The link between autoimmunity and cancer has become a topic of unceasing interest over the past decade. Although it is increasingly evident that the risk of cancer is augmented in patients affected by several types of autoimmune diseases (AID), the nature of the interplay between autoimmunity and cancer remains elusive (2–4). One important question here recalls the old egg-and-chicken dilemma: is the autoimmune background in AID a seedbed for cancer development or, alternatively, may cancer cause autoimmunity?

Rheumatic AID such as systemic lupus erythematosus, rheumatoid arthritis, or Sjögren syndrome promote cancer development after several years of chronic inflammation and also exposure to immunosuppressive drugs (2, 5, 6). In the case of lupus for instance, the broadness of cancer type spectrum is striking, including hematological—mostly virus induced—malignancies but also numerous kinds of solid tumors such as vulva, lung, thyroid, and liver cancer (7). Here, the risk of cancer cannot only be ascribed to the sole autoimmune status but also presumably to iatrogenic immunosuppression.

This view is counterbalanced by the examples of paraneoplastic neurological syndromes, systemic sclerosis, and myositis. Paraneoplastic neurological syndromes with anti-neuronal autoantibodies (aAbs) include a most illustrative example, i.e., anti-Hu encephalitis associated with small cell lung carcinoma (SCLC) (8). Anti-Hu aAbs recognize the HuD autoantigen normally restricted to neurons but ectopically expressed on SCLC cells (9). Expression of the immunogenic HuD molecule by SCLC elicits the production of anti-Hu aAb and CD8⁺ cytotoxic T cells, explaining the parallel clinical course of neurological symptoms and SCLC evolution and demonstrating the direct link between cancer and tissue-specific AID (10). Cancer-induced breakage of tolerance can also be caused by tumoral somatic mutations, as recently highlighted by the case of systemic sclerosis associated with anti-polymerase III (POL3) aAbs. The immunogenic peptides generated by POLR3 gene mutations induce a POL3-specific CD4⁺ T cell response with production of specific antibodies that secondarily target wild-type POL3 by epitope spreading (11).

This present review focuses on the forms of myositides, i.e., dermatomyositis (DM) and immune-mediated necrotizing myopathies (IMNMs) which have been identified as associated with cancer and represent a paradigm of cancer-associated AID.

Autoimmune myopathies or myositides constitute a heterogeneous group of severe acquired myopathies. They are characterized clinically by symmetrical proximal muscle weakness, associated or not with systemic features, and histologically by various levels of myofiber necrosis/regeneration and interstitial mononuclear infiltrates. Clinical and histopathological patterns define different diseases: polymyositis, DM, overlap myositis, sporadic inclusion-body myositis, and IMNM (12–17).

Dermatomyositis affects both adults and children among all ethnic groups with an unbalanced 1/2-sex-ratio in favor of women. Its annual incidence varies from 1.9 to 7.7 cases per million inhabitants according to data in the literature with a peak of frequency in 40–60-year-old adults and in 5–14-year-old children (13, 18). The appearance of specific cutaneous manifestations is typical of DM and is among its diagnostic criteria. Cutaneous manifestations typically consist of erythematous scaly papules over the metacarpophalangeal knuckles (Gottron’s papules) (Figure 1A); a symmetrical reddish-violet periorbital edema that predominates on the upper eyelids (heliotrope erythema) but may affect the rest of the face; lupus-like erythema which involves low neck (V sign) (Figure 1B), shoulders (shawl sign), extensor surfaces of the limbs, dorsal side of hands and fingers and scalp; poikiloderma of the upper trunk (Figure 1C); and centripetal flagellate erythema affecting the trunk and or proximal extremities (Figure 1D). Cutaneous manifestations of DM also include non-specific lesions such as (i) vascular lesions, i.e., periungual erythema with telangiectatic capillary loops, nail fold dilated capillaries visible to the naked eye, cuticular hypertrophy (Figures 1E,F), vasculitis, cutaneous necrosis, or Raynaud’s phenomenon, which are more prevalent in the course of juvenile dermatomyositis and (ii) several other dermatological features such as pruritus (present in 30% of DM), photosensitivity, mucinosis, and calcifications, which are more frequent in children rather than in adults, i.e., 30–70 versus 10%. Dermatological particularities of DM have been reported in several ethnic groups. In Afro-Caribbeans, edema of the face is usually predominant whereas in Eurasians, the Wong-type DM which mimicks a pytiriasis rubra pilaris seems to be more frequent (19).

Specific lesions of DM are histologically characterized by an interface dermatitis with basal layer vacuolar changes that are associated in various degrees with hyperkeratosis, epidermal atrophy, basement membrane thickening, upper dermal edema, pigmentary incontinence, mucine deposits, and light perivascular CD4⁺ T lymphocyte infiltrate of the superficial dermis (Figures 1G–I) (20).

Dermatological features usually precede muscle weakness by 3–6 months but may appear several years before. However, muscular manifestations may be absent, defining the amyopathic form of DM with an associated cancer rate, which is theoretically the same as classic DM (21, 22).

