The components of BCSA and BCCSA are shown in Table 1. BCSA (product ref:11317) and supplement (product ref: 11317a) was obtained from Beijing SanYao Science & Technology Co., Ltd, Beijing, China, and prepared in accordance with manufacturer’s instructions. The composition (per liter of distilled water) of BCCSA is as follows: sucrose 2.0 g, sodium pyruvate 7.0 g, trypticase peptone 10.0 g, NaCl 5.0 g, yeast extract 1.5 g, KH₂PO₄ 1.54 g, phenol red 0.02 g, crystal violet 0.001 g, agar 14.0 g, polymyxin B sulphate 600 000 IU, gentamicin 0.01 g, vancomycin 0.0025 g. The phenol red and crystal violet were prepared as 0.2% and 0.01% aqueous solutions, respectively, and 10 ml of each was added per liter. After autoclaving for 15 min at 121°C, antibiotics were added when the medium is cooled to around 50°C. Soybean-Casein Digest Agar (TSA, ref: 236950), Soybean-Casein Digest Broth (TSB, ref: 211825) and Sabouraud Dextrose Broth (SDB, ref: 238230) was obtained from Becton-Dickinson, Rungis, France.

Twenty strains of Bcc and eleven strains of non-Bcc were purchased separately from China Center of Industrial Culture Collection (CICC), China National Center for Medical Culture Collections (CMCC), China General Microbiological Culture Collection Center (CGMCC) and American Type Culture Collection (ATCC). A further 40 strains of Bcc and 157 strains of non-Bcc were retrieved from water-based pharmaceutical, cosmetic products and pharmaceutical water systems. In total, 60 strains of Bcc were tested including: Burkholderia cepacia (n = 24), Burkholderia aenigmatica (n = 13), Burkholderia cenocepacia (n = 11), Burkholderia contaminans (n = 4), Burkholderia vietnamiensis (n = 3), Burkholderia stabilis (n = 2), Burkholderia multivorans (n = 1), Burkholderia pyrrocinia (n = 1), Burkholderia ambifaria (n = 1).

Bacterial strains used in this study were identified through the analysis of the 16S rRNA gene. The 16S rRNA gene was amplified using the primers 27f (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492r (5′-TACGGCTACCTTGTTACGACTT-3′). The PCR reaction mixture (50 μL) contained: Taq DNA polymerase (5 U/μL) 0.25 μL, upstream and downstream primers (10 μmol/L) 1 μL each, 10 × PCR buffer 5 μL, dNTP mixture (each 2.5 mmol/L) 4 μL, DNA template 1 μL, and ddH₂O to adjust the volume to 50 μL. The PCR reaction conditions were as follows: initial denaturation at 95°C for 4 min; 35 cycles of denaturation at 95°C for 30 s, annealing at 50°C for 45 s, and extension at 72°C for 1 min and 30 s; final extension at 72°C for 5 min. Bcc strains were additionally identified by recA gene sequence analysis. The recA gene was amplified using the primers BCR1 (5′-TGACCGCCGAGAAGAGCAA-3′) and BCR2 (5′-CTCTTCTTCGTCCATCGCCTC-3′) (Mahenthiralingam et al., 2000). The reaction mixture was identical to that used for the 16S rRNA gene amplification. The PCR reaction conditions were as follows: initial denaturation at 94°C for 3 min; 32 cycles of denaturation at 94°C for 30 s, annealing at 56°C for 45 s, and extension at 72°C for 1 min; final extension at 72°C for 10 min. Fungi used in this study were identified through colony observation, microscopic examination, and Internal Transcribed Spacer (ITS) sequence analysis. The ITS gene was amplified using the primers ITS F (5′-GGAAGTAAAAGTCGTAACAAGG-3′) and ITS R(5′- TCCTCCGCTTATTGATATGC-3′). The reaction mixture was identical to that used for the 16S rRNA gene amplification. The PCR reaction conditions were as follows: initial denaturation at 94°C for 10 min; 29 cycles of denaturation at 94°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 30 s; final extension at 72°C for 5 min. The amplified products were sequenced using the Sanger sequencing method, performed by Beijing Novogene Bioinformatics Technology Co., Ltd. The sequences were compared and analyzed using BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi). In addition to sequence analysis methods, Bcc strains were identified by the Omnilog system on the basis of utilization of carbon sources, which is described in section 2.3.

