Detection of Corynebacterium kroppenstedtii in Granulomatous Lobular Mastitis Using Real-Time Polymerase Chain Reaction and Sanger Sequencing on Formalin-Fixed, Paraffin-Embedded Tissues

We retrospectively identified cases of GLM during a 10-year period, using keyword searches within the anatomic pathology information system, in addition to 10 cases of nongranulomatous breast abscess included as controls. Patient demographics including age, ethnicity, body mass index, gravidity and parity, history of lactation and hyperprolactinemia, clinical presentation, radiographic findings, histologic features, Gram stain, and microbial culture results were reviewed. Hematoxylin and eosin–stained slides from all cases were reviewed by a breast pathologist to assess for features of GLM. Only cases with defining histologic features of GLM (ie, nonnecrotizing granulomatous inflammation in and around the lobules) were included in this study (Figure 1, A). All other causes of granulomatous inflammation (infectious and autoimmune) were ruled out clinically and histologically in each case. The GLM cases were further assessed for histologic features of CNGM, defined as clear spaces (microcysts) surrounded by a rim of neutrophils and further surrounded by histiocyte-rich granulomatous inflammation (Figure 1, B), and based on this assessment were subdivided into CNGM and non-CNGM categories. Similarly, 10 nongranulomatous breast abscess cases were reviewed by a breast pathologist to ensure that features of GLM were absent (Figure 2, A and B). These cases were included in the control group. Gram, acid-fast bacillus, and Gömöri methenamine silver histochemical stains were performed on all 67 GLM cases as well as the 10 control cases.

FFPE tissue blocks from all 67 GLM and 10 control cases were retrieved. A representative block was selected from each case for molecular study (1 block per case). Five-μm to 10-μm-thick FFPE scrolls were collected from each FFPE block into an Eppendorf tube using PCR precaution protocol to avoid contamination. FFPE sections were deparaffinized by CitriSolv, and DNA was extracted from FFPE sections using the EZ1 DNA Tissue Kit (Qiagen) on the BioRobot EZ1 System (Qiagen). DNA concentration was measured by the NanoDrop spectrophotometer (Thermo Fisher Scientific).

DNA was amplified by SYBR real-time PCR (PowerUp SYBR Green Master Mix, Thermo Fisher Scientific) with primers specifically targeting the C kroppenstedtii 16S rRNA gene region (C kroppenstedtii–specific primer set)²²  on the ABI 7900HT real-time PCR System (Thermo Fisher Scientific). For each sample, PCR was performed in duplicate. C kroppenstedtii–negative DNA control and non-DNA template controls were included in each run to monitor PCR contamination issues. In addition, a human control gene cyclophilin (Cy FW and RV primer set)²³  was amplified along with the C kroppenstedtii PCR to monitor DNA quality and PCR efficiency.

C kroppenstedtii PCR–positive samples were bidirectional Sanger sequenced using BigDye Terminator Cycle Sequencing Kit (Thermo Fisher Scientific) on the ABI 3130xl Genetic Analyzer. Obtained sequences were analyzed using DNASTAR Lasergene 10 software (DNASTAR, Inc.). Resulting sequences were queried in the GenBank database using BLASTn (accessed on September 2019), and results with the highest alignment scores were analyzed. In addition, sequences from all C kroppenstedtii–positive cases were aligned against C kroppenstedtii partial 16S rRNA sequence (strain DSM 44385, GenBank# NR_074408) using the AliView program (version 1.26, Larsson, A., 2014).

Sixty-seven cases fulfilling histologic criteria for GLM were identified within the 10-year period. In addition, 10 non-GLM cases with histologic features of breast abscess were included in the control group. The clinical characteristics including patient age, ethnicity, body mass index, obstetric history, history of lactation and hyperprolactinemia, and history of diabetes, and the presenting symptoms in GLM and non-GLM patients are summarized in Table 1.

