Assessment of Morphology for Rapid Presumptive Identification of Mycobacterium tuberculosis and Mycobacterium kansasii

Abstract

Mycobacterium tuberculosis often exhibits serpentine cording when grown in liquid medium, whereas Mycobacterium kansasii can be larger and cross-barred. We assessed the use of these morphologic characteristics as a cost-effective method for rapid presumptive identification of isolates from BACTEC bottles. Without specific training, using the Kinyoun acid-fast stain, definitive cording was found in 237 of 373 specimens positive for M. tuberculosis (64%) and cross-barring was recognized within 63 of 76 (83%) of the specimens positive for M. kansasii, giving sensitivities specificities, positive predictive values, and negative predictive values of 63.5, 96, 92, and 79%, respectively, for M. tuberculosis and 83, 95, 59, and 98%, respectively, for M. kansasii. With training and experience, these results improved to 74.5, 98, 96, and 84% and 93, 98, 79, and 98%, respectively. The major improvements were in distinguishing the pseudocording, or loose aggregation of Mycobacterium avium complex from M. tuberculosis and the long beaded forms of Mycobacterium gordonae from M. kansasii. Mycobacterium asiaticum and Mycobacterium szulgai, which rarely occur, are genetically related to M. kansasii and morphologically difficult to distinguish. In defined circumstances, serpentine cording and cross-barring can be used for rapid presumptive identification of M. tuberculosis and M. kansasii, respectively, and as guides for initial probe selection to reduce costs.

The rapid, specific, and inexpensive detection of Mycobacterium tuberculosis and Mycobacterium species other than M. tuberculosis is an important concern, particularly in developing countries with a high incidence of mycobacterial disease and few economic resources. Recently, the development of rapid and specific new methods, such as DNA probes, high-performance liquid chromatography analysis of mycolic acids, and PCR, has aided in detection, but they may not be cost-effective or available to developing countries.

The distinctive morphology of some mycobacteria when grown in liquid culture has been previously noted. Trehalose 6,6′-dimycolate, called cord factor, is a glycolipid present in the cell wall of the mycobacteria that contributes to the virulence of M. tuberculosis and promotes growth as tight, rope-like aggregates of acid-fast bacilli (AFB) in which the long axes of the bacteria parallel the long axes of the cord (8). Others have shown that the presence or absence of this serpentine cording can be used as a cost-effective method for rapid and presumptive identification of M. tuberculosis and as a guide for the initial probe selection (2, 3, 4, 6, 8). The morphology of Mycobacterium kansasii has been described as broad rods that exhibit marked cross-barring which we call ladders. Others authors call this kind of morphology striped bacilli or candy canes (2, 7). This characteristic is not specific but with experience permits one to distinguish M. kansasii from the most common clinical mycobacteria. Although neither of these characteristics is unique to a species, we hypothesized that with experience, the most common clinical mycobacteria could be distinguished.

Our objectives were to evaluate presence of serpentine cording and ladders from Kinyoun smears prepared from positive BACTEC cultures for the rapid and presumptive diagnosis of M. tuberculosis and M. kansasii, respectively. We also assessed the degree of improvement after specific training.

MATERIALS AND METHODS

A total of 12,297 specimens were processed from 5,234 patients at the Veterans Affairs Medical Center, Houston, Tex., were processed. Positive clinical specimens included 852 respiratory secretions (sputum and bronchoscopy samples), 45 blood samples, 31 tissue samples, 24 fluid samples, 21 stool samples, 11 bone marrow samples, and 14 urine samples. In our routine procedure, contaminated specimens were digested and decontaminated with N-acetyl-l-cysteine and 2% NaOH for 15 min, the reaction was stopped by the addition of phosphate buffer, and specimens were centrifuged at 3,000 × g for 15 min. Tissues and normally sterile body fluids were inoculated directly into the media. Each specimen was inoculated into one BACTEC 12B bottle (Becton Dickinson, Sparks, Md.), one Lowenstein-Jensen agar slant, and one Middlebrook 7H11 agar plate. Blood samples were inoculated into BACTEC 13A medium. All BACTEC bottles, incubated at 35°C, were read using a BACTEC 460 instrument twice a week and monitored for 6 weeks. Slants and plates were incubated at 35°C in 5 to 10% CO₂ and observed weekly for 6 weeks. Positive cultures were microscopically examined by Kinyoun acid-fast staining without previous vortexing. The final identification was made by color and morphology of the colonies, biochemical tests, DNA-rRNA hybridization assay (Gen Probe Inc., San Diego, Calif.), and other identification methods such as cellular fatty acid analysis (Microbial Identification System; MIDI, Inc., Newark, Del.).

