positions.

The assay comprised nine treatments: seven presumptive avirulent isolates representing all three wards BWN A1, BWN A5, MLN A1, MLN A2, SMS A1, SMS A2 and SMS A3were selected because they showed rapid colonization during re-isolation in pathogenicity test; the five isolates not tested were BWN A2, BWN A3, BWN A4, MLN A3 and SMS A4), ampicillin-cloxacillin as positive control, and sterile distilled water as negative control. Each bacterial treatment was applied as 20 µl per well of a suspension of approximately 10⁸ CFU·ml⁻¹, matching the lawn inoculum; the antibiotic control was applied as 20 µl per well of a filter-sterilized 1 mg·ml⁻¹ aqueous stock, delivering 20 µg per well. Because living suspensions and a fixed antibiotic dose are not equivalent quantities, no equivalence of concentration between them is implied.

The four wells on a plate received the same treatment and were technical replicates, averaged to give one value per plate; the plate was the experimental unit and three replicate plates per treatment were the experimental replicates, giving n = 3 per treatment and 27 plate values. Individual wells were never treated as replicates. Plates were incubated at 28˚C for 48 h in a completely randomized design, and the total diameter of the clear zone, including the 6 mm well, was measured to the nearest 1 mm. All diameters reported follow this convention.

Analyses were done in R version 4.3.3 with significance at p < 0.05, and descriptive results are given as mean ± SD. Severity at 35 days could not be analyzed by two-way ANOVA because it was zero in 13 of the 14 treatments on both varieties, leaving no within-group variance and completely separated groups, so severity is reported descriptively only. Inference was based on wilt incidence (severity > 0): for each variety, the positive control was compared with the pooled 13 remaining treatments by Fisher’s exact test, and positive control severity was compared between varieties by the exact Wilcoxon rank-sum test. Treatment-by-treatment comparisons were not reported because, with four plants per combination, the smallest possible p-value (0.029) would not survive correction for 13 comparisons. Inhibition zone diameters (n = 3, nine treatments) were analyzed by one-way ANOVA and confirmed with the Kruskal-Wallis test, which supports only an overall treatment difference. Because variances were unequal, Tukey’s test was replaced by Welch’s t-test with Holm adjustment, comparing the two active isolates with the antibiotic control and with each other, and excluding the six treatments with no inhibition zone.

Twenty-seven bacterial isolates were purified from the soil samples: 11 from Bwawani, seven from Mlangarini and nine from Sambasha (Table 1). Twelve had small, dry, circular, uniformly deep-red, non-fluidal colonies on TZC medium: five from Bwawani (BWN A1 - A5), three from Mlangarini (MLN A1 - A3) and four from Sambasha (SMS A1 - A4). Of the remaining 15, seven had large, fluidal, irregular, creamy-white colonies with a pink centre and eight had characteristics consistent with Bacillus spp. All three symptomatic plants gave a milky-white ooze in the stem-streaming test, and isolation from their vascular tissue yielded three fluidal isolates, coded BWN P1, MLN P1 and SMS P1.

Table 1. Location, number of fields sampled and recent cropping history of the three wards of Arusha District Council sampled during the February-March 2025 tomato cropping season, with the numbers of samples collected and isolates recovered per ward.

Table 2. Colony characteristics of the two phenotypic groups on triphenyl tetrazolium chloride medium after 48 h at 28˚C.

Two colony phenotypes were recognized on TZC medium after 48 h at 28˚C and differed in form, margin, elevation, texture, colour and fluidity (Table 2, Figure 1). The fluidal phenotype, designated presumptive virulent, comprised the three isolates from symptomatic plants; the non-fluidal phenotype, designated presumptive avirulent, comprised the 12 soil isolates. Characteristics were consistent across the three replicate plates per isolate.

(a) (b)

Figure 1. Colony morphology of the two phenotypic groups on triphenyl tetrazolium chloride medium after incubation at 28˚C for 48 h. (a) Fluidal, irregular, creamy-white colonies with a pink centre, characteristic of the presumptive virulent phenotype recovered from symptomatic vascular tissue; (b) small, circular, dry, uniformly deep-red, non-fluidal colonies, characteristic of the presumptive avirulent phenotype recovered from rhizosphere soil.