Diagnosis of DM is supported by serum creatine kinase (CK) elevation, which mirrors muscle lysis and by electromyographic data, but diagnosis is confirmed by muscle biopsy except for amyopathic DM. Histological examination of muscle tissue (Figure 2) typically shows perifascicular atrophy, necrotic and regenerative muscle fibers, septal and/or perivascular inflammatory cell infiltrate, and endomysial microangiopathy with membranolytic attack complex C5b-9 capillary staining that may also be present in cutaneous lesions. Although these capillary injuries have long been considered as indirect evidence of an initial endothelial target in DM, recent scientific findings demonstrate that capillary lesions might be non-specific in relationship to ischemia-reperfusion injury of perimysial arcade arteries (23). The role for interferons in DM pathogenesis is increasingly evident, since a typical interferon type 1 pathway signature was found in both muscle (24) and skin (25), with a correlation with disease activity (26).

The outcome of DM is variable with benign forms highly responsive to treatment and severe forms associated with cardiac or lung involvement and/or malignancies. In juvenile DM, cancer association is uncommon. Nonetheless, a poor prognosis may be due to the intensity of vasculitis and vascular damage that involves the skin and the digestive system. Calcifications on the areas around joints may also lead to severe functional impairment in children.

Interestingly, several dermatological manifestations of DM such as cutaneous necrosis, skin vasculitis or pruritus, or histological patterns such as leucocytoclastic vasculitis are reported to be associated with cancer whatever the ethnic group (27, 28). Conversely, several features seem to protect against cancer such as Afro-Caribbean ethnicity (29).

Globally, myositides are rare diseases and epidemiologic data remain scarce. Studies are limited by their retrospective character and small population size. Yet, since its first descriptions (30, 31), the association between cancer and myositis has been confirmed by several studies published between 1975 and 2012, reporting a global malignancy rate from 6.7 to 32% (32–35). Importantly, several studies also point to cancer as the main cause of death in cancer-associated myositis patients (34, 36, 37). Yet, it may be difficult to distinguish between cancer on the one hand, and myositis exacerbation and its complications on the other hand, as the actual cause of death. Most observations highlighted the short delay between the onset of myositis and the discovery of cancer. In some cases, cancer even preceded myositis, reinforcing the view that myositis might be the consequence rather than the cause of cancer and even leading to individualizing “cancer-associated myositis” in one classification in 2005 (14). Cancer risk is particularly established in DM. A recent meta-analysis confirmed adult (but not juvenile) DM as a risk-factor of cancer, with a standardized incident ratio (SIR) for occurrence of cancer of 5.5 [4.3–6.7], mostly peaking at 1 year around DM diagnosis (38). A limitation to previous studies, including those reviewed in this meta-analysis, is that the classification of myositis was essentially based on the widely used Bohan and Peter classification which tends to classify IMNM, overlap myositis, and polymyositis in a single category (1). Whereas this classification remains popular for its practical value in DM, progress in the definition of clinical, pathological, and serological patterns has led to newer classifications that are useful for the diagnosis of inclusion-body myositis (12), overlap myositis (14), or IMNM (13). In the light of these different existing classifications for myositis, a new unifying classification would provide a much-awaited tool.

The detection of myositis-specific aAbs (MSA) has proved most helpful in classifying the different forms and predicting the outcome of myositis. Indeed, in clinical practice, MSA define particular clinical ± histological patterns (39). In DM, anti-Mi-2 are classically associated with absence of cancer, sensitivity to treatment and, overall, good prognosis (40–42). Anti-melanoma differentiation antigen 5 (MDA-5) identifies a subgroup of DM patients with interstitial lung disease, necrotizing cutaneous lesions, skin ulcerations, and tender palmar papules while muscle signs are minimal (43–45). Anti-SAE (small ubiquitin-like modifier activating enzyme) has a low prevalence (1–4%) and patients have classic cutaneous signs of DM. Interestingly, a significant but low rate of cancer has been reported in this group (46–48). Anti-TIF1γ (transcriptional intermediary factor 1 gamma) is the leader of cancer-associated aAbs, with a rate of associated malignancy ranging from 60 to 80% of patients (49, 50). Anti-NXP2 (nuclear matrix protein 2) is another biomarker of risk of malignancy in adult patients, with a 30% rate of cancer among DM patients (51, 52).

Besides DM, IMNM is another form of myositis recently suggested to be associated with malignancy (53–55). IMNM may be subdivided in three groups: seronegative, anti-HMGCR (3-hydroxy-3-methylglutaryl-coenzyme-A-reductase), and anti-SRP (signal recognition particle). In contrast to anti-SRP positive patients, one study has recently suggested that, among IMNM, seronegative and anti-HMGCR positive patients have a significantly higher risk of cancer (56).

