Introduction

Rosai–Dorfman disease (RDD), also known as sinus histiocytosis with massive lymphadenopathy, is an R-group non-Langerhans cell histiocytosis with diverse clinical presentations.^{1, 2} While it was originally described in cervical lymph nodes, it can involve numerous nodal and extranodal sites, including viscera, skin and soft tissue.^3-5 It can also present as either self-limited localised disease curable with excision or an aggressive multisite process necessitating systemic chemotherapy. Similarly, putative aetiologies of RDD are markedly heterogeneous. Recently, mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) pathway mutations, commonly affecting MAP2K1 and KRAS, were detected in up to half of RDD, indicative of a neoplastic process.^6-10 However, other cases are polyclonal and can be associated with malignancy, infections and autoimmune conditions.^{11, 12} Regardless of mutation status, RDD shows consistent expression of p-ERK and cyclin D1, leading to emerging consensus that it results from MAPK/ERK pathway activation via a variety of mechanisms.^13-15

The sinonasal tract is one of the most common sites for extranodal RDD. Sinonasal involvement is present in 11% of RDD patients, of whom 70% also have multisite disease.^16-19 Sinonasal involvement does not drive overall prognosis, although it can be associated with aggressive behaviour in systemic disease.^{16-18, 20} However, localised RDD also can arise in the sinonasal tract and generally has excellent outcomes after surgical excision alone.^16-22 Recently, understanding of sinonasal RDD became more complicated with description of RDD-like histiocytes – a phenomenon defined by the presence of histiocyte aggregates with histological features of RDD in patients with chronic sinusitis symptoms.²³ It is currently unclear whether these histiocytes truly represent a separate process or a localised manifestation of RDD. In this study, we aimed to further characterise the clinicopathological and molecular features of this phenomenon to clarify its prevalence, classification and clinical implications.

Materials and methods

Results

HISTOLOGICAL FINDINGS

The 27 cases with histological features of RDD displayed intact sinonasal mucosa with variable polyp formation (Figure 1A). Prominent aggregates of histiocytes and lymphocytes conferred a checkerboard pattern at low power (Figure 1B). The histiocyte aggregates were largely nodular and compact with a granuloma-like appearance (Figure 1C); confluent growth was rare (Figure 1D). There was abundant oedematous stroma with scattered inflammatory cells between the histiocyte aggregates. While most histiocyte aggregates were dispersed irregularly (Figure 2A), a subset were perivascular (Figure 2B). However, no overt vasculitis or angiodestruction was seen. Other aggregates surrounded seromucinous glands (Figure 2C) or surface epithelium (Figure 2D). Histiocytes were large and epithelioid (Figure 3A), spindled (Figure 3B) or occasionally foamy (Figure 3C), with abundant eosinophilic cytoplasm and round to oval nuclei without grooves or indentations (Figure 3D). They were consistently intermixed with dense lymphocytes and plasma cells (Figure 4A). Emperipolesis was evident on haematoxylin and eosin (H&E) in most cases (Figure 4B). Most cases also showed prominent intermixed eosinophils (Figure 4C) with occasional eosinophilic emperipolesis (Figure 4D). While histiocytes themselves occupied a low volume of the tissue, numerous aggregates were scattered throughout all specimens, comprising the most notable pathological feature of these cases. Five cases had additional histological findings, including allergic fungal sinusitis in three (11%), mycetoma in one (4%) and antrochoanal polyp in one (4%).

IMMUNOHISTOCHEMICAL FINDINGS

Immunohistochemical findings are summarised in Table 1. All 27 cases with histological features of RDD displayed nuclear and cytoplasmic positivity for S100 protein (100%), with diffuse (Figure 5A) to patchy staining (Figure 5B). S100 protein highlighted emperipolesis in all cases where it was not evident on H&E (Figure 5C), such that 100% ultimately had emperipolesis. All 23 cases tested were positive for cyclin D1 (100%) within histiocyte aggregates (Figure 5D). All 25 cases tested were negative for CD1a (0%) and all 21 tested were negative for CD207 (0%). The 11 cases that had prominent histiocytes but lacked histological features of RDD were negative for S100 protein expression (0%).

Discussion

In this study, we confirm the presence of features of RDD in a subset of chronic sinusitis patients. These cases meet criteria for histological classification as RDD, including epithelioid histiocytes with abundant eosinophilic cytoplasm and dense lymphoplasmacytic inflammation. They also display essential immunohistochemical findings of RDD, with uniform S100 protein and cyclin D1 positivity and CD1a and CD207 negativity. All cases also show emperipolesis highlighted by S100 protein. A few additional histological features were notable in this case cohort. Abnormal histiocytes were arranged in discrete aggregates with a granuloma-like appearance. Most aggregates were centred in oedematous stroma, although some showed vasculocentric growth without overt vasculitis or the vasculopathy that has been described in RDD at other sites. ^{25, 26} Prominent intermixed eosinophils were also identified in the majority of cases with eosinophilic emperipolesis. These findings appear virtually identical to cases previously reported as RDD-like histiocytes.²³

Our findings also validate the clinical significance of this histological pattern. Patients with histological features of RDD demonstrated severe pan-sinusitis symptoms and had a substantially higher need for repeat surgery compared to chronic sinusitis patients without these findings. Moreover, all previous or subsequent sinus contents specimens from these patients displayed identical histological features, suggesting that the histiocytic proliferation was instrumental to the clinical phenotype. However, there was no locally destructive, multifocal or systemic disease in any chronic sinusitis patients with features of RDD, indicating that this pattern is associated with isolated sinonasal disease. These clinical features also parallel cases reported as RDD-like histiocytes.²³ Notably, these findings were seen in 3.5% of all sinus contents specimens over 2 years, suggesting that this is an uncommon but recurrent phenomenon. This histological pattern also appears to be under-recognised, as all cases identified retrospectively were originally signed out as variations on chronic or allergic sinusitis.

