Targeted histological evaluation shows high incidence of actinomyces infection in medication-related osteonecrosis of the jaws

Medication-Related Osteonecrosis of the Jaws (MRONJ) is a difficult-to-treat complication of the therapy of osteoporosis and some malignancies cured with bisphosphonates and antiresorptive drugs. The pathomechanism is unclear, but there is increasing observation that Actinomyces infection may play a role in its development and progression. The aim of our study was to demonstrate that histological examination using a validated triple staining procedure for Actinomyces bacteria strains can detect a high rate of Actinomyces infection in patient's samples with MRONJ. 112 previously hematoxylin-eosin (HE) stained samples submitted with the clinical diagnosis of MRONJ were re-evaluated histologically using an appropriate triple special staining validated for the identification of Actinomyces infection. During the first evaluation, when pathologists did not specifically look for Actinomyces, only 8.93% of the samples were reported as positive. In contrast, re-evaluation with triple staining provided a yield of 93.7% positive samples, therefore, we suggest the triple special staining to be standard in MRONJ histology evaluation. These results show that if the clinician suspects Actinomyces infection and brings this to the attention of the pathologist, it could significantly increase the number of correct diagnoses. It serves as an aid for clinicians in therapeutic success of MRONJ by selecting a long-term adequate antibiotic medication which is suitable for the elimination of actinomyces infection.


Objectives
We aimed to re-evaluate hematoxylin-eosin (HE) stained histological samples from patients diagnosed with MRONJ for the presence of Actinomyces.

Materials and methods
Study design and clinical data. Samples were obtained from patients diagnosed with MRONJ between 2011 and 2020 at the Department of Oro-Maxillofacial Surgery and Stomatology and Oral Diagnostics Department, Faculty of Dentistry, Semmelweis University, without any selection. The involved patients belonged to stage 2 and 3 according to American Association of Oral and Maxillofacial Surgeons (AAOMS) classification 5 .
In accordance with the recent guidelines, surgical removal of sequesters was completed with antibacterial treatment as follows: patients started antibiotic therapy 3 days before surgery and continued for 10 days in the postoperative period. In cases when appropriate microbiological data were available, definitive antibiotic therapy was used based on the sensitivity. In other cases, the empirical therapy included amoxicillin-clavulanic acid (875/125 mg, 2 × 1) as the first choice, or clindamycin (300 mg 4 × 1) or doxycycline (100 mg, 2 × 1) in patients with penicillin allergy. Cooling of the area and the use of 0.2% chlorhexidine mouth rinse at least 3 times a day for one week after surgery was recommended 40 .
Altogether 117 samples of 83 patients from the archives of the Department of Pathology and Experimental Cancer Research, Semmelweis University from 2011 to 2020 were re-evaluated. The study design was approved by the local Ethics Committee (SE RKEB 137/2020). The data of all patients were anonymized and procedures were in line with the requirements of the Helsinki Declaration. Due to the retrospective study design there was no need for ethical approval of any informed consent. Hereafter we use the term "evaluation" for the original hematoxylineosin (HE) study and "re-evaluation" for the second examination with specific stains for Actinomyces. A total of 117 archived samples were compared with the original histological findings during the re-evaluation. 3 samples were excluded due to the lack of unambiguous osteonecrosis and 2 other due to the insufficient bone content for further analysis. Finally, 112 archived samples were involved in the study.
All of the involved patients were treated with bisphosphonates, 101 cases with oncological indication, and 11 cases with osteoporosis. The latter group received bisphosphonate therapy only. Multiple (2-4) specimens were available from 33 patients due to persistent or recurrent lesions.
We compared the histological data with available microbiological results of the patients included in this study. Microbiological examination was performed in 39 cases within 3 months before or after the operation. For the calculation of sensitivity and specificity we made a 2 × 2 table with groups of subjects divided according to the triple staining as reference method in columns, and categories according to microbiological results in rows.
Histological analysis. All the included samples were obtained from the surgical treatment of osteonecrosis following bisphosphonate treatment and the clinical diagnosis was MRONJ. Sections from the decalcified, formalin-fixed paraffin-embedded biopsy were stained with HE. Histological criteria included the presence of completely necrotic bone trabeculae and bacterial aggregates along with variable inflammatory infiltrates as it was previously described. During the re-evaluation, beside the HE histomorphology, the presence of Actinomyces was confirmed using Gram (Bio-Optica, Milano, Italy), periodic acid-Schiff (PAS; Surgipath Schiff reagent, Leica Biosystems, Richmond, USA), and Grocott's methenamine silver (GMS; Biognost, Zagreb, Croatia) staining. Slides were evaluated using a Nicone Eclipse E600 POL microscope first, then the most representative sam- www.nature.com/scientificreports/ ples were digitalized using a Pannoramic digital slide scanner (3DHISTECH Ltd., Hungary). Image acquisition was performed using the Case Viewer software (3DHISTECH Ltd., Hungary). Samples containing the characteristic filamentous bacterial colonies ("sulphur granules") showing positivity with all three stainings were regarded as "positive" to Actinomyces.
Original routine histological evaluation had been performed by 14 different pathologists in the described time period and the clinical diagnosis of osteonecrosis was confirmed in all cases. Re-evaluation of the samples was performed by 1 trained pathologist focusing on the presence of Actinomyces infection. During the re-evaluation, the examiner was blinded regarding the previous pathology report of presence or absence of Actinomyces or any microbiological data. To test the examiners reliability, he performed a repeated re-evaluation of the same samples, one and a half year following the first re-evaluation (the examiner was masked to the previous results). The intraobserver agreement was calculated with Cohen's kappa; value is 0.93 (95% truncated CI 0.79-1). That indicates good agreement.
In the samples containing both viable and non-viable bone tissues the two areas were also compared from the perspective of Actinomyces content. Bone was considered vital when viable osteocytes were visible in the lacunae.
Statistical analysis. SPSS Statistics 27 (IBM corp., USA) was used for statistical analysis. Group comparisons were performed using the Related-Samples McNemar Change Test with 95% confidence interval.
The intraobserver agreement was assessed by using Cohen's kappa with its 95% confidence interval was determined; we included the data to a useable structure to dncs_data1.xlsx, that was used as input data. The confidence interval limits was truncated to 0 or 1 if the limits are below 0 or over 1. The calculations were made in R (R Core Team 2021, v4.1.1) using the psych (Revelle 2021.2.1.9) package.