Of 67 GLM cases, 38 (56.7%) showed definitive histologic features of CNGM. Gram stain performed on the histologic sections showed rare Gram-positive bacilli in only 12 of 67 (17.9%) GLM cases. All 12 Gram-positive cases were in the CNGM category and showed rare bacteria exclusively inside the cystic spaces lined by neutrophils (Figure 3). All 10 cases in the control group were negative for bacteria on Gram stain and did not show features of CNGM. Acid-fast bacillus and Gömöri methenamine silver were negative for microorganisms in all 67 GLM cases as well as the 10 control cases. Breast aspirates for microbial cultures were obtained in only 38 of 67 (56.7%) of the GLM cases at the time of the biopsy procedure, of which 22 of 38 (57.8%) were negative for bacterial growth, 14 of 38 (36.8%) were positive for Diphtheroids, 1 grew coagulase-negative Staphylococci, and 1 was positive for Streptococcus angionosus. In the control group, breast aspirates for microbial cultures were obtained in 7 of 10 (70%) cases. Three of 7 were positive for Staphylococcus aureus, 2 of 7 were positive for Actinomyces israelii, and 1 each was positive for Staphylococcus lugdunensis and Streptococcus angionosus.

For molecular testing, human control gene cyclophilin (Cy) was used to determine the suitability of the FFPE samples for PCR-based assay, and all 77 samples showed Cy amplification by PCR. In the GLM category, 46 of 67 (68.7%) cases were positive for C kroppenstedtii by 16S rRNA SYBR real-time PCR. All PCR-positive cases were confirmed by Sanger sequencing. Clean sequences were obtained from all 46 cases with sequence lengths ranging from 104 bp to 136 bp. BLASTn search results indicated that sequences from these 46 cases were a 100% match with the C kroppenstedtii partial 16S rRNA sequence (strain DSM 44385, GenBank# NR 074408). In addition, the alignment study by the AliView program also confirmed that all 46 sequences were aligned to the C kroppenstedtii partial 16S rRNA sequence (see Figure 4). The remaining 21 of 67 (31.3%) of GLM cases in our study were either negative (n = 17) for C kroppenstedtii real-time PCR or weak positive but the sequence failed to confirm the C kroppenstedtii strain (n = 4). All 10 control cases were negative for C kroppenstedtii by PCR. Our findings suggest that C kroppenstedtii is specific to GLM and is not seen in association with nongranulomatous breast abscess (P < .001, Fisher exact test).

Histologically, 28 of 46 (60.9%) of C kroppenstedtii PCR–positive cases were in the CNGM category. Gram stain was positive for Gram-positive bacilli in only 8 of 46 (17.4%) of the C kroppenstedtii–positive cases by SYBR real-time PCR. Of note, 4 cases negative for C kroppenstedtii by PCR were positive for Gram-positive bacilli by Gram stain on tissue sections. Microbial cultures were obtained in only 25 of 46 of C kroppenstedtii–positive cases. Of the 25, 10 (40%) were positive for Diphtheroids and 1 (4.0%) was positive for coagulase-negative Staphylococci, while 14 (56%) were negative for microbial growth. The clinical, histologic, Gram stain, and microbial culture findings in C kroppenstedtii PCR–positive versus negative cases are summarized in Table 2.

The etiology of GLM is not well established, and the disease is often regarded as idiopathic; however, recent research indicates that human breast tissue is not sterile but contains a diverse microbiome that includes Corynebacteria species.^24–26  Dysbiosis involving this microbiome may contribute to the development of several disease states.²⁵  The association between GLM and lipophilic Corynebacteria in cultures was communicated for the first time in 1996 by Binelli et al.²⁷  Subsequently, Taylor et al⁵  demonstrated the presence of lipophilic bacteria with Coryneform morphology inside the neutrophil-lined cystic spaces in histologic sections of GLM. This prompted them to perform 16S rRNA gene sequencing on selected culture specimens that yielded C kroppenstedtii as the most common pathogen. This was followed by a study by Yu et al²⁰  in which the microbiota of GLM was studied using 16S rDNA metagenomic sequencing on fresh breast aspirates. They found Corynebacteria in all 19 of their patients with C kroppenstedtii being the most common species. A similar study using 16S rRNA gene sequencing on fresh breast samples from 15 CNGM patients by Johnstone at al⁸  showed C kroppenstedtii to be the most prevalent pathogen in these patients.