Immune-mediated necrotizing myopathies are a recently described entity, based on specific histological pattern with poor inflammatory infiltrate and presence of significant necrotizing and regenerative fibers (13, 72). As mentioned above, three IMNM subsets have been identified according to serologic status: anti-SRP, anti-HMGCR, or negative serology. Anti-HMGCR and seronegative IMNMs seem to be associated with a higher risk of cancer, with a SIR score of 2.79 and 8.35, respectively (56). Between 13 and 36% of anti-HMGCR positive patients have an associated cancer (56, 68, 69). A genetic study found that HLA-DRB1*11:01 is associated with a higher risk of anti-HMGCR IMNM in both white American and African American adult populations, whereas HLA-DRB1*07:01 seem to be associated with a risk of anti-HMGCR myositis in small series of children (86, 87).

Anti-HMGCR aAbs were discovered in 2010 in a group of patients who developed myositis after exposure to statins, without resolution by stopping statins (65, 66). Histological features corresponded to IMNM (88–90). While the association with statins was confirmed in recent studies, it involved a wide range between 37 and 94% of patients (56, 91). Thus, statins are not necessarily required to develop anti-HMGR myositis. Clinically, patients present severe and acute muscle weakness, dramatically elevated CK level, but extra-muscular disorders are uncommon (65, 66, 71). Anti-HMGCR aAbs have also been identified in the sera of juvenile myositis, without exposure either to statins or cancer. Clinical features in children may wrongly shift toward muscular dystrophy (92).

Intriguingly, all three aAb targets in cancer-associated myositis are involved to some extent in cancer pathogenesis.

In the model proposed by Joseph and colleagues in systemic sclerosis, the immune response directed against POL3 was initially an antitumor immune response. Oncogenesis results from random and additive mutations on several genes some of which are involved in mitotic checkpoint, DNA repair, or differentiation, under the effect of viruses, oncogenes, or radiation. Later, tumoral cells can over-express, ectopically express, or express mutated forms of distinct proteins. The immune system may recognize these newly synthetized forms as neoantigens, which may ultimately lead to T cell and B cell responses (142, 143).

In many human cancers, TIF1γ is considered as a tumor suppressor by inhibiting the TGF-β pathway. HMGCR also seems to be involved to some extent in oncogenesis through its role in metabolic pathways. Today, NXP2 is not directly known to be involved in cancer but it interacts with the well documented p53 tumor suppressor. Therefore, the targets of these three MSA appear to be proteins involved in cellular pathways that intervene in tumor promotion. Thus, it is reasonable to hypothesize that somatic mutations of TIF1γ, NXP2, or HMGCR genes in tumors may provoke a specific antitumoral immune response which may secondarily extend to the target organs of myositis (muscle, skin) by cross-reactivity and/or a process of epitope spreading (Figure 3). It is tempting to speculate that absence of cancer in some myositis patients may result from an efficacious antitumoral immune response: in this view, myositis might be the immunological price to pay for tumor eradication.

The hypothesis developed herein implies that the abovementioned antitumoral response provokes/sustains myositis. How this response could promote myositis is unclear. Indeed, the three TIF1γ, NXP-2, and HMGCR proteins are ubiquitously expressed, raising the question of how they could be muscle/skin-specific. In this regard, it should be remembered that most if not all MSAs, associated with cancer or not, are specific for ubiquitously expressed, intracellular proteins such as t-RNA synthetases, MDA-5, and signal recognition particles.

Some elements support the direct pathogenicity of aAbs in some forms of myositis. For instance, aAb level correlates with clinical evolution in IMNM (72, 144–146). The pathogenic effect of aAbs is particularly real for anti-HMGCR aAb. Indeed, Arouche-Delaperche and colleagues recently demonstrated that anti-HMGCR aAbs promoted muscle atrophy and impaired regeneration and expression of inflammatory cytokines (147). A similar effect could be expected for anti-TIF1γ and NXP2 aAbs. TIF1γ has been shown to be over-expressed in regenerating muscle (148), which could explain the recognition of TIF1γ in muscle by aAb.

Most myositis clinical trials exclude patients with cancer, limiting the spectrum of our therapeutic knowledge. While beneficial for treating myositis, immunosuppression is of difficult use in the presence of cancer. Reciprocally, myositis alters the patient’s condition and complicates the therapeutic strategy, i.e., administering antitumoral drugs or performing surgery. Guidelines will be required to help manage patients in this context. The hypothesis of myositis triggered by cancer, rather than the opposite, supports a therapeutic strategy of performing an active antitumoral treatment compatible with the patient’s myositis status. Among new cancer therapies, immune checkpoint inhibitors may expose the patient to an exacerbation of autoimmunity that may yield a myositis flare. Since aAbs may be directly involved in myositis pathogenesis, therapeutic trials evaluating plasma exchanges are welcome.

Taken together, available data point to a unicist view of cancer-associated myositis, in which AID may result from an antitumoral response. This response may be triggered by mutations, overexpression, or posttranslational modification of the autoantigen in tumor. Elucidating these mechanisms will provide strong clues to better understand the potential role of cancer as a cause of autoimmunity.

The first version of this review was written by AA and NC. OB revised it critically for important intellectual content. All authors approved the version to be published.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.