Ethical approval. The study design was approved by the Semmelweis University Regional and Institutional
Comitte of Science and Research Ethics 137/2020 (Hungary). Due to the retrospective study design there was no need for ethical approval of any informed consent: Semmelweis University Regional and Institutional Comitte of Science and Research Ethics 155/2012 (Hungary). The study was done according to the declaration of Helsinki.

Results
The evaluated samples contained necrotic, acellular bone trabeculae with bacterial aggregates between them ( Fig. 1.), usually surrounded by only mild or even absent inflammatory infiltrate (Fig. 1A). Reactive changes of the bone showed similar characteristics and were absent in the majority of cases. Actinomyces aggregates showed the characteristic radial arrangement on HE stained sections in "positive" samples ( Fig. 1A), along with PAS, GMS and Gram positivity ( Fig. 1C-E, respectively). In cases showing more severe inflammatory reaction, the infiltrate was usually confined to certain areas of the specimen and uninvolved necrotic bone was also present (Fig. 1E). It could also be demonstrated in some specimens that Actinomyces was present in the necrotic bone, while in viable and inflamed areas, the bacterial colonies were not detected (an example is demonstrated on Fig. 2). In 3 cases, fungi were also detectable (Fig. 1F).
Among the 112 archived samples 102 (91.07%) were reported as Actinomyces negative during the first routine histological evaluation with HE stainings, of which 95 were found to be positive during the re-evaluation with triple specific stains. Compared with the original report 7 samples (6.25%) were found to be negative during the re-evaluation. In the evaluation, the presence of Actinomyces was detected in 10 samples (8.93%) and this result was confirmed in all cases in the re-evaluation, so in the end a total of 105 (93.75%) samples were found to be positive. The result proved highly significant (p = 0.0000) ( Table 1).
As a result of the microbiological examination the presence of Actinomyces could be detected in cultures in only 2 of 39 cases (5.13%), all of which were evaluated histologically as "positive" during the re-evaluation. No samples were histologically negative and microbiologically positive. 2 of 39 samples were negative with both histological and microbiological examinations. Comparison of triple staining histology and microbiology results resulted in high specificity (1) and very low sensitivity (0.054). Negative predictive value of routine microbiological test was 0.054.