None of the 12 presumptive avirulent isolates produced wilt symptoms on either variety at any assessment up to 35 days: mean severity was 0.00 ± 0.00 and wilt incidence 0 of 4 plants for every isolate on both varieties (Table 3). Plants treated with sterile distilled water likewise remained symptomless. The presumptive virulent control caused severe wilt on both varieties, all four plants wilting on each, with mean severity 4.50 ± 0.58 on Tanya and 3.50 ± 0.58 on Tengeru 97.

Table 3. Mean disease severity and wilt incidence on the tomato varieties Tanya and Tengeru 97 at 35 days after inoculation.

Wilt incidence in the positive control (4 of 4) differed from the pooled incidence of the 13 remaining treatments (0 of 52) on both varieties (Fisher’s exact test, p = 2.7 × 10⁻⁶). The experiment therefore demonstrates complete separation between the positive control and the candidate isolates under the conditions tested, and no measurable variation among the candidates themselves. Because every candidate and the negative control scored zero at every assessment, the data contain no information with which to rank the candidates, and no such comparison is attempted.

Mean severity in the positive control was numerically higher on Tanya than on Tengeru 97, but with four plants per variety the difference was not statistically supported (exact Wilcoxon rank-sum test, W = 14, p = 0.114), so no conclusion is drawn about relative susceptibility.

Plants inoculated with the presumptive virulent isolate developed vascular browning and their stem sections released a milky exudate (Figure 2). Re-isolation from these plants and their pot substrate yielded colonies morphologically identical to the inoculated isolate, satisfying Koch’s postulates; no Ralstonia-like colonies were re-isolated from the negative control.

Re-isolation from the symptomless plants and pot soil inoculated with the 12 candidate isolates were recovered from Tengeru 97 with codes, BWN A1, BWN A5, MLN A1, MLN A2, SMS A1, SMS A2 and SMS A3 had 3/4 plants. Recovery of the inoculated isolate indicates that the absence of wilt reflected an inability to cause disease after establishment, whereas failure to recover it leaves open that the isolate did not persist, in which case absence of disease cannot be attributed to avirulence.

(a) (b)

Figure 2. Response of tomato plants 35 days after root-drench inoculation with 30 ml of a suspension of approximately 108 colony-forming units·ml⁻¹. (a) Plants inoculated with a presumptive avirulent isolate, showing no wilt symptoms; (b) plants inoculated with the presumptive virulent isolate, showing wilting and collapse.

Seven of the 12 presumptive avirulent isolates, representing all three wards, were tested. Two produced clear inhibition zones: BWN A1, 16.4 ± 0.5 mm, and BWN A5, 16.2 ± 0.4 mm (Table 4, Figure 3). The ampicillin-cloxacillin control produced 15.4 ± 0.3 mm.

Table 4. Mean total diameter of inhibition zones, including the 6 mm well, produced by seven presumptive avirulent isolates and two controls against the presumptive virulent test isolate in an agar well diffusion assay on casamino acid-peptone-glucose medium after 48 h at 28˚C. Means sharing the superscript letter a do not differ significantly (Welch’s t-test with Holm adjustment, p > 0.05); no zone was detected in the remaining treatments, which had zero variance and were not entered in pairwise tests.

Figure 3. Agar well diffusion assay of isolate BWN A1 and BWN A5 against the presumptive virulent test isolate on casamino acid-peptone-glucose medium after 48 h at 28˚C, showing a clear inhibition zone. Wells were 6 mm in diameter and received 20 µl of a suspension of approximately 10⁸ colony-forming units·ml⁻¹. Diameters in Table 4 are total and include the well.

All values are total diameters and include the 6 mm well. No zone was recorded for MLN A1, MLN A2, SMS A1, SMS A2, SMS A3 or the negative control. Treatments differed significantly (F = 3447; df = 8, 18; p < 0.001), confirmed by the tie-corrected Kruskal-Wallis test (H = 25.9; df = 8; p = 0.001). These tests show an overall difference among treatments but do not identify which treatments differ. In pairwise comparisons by Welch’s t-test with Holm adjustment, neither BWN A1 (difference 1.0 mm; [adjusted p = 0.16]) nor BWN A5 (difference 0.8 mm; [adjusted p = 0.16]) differed significantly from the antibiotic control, and the two isolates did not differ from each other ([adjusted p = 0.62]) (Table 4). The numerically larger zones of the two isolates therefore cannot be distinguished from the antibiotic control with three plates per treatment.