Routine technologists made the initial microscopic evaluation of 899 slides to differentiate the presence or absence of serpentine cording (tight, rope-like aggregates of AFB in which the long axes of the bacteria parallel the long axes of the cord) (Fig. 1) and ladders (broad rods that exhibit marked cross-barring) (Fig. 2). After training, 110 slides were subsequently reviewed (the others were not available) without knowledge of the previous microscopic or culture results to determine if the recognition of serpentine cording and cross-barring improved with training and experience. A third category, which we called pseudocording (loose aggregates of mycobacteria) (Fig. 3), was recognized.

FIG. 1.

Microscopic morphology of M. tuberculosis grown in BACTEC broth. The panels show M. tuberculosis exhibiting different serpentine cording.

FIG. 2.

Microscopic morphology of M. kansasii, exhibiting cross-barring (ladders).

FIG. 3.

Microscopic morphology of M. avium, exhibiting loose aggregates; usually they are found singly or in small aggregates.

Sensitivity (SE), specificity (SP), positive predictive value (PPV), and negative predictive value (NPV) were calculated before and after training.

RESULTS

From 1995 through 1998, 12,297 specimens from 5,234 patients were processed; 998 (8%) from 512 patients were positive for AFB. A total of 899 (91%) of the mycobacterial isolates grew from BACTEC bottles, and these were evaluated for the presence or absence of serpentine cording and ladders. The overall distribution of the mycobacteria is shown in Table 1. The percentage of M. kansasii, at about 10%, is higher in Texas than in some other geographical areas (1), and thus the ability to make a presumptive identification is more important. Table 2 shows the percentages of specimens that were smear positive and that grew in the BACTEC bottles only. Those that eventually grew M. tuberculosis were more frequently smear positive (57%). In addition, 64% (231 of 359) of the smear positive specimens were M. tuberculosis.

TABLE 1.

Distribution of mycobacteria by year

Species of mycobacteria	No. (%) of isolates in:
	1995 (n = 246)	1996 (n = 322)	1997 (n = 251)	1998 (n = 179)	Total (n = 998)
M. tuberculosis	107 (43)	141 (44)	98 (39)	62 (35)	408 (41)
M. avium	72 (29)	90 (28)	86 (34)	74 (41)	322 (32)
M. kansasii	31 (13)	33 (10)	16 (6)	14 (8)	94 (9.5)
M. fortuitum- M. chelonae	9 (4)	17 (5.5)	10 (4)	9 (5)	45 (4.5)
M. xenopi	6 (2)	4 (1)	2 (1)	0 (0)	12 (1)
Mycobacterium speciesa	5 (2)	4 (1)	2 (1)	4 (2)	15 (1.5)
M. gordonae	14 (6)	20 (6)	29 (12)	16 (9)	79 (8)
Othersb	2 (1)	13 (4)	8 (3)	0 (0)	23 (2.5)

^aProbe negative and not further identified. 

^b1995, one M. asiaticum isolate and one M. marinum isolate; 1996, seven M. szulgai isolates, three M. lentiflavum isolates, two M. scrofulaceum isolates, and one M. smegmatis isolate; 1997, two M. asiaticum isolates, two M. lentiflavum isolates, two M. szulgai isolates, one M. mucogenicum isolate, and one M. scrofulaceum isolate. 

TABLE 2.

Characteristic of cultures, 1995 to 1998

Characteristic	No. (%) of isolates (n = 998) of:
	M. tuberculosis(n = 408)	M. avium(n = 322)	M. kansasii(n = 94)	M. fortuitum- M. chelonae(n = 45)	M. xenopi(n = 12)	Mycobacterium speciesb(n = 15)	M. gordonae(n = 79)	Others (n = 23)
Smear positivea	231 (57)	77 (24)	38 (40)	5 (11)	0 (0)	2 (13)	0 (0)	6c (26)
BACTEC positive	373 (91)	293 (91)	76 (81)	36 (80)	12 (100)	14 (93)	75 (95)	20d (87)

^aOriginal smear. 

^bProbe negative and not further identified. 

^cSix M. szulgai isolates. 

^dEight M. szulgai isolates, five M. lentiflavum isolates, three M. asiaticum isolates, two M. scrofulaceum isolates, one M. marinum isolate, and one M. smegmatis isolate. 

Table 3 shows the original morphologic assessment of Kinyoun-stained smears from BACTEC 12B bottles. Definitive cording was found in 237 of 373 specimens (64%) which were positive for M. tuberculosis. M. kansasii was recognized with cross-barring in 63 of 76 (83%) of specimens. However, we initially thought that of the other mycobacteria, 20 of 590 (3%) showed cording and 43 of 822 (5%) showed ladders.

TABLE 3.