Discussion
MRONJ is a multifactorial, not a life-threatening disease, but significantly impairing the quality of life. There are still many questions about the pathogenesis, but most authors agree that infections may play a major role in its development. Many of them have raised the possibility that Actinomyces strains are potentially involved in this process, in addition to other bacteria 4,9,13,20,[41][42][43] . A highly variable proportion of Actinomyces positivity can be found in the literature, however, in the studies with lower prevalence the methods of Actinomyces detection, including microbiology or HE staining of histological samples, were probably inadequate regarding sensitivity 25 . Cerrato's analysis of 30 publications showed that Actinomyces positivity occurred in 96.4% to 25.4% of the samples 7 . These variances are presumably related not only to methodological differences but whether the study design was focused on the presence of Actinomyces or not.
In our study, we re-evaluated the histopathologycal samples in our database that had been received with a MRONJ diagnosis over a 10-year period to examine the frequency of Actinomyces. To the best of our knowledge there has been no study to date that has re-stained and re-evaluated previously tested samples. The re-evaluation of archived samples with a special focus on and using appropriate stains for Actinomyces, showed positivity in www.nature.com/scientificreports/ 93.75% compared to the 8.93% of the original routine evaluation. There were no cases where the sample initially found to be positive was found as negative in the re-evaluation. Only 7 out of the 112 samples proved to be negative in both the routine and the re-evaluation. These results suggest that one of the reasons for the significant discrepancy in the literature reports on the prevalence of Actinomyces may be a methodological problem. The routinely used HE staining resulted in a sparse detection of Actinomyces, which therefore seems more likely to be an accidental finding on MRONJ samples.
The sensitivity of microbiological cultures is very poor regarding the detection of Actinomyces species 27,29,51,52 . Microbiological results were available in case of 39 samples in the patient's records. Actinomyces could only be detected in two samples, representing a prevalence of 5.13% versus the 93.75% result of histology. All two samples were evaluated histologically as "positive" during the re-evaluation. These data confirm the well-known fact that microbiology is not a proper method to detect Actinomyces and underlie the importance of adequate histological assessment.
In dental plaques, Actinomyces species create co-aggregates with other bacteria, mainly Fusobacteria, Prevotella, Eikenella and Veillonella strains, during the formation of the biofilm. The role of this cooperation seems to be basically important in the pathogenesis of osteomyelitis as well, in which these species adhere to collagen fibers and promote the development of osteonecrosis 43,44 . During the review of the microbiological data, we found that Prevotella, Fusobacteria, Parvimonas, Veillonella, Eikenella, Staphylococcus, Streptococcus and Klebsiella species were most frequently found in the samples.
The potential causal role of Actinomyces in the pathogenesis of MRONJ is still an unresolved issue 9,15,36,37 , however, and increasing number of authors suggest a causal role of Actinomycetes in the development of MRONJ based on the higher prevalence detected by histological evaluation 7,20,22,37,39,41,44 . Russmueller et al. found a high (89%) prevalence and concluded that Actinomyces strains play a prominent role in MRONJ and will change our understanding of it. In a systematic review published in 2020, Cerrato concluded that osteomyelitis caused by Actinomyces and MRONJ may have a common origin with the notion that the lower prevalence reported in earlier series is at least partially related to methodological problems. Our study supports Cerrato's latter view.
The therapy of MRONJ is generally based on the recommendation of AAOMS (American Association of Oral and Maxillofacial Surgeons). According to the AAOMS principles, systemic antibacterial therapy is recommended from stage 3, but the above mentioned position paper does not mention the length of antibiotic therapy.
The MRONJ protocols of different countries usually recommend a few weeks of antibiotic treatment after surgery 12 . In contrast, the recommended antibiotic treatment protocols for actinomycosis of the jawbone start www.nature.com/scientificreports/ with intravenous broad-spectrum antibiotics (amoxicillin-clavulanic acid, ampicillin-sulbactam) for 3-4 weeks and continue with oral penicillin for 9-12 months or longer 45,47,48 , whereas Actinomyces strains are sensitive to beta-lactam compounds, but these compounds show very poor potential to penetrate fibrotic, necrotic and inflamed tissues and large bacterial aggregates 49 . Therefore, following a few weeks of intravenous therapy penicillin V can be used for the long term, significantly decreasing the incidence of gastrointestinal side effects compared to penicillins with beta-lactamase inhibitors 27,43,44,47,[50][51][52] .  www.nature.com/scientificreports/ The proportion of Actinomyces positive samples in our study were undoubtedly higher in comparison with some previous reports, although not far from the result of Rusmueller et al., or even lower than others, Anavi-Lev, Lee, Naik, Hansen, Franco-Pretto 16,37,39,45,46,53 . However, the antibiotic regime used in our patients is unlikely to affect the proportion of Actinomyces positive cases, since it is started only three days before the sampling. It is important to note that these short preoperative treatment is insufficient to eliminate Actinomyces.
Summarized, it would be important for the clinician to consider the possibility of Actinomyces infection in all MRONJ samples and bring this to the attention of the pathologist, who will then search for it using validated methods. This will presumably increase the frequency of detected Actinomyces infections and allow for the selection of an appropriate therapy. The focused histological examination, including triple stains and appropriate clinical issues, can be significant, and result in much higher detection rate of Actinomyces in comparison with isolated HE staining. This is highlighted by the fact that the microbiological detection of Actinomyces using conventional techniques shows poor sensitivity.
Additionally, long-term antibiotic treatment protocols should be considered in the postoperative pharmacotherapy of MRONJ, in order to eradicate Actinomyces species. The currently applied too short antibiotic therapy may contribute to the large number of therapeutic failures.

Data availability
The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.