Pretreatment of Urine Samples with SDS Improves Direct Identification of Urinary Tract Pathogens with Matrix-Assisted Laser Desorption Ionization–Time of Flight Mass Spectrometry

Processing of urine samples.We analyzed 71 urine clinical samples with bacterial counts of ≥10⁵ CFU/ml on blood agar after 18 h of incubation in an aerobic atmosphere at 37°C and with a score of <2 in MALDI-TOF MS identification. Samples showing mixed cultures were discarded.

Routine urine samples were initially processed for direct microorganism identification with MALDI-TOF MS, according to previously reported methods (7). An aliquot of each sample was stored at 4°C. When the bacterial count was ≥10⁵ CFU/ml and the MALDI-TOF MS identification score value was <2, the stored aliquot was spread again on blood agar for checking that bacterial count had not changed significantly (count modifications of <5% were considered acceptable) and processed again for MALDI-TOF MS identification, after SDS pretreatment, always before 24 h of storage. Comparison between improvement rates in Gram-positive and Gram-negative bacteria was performed by using the Fisher exact test with the mid-P method. Statistical significance was considered when the P value was <0.01.

Differential procedure. (i) MALDI-TOF MS.Samples were processed as described before (7). Briefly, urine (3 ml) was centrifuged at 2,000 × g for 30 s to remove leukocytes. The supernatant was centrifuged at 15,500 × g for 5 min to collect bacteria. The pellet was washed once with deionized water.

(ii) MALDI-TOF MS with SDS pretreatment.A total of 600 μl SDS 10% was added to 3 ml of urine and vortexed for 2 min. After 5 min of repose on the benchtop, the samples were centrifuged at 15,500 × g for 5 min to collect bacteria. The pellet was washed once with deionized water and transferred to a new tube.

Common procedure.The pellets from samples processed by these two ways were centrifuged at 15,500 × g for 5 min and underwent ethanol-formic acid extraction and MALDI-TOF MS as described previously (7).

No samples showed significant changes of bacterial count between the first spread and the spread of the stored aliquot. Results on whole identification improvement appear in Table 1. Identification reliability increased in 33/71 samples (46.5%). In 22 samples (31%), identification reliability improved by more than one step, from “no reliable identification” (score of <1.7) or “no peaks found” to “identification reliable to the species level” (score of ≥2). Identification level remained unchanged or had changes with no repercussion for identification purposes (from “no peaks found” to “no reliable identification” or vice versa) in 35 samples (49.3%) and decreased in only 3 samples (4.2%).

Improvement was more frequent in urine samples infected with Gram-negative than Gram-positive bacteria. In whole, identification improved in 25/46 (54.3%) samples infected with Gram-negative bacteria and in 8/25 (32%) samples infected with Gram-positive bacteria. Nevertheless, no statistical significance was obtained with the Fisher exact test with mid-P method (P = 0.039). Behavior by species is shown in Table 2.

In a previous study by our group (7), MALDI-TOF MS identified correctly the etiologic microorganism to the species level in 91.8% of urine samples when the bacterial count was >10⁵ CFU/ml, while reliability dropped sharply with lower bacterial counts.

Studies with experimental inocula confirmed this threshold. Minimal bacterial counts ranging between 8 × 10⁴ and 1.5 × 10⁵ CFU/ml were necessary for obtaining good protein profiles. A recent study with a very similar design reported similar bacterial thresholds (8). Inocula required for obtaining MALDI-TOF MS score values of ≥2.0 ranged between 6 × 10⁴ CFU/ml (Proteus mirabilis and Klebsiella pneumoniae) and 1 × 10⁶ (Enterococcus faecalis). Obtaining scores of ≥2.0 for Escherichia coli required bacterial counts of ≥5 × 10⁵. Results on MALDI-TOF MS identification from real urine samples were also around 90%, and the authors also concluded that samples with bacterial counts of <10⁵ CFU/ml are not reliable for MALDI-TOF MS identification.

Sample pretreatment with SDS, a compound that lysates cell membranes, might release microorganisms and increase method sensitivity. Similar methods have been proposed for MALDI-TOF MS direct identification from blood cultures, based on similar theoretical considerations, yielding identification improvement rates of 11% to the genus level and 6% to the species level (9).

The new method allowed a correct identification of 46% of samples that could not be identified with the first described method. Accepting a 92 to 95% level of correct identifications using the standard method, according to our previous results (7), the new method applied on the preliminary identification failures would mean a 2.5 to 4% level of correct identifications, reaching figures around 96 to 97.5%. As happened with the standard method, the results are better for Gram-negative microorganisms. Nevertheless, 32% of urine samples harboring Gram-positive bacteria that had not been identified by the standard method were identified with the SDS pretreatment.

According to our whole results, urine samples reported as presumptively positive by the screening might be directly processed by MALDI-TOF MS, and those for which MALDI-TOF MS does not give a reliable identification are tested again with SDS extraction. This scheme would report the identification of urine tract pathogens soon after receiving the sample in >95% of cases. This might allow for better adjusted empirical treatment in many community-acquired UTIs. Obviously, moderate to severe infections would still require antibiogram in the usual way. Even so, the fast identification of the pathogen may be useful in these patients. The repercussion on the laboratory workflow, the impact in the patient care, and the cost/benefit rate are, in our opinion, the points that need to be studied in the future.

The study did not include any mixed samples. In a previous study (5) of three mixed samples, two did not lead to any reliable identification by MALDI-TOF MS, but one led to E. coli identification. Results obtained with mixed samples probably depend on two factors: the bacterial count of each single population and the proportion between populations. Populations with low bacterial counts would not be detected, as happens in monobacterial infections. Thus, these populations would be ignored, both if they are contaminant and if they are potential pathogens. On the other hand, these populations probably would not interfere with identification of microorganisms with high bacterial count. When we have two or more populations with high bacterial counts, it leads to bizarre protein profiles that would not match with any profile in the database. Nevertheless, newer software versions seem to be able to identify mixed cultures, and thus this problem might be avoided, should both populations attain bacterial counts high enough to be reliably detected and identified.

In conclusion, we have developed a new method, based on SDS pretreatment of urine samples, which allows the identification of urinary tract pathogens directly by MALDI-TOF MS in samples that, while having high bacterial counts, had not been identified by using the standard method. This method allows for direct identification in 46% of the samples with bacterial counts of >10⁵ CFU/ml in which MALDI-TOF MS had not reached a reliable pathogen identification. The improvement is higher for Gram-negative bacteria but affects both Gram-positive and Gram-negative microorganisms. Since >90% of urinary tract pathogens with high bacterial counts are correctly identified by using conventional MALDI-TOF MS as previously described (5), in our opinion this method should be restricted to screening-positive, conventional MALDI-TOF MS procedure-negative samples.