Nonspecificity of Immunohistochemistry for Mycobacteria Species Using a Rabbit Polyclonal Antibody

The archive of the University of Michigan (Ann Arbor) was searched during a period of 6 years for a variety of cases of biopsied or resected untreated infections (both mycobacterial and nonmycobacterial) from a range of body sites and microorganisms, including bacteria, fungi, and parasites. Cases were excluded if blocks could not be located and if the original histochemical stains were not available for review. Forty cases were ultimately identified. Associated histochemical stains, including Twort tissue Gram stain, Grocott-Gomori methenamine silver stain, Fite modified acid-fast stain (Fite), Ziehl-Neelsen acid-fast stain (ZN), Warthin-Starry stain, and treponeme IHC were retrieved and reviewed when available. The medical charts were also reviewed for culture and PCR microbiologic studies.

Three pathologists (one with expertise in infectious disease pathology, one with expertise in IHC interpretation, and one general pathologist) reviewed the hematoxylin-eosin–stained slides and their respective histochemical stains. Sections of 4-μm thickness were deparaffinized, and heat-induced epitope retrieval was performed on the Ventana Benchmark Ultra immunostainer using cell conditioning 1 buffer from Ventana Medical Systems at 95°C for 36 minutes. After blocking endogenous peroxidase activity, the slides were incubated with a predilute polyclonal antibody to MTB (cat no. PP140AA, BioCare) for 32 minutes at 37°C. Immunoreactivity was detected by using the UltraView Universal DAB detection kit (Ventana Medical Systems). The assay was then rerun in a subset of both positive and negative cases, in addition to a negative reagent control in a separate subset of cases. The cases were then rereviewed with the MTB immunostain to evaluate immunoreactivity. Positive staining was defined as immunoreactivity for organisms with the appropriate morphology within a characteristic inflammatory background, such as granulomatous or mixed granulomatous and suppurative inflammation. False-positive staining was defined as immunoreactivity without appropriate organism morphology and/or lacking a characteristic tissue reaction for mycobacterial infection.

The details of the 40 cases that were evaluated, including specimen types, morphologic findings, histochemical stains, and microbiologic studies, are provided in Tables 1 through 3.

The cases evaluated included confirmed or presumed examples of Actinomyces, Aspergillus, Bartonella henselae, Blastomyces, Candida, Coccidioides, Enterococcus, E coli, Histoplasma, Helicobacter heilmannii, Helicobacter pylori, Leishmania, M avium intracellulare, Mycobacterium chelonae, Mycobacterium haemophilum, M leprae, Nocardia, P aeruginosa, Malassezia, Porphyromonas, Pneumocystis, Tissierella praecuta, and Treponema pallidum. Cases of chromoblastomycosis, dermatophytosis, onychomycosis, and spirochetosis were also included.

Twenty-six of the 40 cases (65%) were biopsy specimens, including core needle biopsy, transbronchial biopsy, gastrointestinal biopsy, skin punch biopsy, and toenail clipping. Other types of specimens included pulmonary wedge resection, lung lobectomy, splenectomy, laryngectomy, tonsillectomy, bronchial washings, and soft tissue debridement.

The MTB immunostain was positive in all 8 cases of mycobacterial infection (20% of total cases; Table 1), which were also positive by AFB stains (Figure 1, A through F). Interestingly, the staining pattern observed in cases of NTM showed a dotlike positivity rather than staining the entire rod, and the Fite or ZN stain highlighted the morphology of the organism much more clearly.

The MTB immunostain showed immunoreactivity in 15 (38% of total) of the 40 cases (Table 2; Figures 2, A through C, and 3, A through D), including Actinomyces, Blastomyces, Coccidioides, dematiaceous fungi, Enterococcus, Leishmania, Nocardia, and mixed bacterial flora (predominantly skin flora). In terms of bacteria (Figure 4, A and B), immunoreactivity was most often seen with clusters of cocci or with filamentous organisms such as Actinomyces, but was also observed with fungal organisms and Leishmania. The MTB immunostain was negative in the remaining 17 cases (43% of total; Table 3). A subset of these cases was restained to ensure reproducibility, and the assay performed similarly. The negative reagent control was also negative, with no nonspecific immunoreactivity.

The pathologic identification of AFB, including both MTB and NTM, is a very challenging activity for anatomic pathologists that has significant treatment implications. In terms of surgical pathology specimens, pathologists continue to rely primarily on histochemical stains such as ZN and Fite. However, these stains can be technically difficult to perform and equally difficult to interpret, particularly when the number of organisms is low. Furthermore, tissue is often not reserved for culture, and molecular diagnostic methods may not be suitable for use with routinely processed tissue.

For these reasons, the development of a mycobacterial IHC has been heralded with great enthusiasm. Anti-MPT64, a different monoclonal antibody against an MTB complex–specific secreted protein, initially showed promise in many studies,^6–18  including high specificity^{9–12,15,16,18}  and similar performance to Fite^6,7  and PCR,⁸  but was limited by variable protein secretion, protein mutations, and tissue accumulation,^11,15  necessitating a more practical antibody such as polyclonal MTB. A polyclonal antibody with multiple epitopes is advantageous in this setting, but specific epitopes targeted, including the Biocare antibody, are proprietary to the manufacturer and not indicated on the data sheet. Most of the published literature on IHC stains for use in mycobacterial infections (both the Biocare antibody and others) has focused on cases that were suspected to have MTB infection by clinical examination or pathologic findings.^{6–10,16,18–22}  Few detailed studies have investigated the cross-reactivity of these antibodies with other organisms. To our knowledge, we report the first study specifically investigating the cross-reactivity of the Biocare MTB antibody with a variety of pathogens, tissue types, and inflammatory responses.