A 96-well automated Omnilog system with GENIII Micro plates (Biolog Inc., Hayward, CA, USA) was used for the characterisation of carbon source utilization. Microplates were set up and analyzed following the manufacturer’s instructions. Single colonies of 60 strains of Bcc were subcultured on TSA plates separately. After 24 h incubation at 33°C, colonies were collected with a cotton swab and suspended in 18 mL in oculation fluid A (IF-A, Catalog No.72401) to obtain a homogeneous suspension. The turbidity of the bacterium suspensions was set to 90 to 98% using the Biolog Turbidimeter. GENIII MicroPlates were inoculated with 100 μL bacterium suspension per well and incubated at 33°C. The OmniLog system includes an incubator and a charged-coupled device camera system. The OmniLog digital camera measures the color level of each well in OmniLog Units (OU). Tetrazolium redox dyes indicate the extent of microbial utilization of carbon sources and sensitivity to chemicals through changes in color intensity. A positive result is indicated by the reaction wells turning purple, while a negative result is indicated by the reaction wells remaining colorless. Metabolic activity was determined by the system at 15-min intervals automatically during a whole 36 h incubation. The metabolic results were shown by Omnilog software (Biolog Inc., USA, Version 4.20.05). By comparing the carbon source utilization profile of the tested strain with those in the database, Omnilog system provided an identification result at the species level.

Quantitative recovery of Bcc on selective media was assessed with four strains: ATCC 25416 = CICC10857 contained in USP (general chapter <60>), CMCC(B)23005, CMCC(B)23006, CMCC(B)23010 contained in the ChP (General chapter 1109). They were subcultured in TSB at 33°C for 18 to 24 h, then performed by adding 100 μL of suspension to 0.9 mL of 0.9% NaCl solution for serial tenfold dilutions. For each strain, an inoculum of approx. 10³ CFU/mL was intended to be obtained. A 100 μL aliquot of each suspension was spread on the two selective agar plates, as well as TSA. For each suspension, two petri dishes were used. Colonies were counted after incubation at 33°C for 48 to 72 h. Each strain was subjected to 10 experimental replicates. TSA is a non-selective medium and served as a control. The ratio of recovery (%) was determined by colony-forming unit (CFU) counted on selective media to that on TSA for each strain.

To further expand the scope of investigation, the recovery assay was also conducted using the additional sixteen standard strains employed in this study. Strains were inoculated into TSB individually and incubated for approximately 24 h to achieve an initial bacterial concentration of around 10⁸ CFU/mL. Subsequently, serial 10-fold dilutions of the cultures were performed using 0.9% NaCl solution, with the original culture defined as dilution level 0. From each dilution series, 100 μL of the 10⁻⁵ and 10⁻⁶ dilutions was spread onto TSA, BCSA and BCCSA plates using a spreader, with two replicate plates for each dilution. The plates were then incubated at 33°C for 48 h, after which the number of CFU was recorded. The recovery assay for each strain was repeated three times.

A total of 60 Bcc strains were subcultured in TSB at 33°C for 18 to 24 h. An inoculum with an approximate concentration of 10⁶ to 10⁸ CFU/mL was preparared from these cultures. About 10 μL of each dilution was performed three-zone streaking inoculation on different selective media using a disposable inculating loop. The inoculated plates were incubated at 33°C for 24 and 48 h. The plates were examined at both time points to assess bacterial growth. After of incubation. The number of strains that grew in the main inoculum area, the second quadrant and the third quadrant were recorded at each observation time point. This approach allows for a detailed assessment of the distribution and growth dynamics of the Bcc strains across different zones of the streaked plates.

Fifteen strains of Bcc were initially cultured in TSB at 33°C for 24 h to obtain a uniform and active culture. An inoculum with an approximate concentration of 10⁶ CFU/mL was prepared by appropriate dilution of the overnight culture. BCSA and BCCSA were prepared as liquid media according to their respective formulations, excluding agar, phenol red, and crystal violet. These components were omitted to facilitate optical density measurements without interference from solidifying agents or pH indicators. For the automated growth curve analysis, 20 μL of the prepared inoculum was mixed with 180 μL of the liquid culture medium (either BCSA or BCCSA) and transferred into the wells of a Bioscreen plate (Labsystems, Helsinki, Finland). Each condition was tested in quadruplicate to ensure statistical robustness. Control wells contained only the culture medium without inoculum. The Bioscreen C system (Labsystems, Helsinki, Finland) was used for automated growth curve analysis. The instrument was preheated to 33°C to ensure consistent temperature conditions throughout the experiment. The absorbance of the cell suspensions at 600 nm was measured every 10 min over a period of 72 h. Prior to each measurement, the cultures were automatically agitated for 15 s to ensure uniform distribution of cells.