This unique clinical and pathological profile raises questions as to whether these cases are true RDD or a separate phenomenon. A group of nearly identical cases were previously described as RDD-like histiocytes because the overall clinicopathological picture was of chronic sinusitis, rather than RDD.²³ Indeed, several features noted in this series also support separate classification. First, the abnormal histiocytes are quantitatively and qualitatively limited. Some authors have suggested that a diagnosis of RDD should only be made if pathological findings comprise more than 10% of a specimen¹ – a relatively arbitrary percentage that is not yet widely accepted. The lesional histiocytes themselves do not meet this threshold in any cases in our cohort, although surrounding inflammatory milieu and oedema exceed this volume if quantified. Additionally, the histiocyte architecture is more discrete than in conventional RDD, with a nodular, granuloma-like pattern and intervening oedematous stroma rather than diffuse sheets. Furthermore, multiple features point toward a localised and reactive aetiology. These histiocytic proliferations were identified in patients with symptoms restricted to the sinonasal tract. Despite association with recurrent sinusitis, all cases showed benign behaviour with no systemic disease in long-term follow-up. A subset even demonstrated concomitant findings including allergic fungal sinusitis, mycetoma and antrochoanal polyp, similarly pointing to a reactive aetiology. These findings all suggest that pathogenesis of these limited sinonasal cases is different than conventional RDD.

Nevertheless, other findings suggest this phenomenon should be part of the RDD spectrum despite its unique manifestations. Despite the relatively limited volume of histiocytes compared to conventional RDD, they were a prominent feature throughout all cases that starkly differentiated them from other chronic sinusitis specimens. As this impression differed from the previous report of RDD-like histiocytes, which also were seen at higher prevalence,²³ it is possible that we selected more florid cases. Moreover, although no cases had MAP2K1 or KRAS mutations characteristic of conventional RDD, two harboured oncogenic alterations in NF1 and KEAP1. NF1 mutations have been reported in other non-Langerhans histiocytosis, including juvenile xanthogranuloma and Erdheim–Chester disease, and also activate the MAPK/ERK pathway via RAS signalling.^{27, 28} The significance of KEAP1 mutation is less clear, as this gene functions in separate pathways and has not been implicated in histiocytosis.²⁹ Regardless, the presence of oncogenic mutations in cases histologically indistinguishable from the remainder of the cohort suggests that these histiocytic proliferations are not uniformly reactive. Additionally, all cases tested expressed cyclin D1 in the abnormal histiocytes. Cyclin D1 positivity, which is absent in reactive histiocytes, has previously supported MAPK/ERK pathway activation in RDD.¹³ As such, this finding indicates unity with other RDD cases regardless of mutation status. Other authors have proposed that abnormal cytokines produced by lymphoma might drive MAPK/ERK pathway activation in non-neoplastic histiocytes, leading to malignancy-associated RDD.^{14, 30} It is conceivable that the altered cytokine milieu in chronic sinusitis^{31, 32} could spark a similar phenomenon. With diagnostic histological features, MAPK/ERK pathway activation and occasional oncogenic mutations, these histiocytes seem to fit within the RDD spectrum.

Ultimately, given similarities and differences from conventional RDD, we think it prudent at this point to regard this phenomenon as a limited form of RDD that carries minimal risk of systemic disease but high likelihood of recurrent sinus symptoms. While the term RDD-like histiocytes may be appropriate for very focal manifestations, we believe that cases with prominent histiocytes throughout the specimen merit recognition on the RDD spectrum. We currently sign these cases out with a top-line of ‘Limited sinonasal Rosai–Dorfman disease’ and a note indicating that these findings carry an increased risk of recurrent chronic sinusitis but are not associated with systemic disease. In some ways, this localised indolent presentation is analogous to limited histiocytoses in other epithelial sites, such as gastrointestinal Langerhans cell histiocytosis in adults, which has mild localised symptoms but no systemic involvement.³³ It is also analogous to special-site considerations for other haematopoietic lesions such as duodenal-type follicular lymphoma, which is extremely indolent.³⁴ Further research to define the minimum boundaries of RDD may help to address this phenomenon as well as other reactive or secondary RDD manifestations, leaving the door open for more nuanced classification in the future. Regardless of terminology, disproportionately high recurrent sinusitis in patients with these findings suggests that recognition of this phenomenon is clinically relevant for the management of sinus disease.

Conflicts of interest

All authors certify that they have no affiliations with or involvement in any organisation or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.

Ethics approval statement

All procedures performed in this retrospective data analysis involving human participants were in accordance with the ethical standards of the institutional review board (IRB) (Johns Hopkins Medicine IRB 00176183 and UT Southwestern IRB 112017–073), which did not require informed consent.