Data on the yield of C kroppenstedtii using 16S rRNA on FFPE tissues from GLM patients are limited. In 2018, Gautham et al²⁸  described Corynebacteria within neutrophil invested microcysts in 4 of 5 patients with CNGM using Gram stain on histologic sections; however, their attempt at performing 16S rRNA PCR on FFPE tissues on these cases was unsuccessful. In 2018, Fuji et al²²  for the first time demonstrated C kroppenstedtii genome encoding 16S rRNA in DNA extracted from FFPE sections of GLM cases using a C kroppenstedtii–specific primer set yielding 7 positive cases of 18. Most recently, Naik et al²¹  identified Corynebacteria by 16S rRNA sequencing on DNA extracted from FFPE tissues in 12 of 23 GLM cases. We performed this study to test the strength of this reported association of C kroppenstedtii with GLM via SYBR real-time PCR on FFPE tissues at a larger scale and to compare the rate of C kroppenstedtii detection in GLM cases with nongranulomatous breast abscess cases. Our study showed that 46 of 67 (68.7%) GLM cases were positive for C kroppenstedtii 16S rRNA by SYBR real-time PCR. All 46 cases had the highest alignment score with C kroppenstedtii partial 16S rRNA sequence. Our yield of detecting C kroppenstedtii by SYBR real-time PCR on FFPE tissues of 68.7% is higher than reported by Fuji et al²²  and Naik et al²¹  in their studies (28.8% and 52.1%, respectively). In addition, in our study, all 10 cases of nongranulomatous abscess in the control group were negative for C kroppenstedtii DNA suggesting that C kroppenstedtii is highly prevalent and specific to GLM and does not appear to be associated with nongranulomatous breast abscess (P < .001). Among the C kroppenstedtii–positive cases, the majority showed CNGM histology (60.9%).

The yield of Gram stain for Corynebacteria in FFPE tissue sections in our study was significantly lower than by PCR (17.9% versus 68.7%). Corynebacteria are notorious for staining poorly in clinical samples,²⁹  and the sensitivity is even lower in Gram-stained FFPE tissue sections. This is also explained by the fact that the bacteria in GLM cases, as shown in our study, are scant and found exclusively inside the neutrophil-rimmed cystic spaces. Microbial culture data were only available in 38 of 67 (56.7%) GLM cases. This is because many of our patients presented with a palpable breast mass that was clinically concerning for malignancy and samples for cultures were not procured at the time of the biopsy procedure due to low clinical suspicion for an infectious or inflammatory process. Of the 38 GLM samples that were cultured, 16 were positive for bacterial growth (42.1%). Overall, 14 of 38 (36.8%) cases with available culture data were positive for “Diphtheroids.” Species-level identification was not pursued in any of these 14 cases likely because many Corynebacteria are part of the normal microflora of human skin, mucous membranes, and body fluids and are usually regarded as commensals/contaminants in the lab. Nevertheless, in our study Diphtheroids were the most common isolates among the positive cultures. Our study shows that the yield of detecting C kroppenstedtii in GLM cases is significantly higher by SYBR real-time PCR on FFPE tissues as compared with Gram stain on histologic sections (68.7% versus 17.9%). The yield of detecting C kroppenstedtii by microbial cultures versus SYBR real-time PCR on FFPE tissues is difficult to analyze in our study as the cases positive for Diphtheroids were not pursued further for speciation; however, it is likely lower than PCR given that only 14 of 38 (36.8%) cases with available culture data were positive for Diphtheroids (versus 68.7% by PCR). If we take into consideration the limitations of performing PCR on FFPE tissues, the yield is likely higher than 68.7% and can be further improved by the use of fresh samples.