The absorbance data obtained from the Bioscreen C system were exported to Microsoft Excel 2010 for further processing. The average absorbance values for each type of culture medium were calculated from the quadruplicate measurements. Growth curves were generated by plotting the average OD600 values against incubation time (in hours). To eliminate the influence of medium color on the absorbance readings, the OD600 values of the experimental wells were corrected by subtracting the corresponding values from the control wells. The maximum specific growth rate (μ_max) for each strain in different media was determined by fitting the logarithmic phase of the growth curves, ensuring that the R² value was ≥0.99.

Non-Bcc bacterial strains (n = 156) were subcultured in TSB at 33°C for 18 to 24 h. Yeasts (n = 10) were subcultured in SDB at 23°C for 24 to 48 h. The spore suspensions of filamentous fungi (n = 2) were prepared and stored by our laboratory. An inoculum with an approximate concentration of 10⁶ to 10⁸ CFU/mL was preparared from bacterial cultures. An inoculum with an approximate concentration of 10⁵ CFU/mL was preparared from fungal cultures. About 10 μL of each dilution was performed three-zone streaking inoculation on different selective media using a disposable inculating loop. The inoculated plates were incubated at 33°C for 72 h. A strain was recorded as “growing” by colonies present in the main inoculum area.

Difference in performance of recovery for each Bcc strain of the two media was investigated for statistical significance using Students’ test. Difference in performance of growth curves of the two media was investigated using McNemar’s test with the continuity correction applied. Statistical significance was assigned to a probability (P) value of < 0.05. For the sixteen other standard strains, a generalized linear mixed model (GLMM) with a negative binomial distribution (nbinom2 parameterization) was fitted to assess the effects of media type on bacterial colony counts, while accounting for between-strain variability. The model was specified as:

Where: fixed effect: media (categorical: BCSA, BCCSA, TSA as control); random intercept: strain (16 levels) to model subject-specific variability; distribution: negative binomial (log link) to handle overdispersion (variance > mean). Model fitting was performed using the glmmTMB package (v1.1.7) in R (v4.5.1). Significance of fixed effects was evaluated via Wald z-tests with α = 0.05. Residual diagnostics of the model were conducted using the DHARMa package (v0.4.7).

Table 2 provides an overview of the carbon source utilization patterns of Bcc strains on the Omnilog GenIII reaction plate. The results indicated that several carbon sources, including D-Maltose, D-Cellobiose, D-Pine Disaccharide, α-D-Lactose, L-Rhamnose, Inosine and D-Lactic Methyl Ester were not utilized by any of the tested Bcc strains. In contrast, carbon sources such as α-D-Glucose, D-Galactose, L-Fucose and Glucuronide were utilized by the strains, as evidenced by positive reactions. Notably, B. ambifaria (n = 1), which differed from other Bcc strains, was unable to utilize certain carbon sources including D-Mannose, D-Fructose, D-glucose-6-Phosphate, D-Fructose-6-phosphate, D-Gluconic Acid and L-Malic Acid. This finding was consistent with the description in Berger’s Manual of Systematic Bacteriology,which states that Bcc strains are unable to utilize α-D-Lactose (Garrity et al., 2005). This inability to utilize α-D-Lactose was a crucial factor in the formulation adjustment of BCSA. Based on the carbon utilization characteristics of Bcc, lactose in BCSA was replaced with sodium pyruvate, and the content of sucrose was adjusted to develop a new selective culture medium BCCSA.

The content of phenol red in BCCSA is 0.02 g/L, and the content of crystal violet is 0.001 g/L. Given that the pH indication of phenol red ranges from 6.8 to 8.4, the initial pH value of BCCSA was adjusted to approximately 6.2 using dipotassium hydrogen phosphate as the pH regulator, resulting in a yellow coloration of the agar. During growth, peptone in BCCSA is utilized by Bcc as a nitrogen source, producing ammonia via deamination or urease activity. The hydrolysis of ammonia to form hydroxide ions leads to an increase in the pH of the medium, which in turn induces a noticeable shift in the medium’s color from yellow to red. This color change facilitates convenient and clear observation by experimenters.

Table 3 showed the recovery rate results of the four standard strains commonly employed as reference strains in pharmaceutical quality control tests, on the two selective media. For each strain, the recovery rate was calculated as the ratio of CFU on each selective medium to those on TSA. Each strain was tested in 10 replicates. For both media, recovery rates of 4 strains satisfied a factor ranging from 0.5 to 2. The recovery rate on BCCSA was significantly higher compared to both BCSA of CICC10857 = ATCC25416 (p < 0.01, Student’s t test) and CMCC(B)23010 (p < 0.01). The differences in recovery rate between the two selective media were not found to be statistically significant for an additional two strains. Figure 1A showed the appearance and indication reactions of the four strains on the two media at 72 h of incubation. Delayed growth on BCSA was manifested in smaller colonies than BCCSA.