Overall, our study supports the previous literature and shows that the MTB antibody is highly sensitive for a variety of mycobacteria. Our findings are similar to manufacturer reports describing reactions with NTM species (including M avium),⁴  and we also demonstrated immunoreactivity with M leprae, M haemophilum, and M chelonae. Immunoreactivity with Nocardia is also well reported with MTB,^6,7  which is believed to be because of cross-reactivity with cell wall mycolic acids that can be seen among Corynebacterineae such as Nocardia, Rhodococcus, and Corynebacterium, so the positive staining in our case of pulmonary Nocardia is not surprising.

As noted above, the MTB antibody demonstrated inferior morphologic detail when compared with Fite and ZN stains in cases of mycobacterial infection. Overall, the antibody yielded a dotlike pattern of reactivity with mycobacteria, and only focally stained the entire rod. This dotlike pattern of reactivity is significant in that a similar pattern is seen with cocci, and both mycobacterial and nonmycobacterial infections may have a suppurative and/or granulomatous inflammatory reaction. Lack of understanding of this staining pattern, as well as the nonspecificity of the antibody, could result in misidentification of an organism. Except for Nocardia, nontubercular cases with bacillary morphology (eg, Actinomyces and Pseudomonas) would also be expected to be negative for Fite and ZN stains.

Further complicating morphologic interpretation, IHC protocols vary significantly depending on the autostaining platform used, necessitating that all protocols undergo rigorous optimization to determine the optimal staining conditions. Given the differences that exist among platforms, suggested manufacturer protocols are merely recommended starting points and not intended to be unchangeable. The Biocare MTB antibody data sheet also does not provide recommended antigen retrieval conditions, but does suggest 30-minute incubation. Our optimization of the MTB antibody tested several different retrieval conditions, antibody incubation times, and temperatures. We experimentally determined that the optimal conditions for the MTB antibody were antigen retrieval with cell conditioning 1 buffer and an incubation of 32 minutes at 37°C, which is similar to the manufacturer recommendations.

Overall, the MTB antibody is nonspecific, and reacted with a wide variety of bacteria, fungi, and even parasites. The interpretation was often confounded by nonspecific staining in background bacterial cocci, such as occurred in all 3 culture-proven P aeruginosa cases, which could lead to misdiagnosis if the correct organism morphology is not appreciated. The immunoreactivity extended to several other types of tested bacteria, fungi, and Leishmania. Some studies have also demonstrated other types of nonspecific staining with MTB. For example, one study using the MTB antibody reported higher sensitivity for NTM than MTB.⁷  Another study using this antibody described 4 cases of nonspecific staining in eosinophils, which is further problematic as eosinophils are associated with a wide variety of inflammatory reactions.²³  The nonspecificity of the MTB antibody emphasizes careful consideration of not only the immunoreactivity, but also the appropriate organism morphology in the appropriate tissue background, for accurate interpretation. This is a critical caveat, as it is well-known in common practice that morphology is not routinely assessed in IHC review, which could lead to false-positive interpretation with this antibody. This is particularly true given that some of the organisms in this study with inflammatory backgrounds similar to mycobacterial infections showed false positivity with the MTB antibody. In some of our cases that showed immunoreactivity for other types of organisms, the lack of an inflammatory response or the presence of an inflammatory response other than what one would expect with mycobacterial infection would be a clue that these were not truly mycobacterial infections. However, several cases containing nonmycobacterial organisms that were positive with the antibody showed granulomatous or necrotizing granulomatous inflammation, which would be expected in many mycobacterial infections. In addition, NTM may show a wide variety of inflammatory responses, some of which may not be granulomatous.²⁴  The morphology of some of the pathogens that were immunoreactive (for example, Blastomyces) is also quite different from that of small rod-shaped bacteria, but can be difficult to discern if organisms are fragmented in a small biopsy.

There is no question that IHC for mycobacteria has its place in diagnostic pathology, as it has many benefits, including reasonable cost, faster and simpler processing, and easier interpretation in significantly less time.^6–8,17  It can be performed in routine, low-volume laboratories, and is ideal for low-resource areas, such as socioeconomically disparate communities and foreign countries that are at higher risk for MTB.^8,13,14  However, it is very important to be aware of the potential for cross-reactivity with a variety of organisms, particularly those that may have a granulomatous and/or necrotizing tissue reaction similar to the inflammatory patterns expected with mycobacteria. Although some publications have suggested replacing conventional histochemical stains with mycobacterial IHC,^6,7  our findings demonstrated nonspecific staining with mycobacterial IHC and urge caution so that nonmycobacterial infections, background bacteria, or even normal cells are not overinterpreted as mycobacterial infection. Because of the nonspecificity, mycobacterial IHC could, instead, be used as a screening test to triage specimens for molecular testing, as previously suggested by Crothers et al.⁶  Given the nonspecificity of the MTB antibody, we decided not to use the antibody in our laboratory. We, instead, chose to continue the traditional use of AFB stains, which perform well in our laboratory setting. However, this is not true for all laboratories; some may benefit from more widespread use of the MTB antibody, keeping in mind its limitations.

In summary, the MTB antibody is highly sensitive but nonspecific for mycobacterial infection. In the first study that tested the MTB antibody against a wide range of pathogens, tissues, and specimen types, we showed that the antibody is reactive with a wide variety of pathogens, both bacterial and nonbacterial. Mycobacterial IHC is a valuable diagnostic tool, but should be implemented and interpreted with caution to avoid false-positive results that can lead to unnecessary and harmful medical treatment.

The authors gratefully acknowledge Paul Lephart, PhD, and Michael Bachman, MD, PhD, for assistance in reviewing the manuscript.