We recommend that fresh samples be obtained in all cases of suspected GLM for microbial studies. This will not only improve the detection rate of C kroppenstedtii by PCR in these cases but also allow antimicrobial susceptibility testing in culture-positive cases that may open doors to targeted antibiotic therapies and hopefully improved clinical outcomes. At present, there is no consensus in the literature regarding the treatment for GLM. The proposed treatment regimens include steroids, immunosuppressants, antibiotic therapy, surgical debridement, and watchful waiting.^30–33  Irrespective of the treatment strategy employed GLM has a high rate of relapse.^34,35  An infectious etiology has long been a leading hypothesis with regard to GLM, but the frequent negative microbial cultures¹³  and the lack of response to antibiotic therapy¹⁹  have led to some skepticism. The negative cultures are likely attributable to the lipophilic nature of Corynebacteria, which typically require specific media and prolonged incubation. Furthermore, “Diphtheroids” in smears and colonies are frequently regarded as commensals or contaminants in the lab and are not pursued further, as seen in our study. In addition, Corynebacteria are poorly susceptible to the beta-lactam antibiotics that are traditionally prescribed for breast infections and many are multidrug resistant.³⁶  In recent years several case reports and case series claiming successful treatment of GLM solely with an extended course of lipophilic antibiotics targeting C kroppenstedtii have been described.^17,37–40  Large-scale studies focused on improved detection of C kroppenstedtii in GLM and treatment with targeted lipophilic antibiotics are needed to ensure a convincing response to therapy.

Of note, 21 of 67 (31.3%) of GLM cases in our study were negative for C kroppenstedtii by PCR. These 21 cases include four Gram-stain positive and 9 cases positive for Diphtheroids in microbial samples. This discrepancy may be in part due to formalin-related nucleic acid degradation.⁴¹  Second, only 1 block per case was selected for PCR, hence, a false-negative result due to sampling error is a very likely possibility. Third, many of our FFPE specimens were small core biopsies, hence, the lesional tissue in the block may have been exhausted during processing. Last, the contents of the cystic spaces may be lost with deparaffinization and DNA extraction leading to false low yield. The use of fresh biopsy samples obtained for microbial studies might improve the detection sensitivity by PCR-based methods.

Certain important limitations of our study should also be noted. First, several factors may influence PCR results, such as small tissue size, variability in tissue processing, contamination of tissue or blocks, and the effects of formalin fixation upon nucleic acid quality and quantity. We have attempted to mitigate these factors through the use of analytic controls as well as case controls to minimize false-negative or false-positive results. Also, since we expected relatively small nucleic acid fragments in FFPE tissue, only a C kroppenstedtii–specific primer set (with PCR product size of 134 bp) was selected for this study. The use of this species-specific primer precludes the detection of other bacteria in our study. Larger studies using multiplex PCR techniques are needed to exclude the presence of other bacterial species in GLM.

We conclude that C kroppenstedtii is highly prevalent in GLM, supported by the detection of C kroppenstedtii DNA in 68.7% of GLM FFPE tissues by real-time PCR. In contrast, C kroppenstedtii was not found in nongranulomatous breast abscesses in our study. The prevalence of C kroppenstedtii is higher in GLM with histologic features of CNGM as compared with non-CNGM. Gram stain has low sensitivity for detecting Corynebacteria in GLM tissue sections and when positive shows bacteria exclusively inside the neutrophil-rimmed cystic spaces. Microbial cultures positive for “Diphtheroids” in smears and colonies should not be dismissed as commensals/contaminants in the setting of GLM, and species-level identification should be pursued in all cases. We strongly recommend obtaining fresh biopsy samples for microbiologic studies in all cases of suspected GLM. This will not only improve the detection rate of C kroppenstedtii by PCR in these cases but also allow antimicrobial susceptibility testing in culture-positive cases that may open doors to targeted antibiotic therapies and hopefully improved clinical outcomes.