BCSA and BCCSA were evaluated using 60 strains, which included 20 strains from international collection and 40 strains isolated from pharmaceutical and cosmetic products. Recovery of Bcc strains on the two agar is illustrated in Figure 1B. After 24 h incubation at 33°C, all 60 strains were detected growth on both selective media. On BCCSA, 49 strains grew into the third quadrant of the inoculum area, 8 strains into the second quadrant, and 3 strains remained in the first quadrant. On BCSA, the distribution was 40 strains in the third quadrant, 15 in the second quadrant, and 5 in the first quadrant. After 48 hous of incubation, one strain isolated from oral liquid (YP20200287) still exhibited growth in the first quadrant with several colonies on BCSA. The number of strains that grew into the third quadrant on BCCSA and BCSA were 58 and 54, respectively.

Figure 2B showed the original colony counts distribution of sixteen Bcc strains on three media. The overall effect of media was not statistically significant [χ²(2) = 5.43, p = 0.066]. However, post-hoc tests suggested a marginal reduction in colony counts for BCSA compared to the TSA (β = −0.203, SE = 0.093, z = −2.19, p = 0.028). BCSA showed a 18% reduction compared to TSA, with BCCSA did not show a significant difference to TSA (β = −0.039, SE = 0.092, z = −0.42, p = 0.675). The forest plot was shown as Figure 2A. Between-strain variability was modest (σ² = 0.098, SD = 0.314), indicating that 7.2% of variance (ICC) was attributable to strain differences after controlling for media effects. The model diagnostics were conducted using DHARMa’s simulated residuals approach, which confirmed the adequacy of the negative binomial GLMM specification. Residual analysis showed no significant deviations from model assumptions, with the uniformity test (KS-test: p = 0.84) indicating proper distributional fit, the dispersion test (p = 0.87) confirming appropriate handling of overdispersion, and no evidence of problematic outliers (p = 0.38). Visual inspection of the Q-Q plot and residual-versus-fitted plot further supported these conclusions, as all diagnostic plots displayed the expected patterns without systematic deviations. The non-significant p-values (all > 0.05) collectively suggest that the model’s distributional assumptions were met and the random effects structure adequately accounted for between-strain variability. These results justify the use of inferential statistics from this model for drawing conclusions about media treatment effects.

To accurately depict the dynamic growth of Bcc strains in both media, the automated growth curve determination method was employed to assess the growth curves of 15 Bcc strains in liquid selective media in this study. The inclusion of agar was omitted during the preparation process. Additionally, phenol red and crystal violet were excluded from the formulations due to their potential interference with OD value readings caused by changes in medium pH resulting from strain inoculation. Figure 1C showed growth curves of four of the 15 Bcc strains in liquid system of BCSA and BCCSA. The majority of microbiologists consider the “maximum specific growth rate” (μ_max) as the measure of the growth intensity of an organism. Mathematically, μ_max is defined as the slope of the (sigmoid) growth curve at its inflection point (Peleg and Corradini, 2011). Table 4 gives the μ_max of 15 Bcc strains in the liquid systems of BCCSA and BCSA. The μ_max in BCCSA was greater than that in BCSA for all strains except CGMCC23882 (B. contaminans) and CGMCC1.10511 (B. ambifaria). The differences were significant according to the McNemar’s test (p = 0.02).

Table 5 summarizes the growth performance of 168 non-Bcc strains evaluated for selectivity on BCSA and BCCSA. For the 168 strains tested, the two media had identical inhibition ability to Gram-positive bacteria and fungi, with one strain of Bacillus altitudinis, one strain of Candida theae, and one strain of Wickerhamomyces anomalus grown. 4 strains of Ralstonia pickettii and 2 strains of Achromobacter xylosoxidans grow well on both media. One strain of Stenotrophomonas maltophilia was detected weak growth with several colonies only on first quadrant of the inoculum area of both media. Difference mainly occurred in the media selectivity for strains from pharmaceutical water system, which accounted for 69% of the total number of strains tested. A single strain representing each of the taxonomic category of Burkholderia gladioli, Phyllobacterium myrsinacearum, Sphingomonas leidyi, Pseudochrobactrum, and 2 strains of Sphingomonas paucimobilis grow on BCCSA, while not on BCSA. Although there was no statistically significant difference in their ability to inhibit non-Bcc strains (p = 0.05), the nutrient formula of BCCSA seems to be more suitable for the growth of objectionable microorganisms from water system, which is more favorable for the pre-screening of contamination risks in water-based products.