Original morphologic assessment from BACTEC bottles

Morphology	No. (%) of isolates (n = 899) of:
	M. tuberculosis(n = 373)	M. avium(n = 293)	M. kansasii(n = 76)	M. fortuitum- M. chelonae(n = 36)	M. xenopi(n = 12)	Mycobacterium speciesa(n = 14)	M. gordonae(n = 75)	Others (n = 20)
Cording positive	237 (64)	10 (3)	1 (1)	3 (8)	0 (0)	1 (7)	4 (5)	1b (5%)
Ladder positive	2 (0.5)	16 (5)	63 (83)	1 (3)	0 (0)	2 (14)	18 (24)	4c (20%)

^aProbe negative and not further identified. 

^bM. szulgai. 

^cTwo M. asiaticum isolates and two M. szulgai isolates. 

After training, a review of 110 available Kinyoun stained smears (Table 4) showed 18% more cording associated with M. tuberculosis and 83% (five of six) less cording associated with Mycobacterium species other than M. tuberculosis. As shown in Table 5, the SE, SP, PPV, and NPV for cording increased from 63.5, 96, 92, and 79% to 74.5, 98, 96, and 84%, respectively. Similarly, 12.5% more ladders were associated with M. kansasii and 70% (7 of 10) of the ladder designation were removed from non-M. kansasii strains (Table 4). Thus, with training (e.g., distinguishing the long forms associated with M. gordonae from the short forms associated with M. kansasii [Fig. 4]), the SE, SP, PPV, and NPV improved to 93, 98, 79, and 99% respectively (Table 5).

TABLE 4.

Morphologic assessment of a subset of isolates (n = 110) before and after review

Species of mycobacteria (n)	Cording			Ladders
	Originall no. of isolates from BACTEC	No. of isolates after review	% Improved with training	Original no. of isolates from BACTEC	No. of isolates after review	% Improved with training
M. tuberculosis (46)	27	32	18	0	0	NAb
M. avium (35)	3	1	67	2	0	100
M. kansasii (9)	0	0	NA	8	9	12.5
M. gordonae (14)	2	0	100	8	3	62.5
M. fortuitum-M. chelonae (4)	0	0	NA	0	0	NA
Mycobacterium species (2)a	1	0	100	0	0	NA

^aProbe negative and not further identified. 

^bNA, not applicable. 

TABLE 5.

Overall performance

Morphology	SE (%)	SP (%)	PPV (%)	NPV (%)
Cording
First reading	63.5	96	92	79
After review	74.5	98	96	84
Ladder
First reading	83	95	59	98
After review	93	98	79	99

FIG. 4.

Microscopic morphology of M. gordonae, exhibiting long beaded forms.

DISCUSSION

Several investigators have shown that M. tuberculosis has the ability to form serpentine cording from liquid medium with a sensitivity of 23 to 95% and a specificity of 95 to 100% (2, 3, 4, 5, 8). This phenomenon has been used as a rapid presumptive diagnostic test and a cost-effective guide to selection of nucleic acid probes (3, 6). Our results are in the middle range, as we found presence of serpentine cording in 64% of M. tuberculosis growing from liquid cultures (BACTEC).

The ability of other mycobacteria to form serpentine cords in BACTEC cultures is rare. In our study, cording was present in an average of 3 to 8% of the most common nontuberculous mycobacteria (M. avium, M. kansasii, and M. gordonae). Kaminski and colleagues (3) emphasized that their performance, with respect to recognition of serpentine cord formation, improved over time. Our results also improved after training. Eighteen percent more cording was recognized, and pseudocording (this was a looser aggregation than reference 4 shows) was distinguished from true cording in five of six (83%) of the nontuberculous mycobacteria. Because of this, the SE, SP, PPV, and NPV improved from 63.5, 96, 92, and 79% to 74.5, 98, 96, and 84% respectively. We reemphasize that without training and experience, serpentine cords can be missed or confused with pseudocords.

M. kansasii has special morphology that usually can be differentiated from that of other mycobacteria (2, 7; J. E. Clarridge, T. Raich, R. Shawer, D. Rhodes-Webb, and S. Karkhanis, Abstr. 94th Gen. Meet. Am. Soc. Microbiol. 1994, abstr. U-86, p. 211, 1994). In our group, 76 M. kansasii isolates grew from BACTEC cultures and 63 (83%) exhibited ladders. The SE, SP, PPV, and NPV before training were 83, 95, 59, and 98%, respectively. After review, these results improved to 93, 98, 79, and 99%, respectively, when we learned to differentiate M. gordonae from M. kansasii. However, even after training and with additional specimens examined after the study period, we could not distinguished M. kansasii from M. szulgai and M. asiaticum, which are similar taxonomically and in the disease produced. The presence or absence of cording and cross-barring, detected on microscopic examination of Kinyoun-stained smears prepared from BACTEC 12B bottles, can be used as a cost-effective method for presumptive identification of M. tuberculosis and M. kansasii and as a guide for initial probe selection.