Genetics of Neurosarcoidosis

Abstract

Sarcoidosis is a systemic, inflammatory, granulomatous disease characterized by great variability in organ involvement, clinical course, and severity. While pulmonary manifestations are almost universal, the central and peripheral nervous systems can also be affected. Neurosarcoidosis occurs in ~5–15% of cases and is among the manifestations with the highest morbidity and mortality. It is known that sarcoidosis has genetic underpinnings and while multiple studies aimed at identifying associations to sarcoidosis susceptibility and prognosis, very few studies have focused on neurosarcoidosis. This review summarizes the genetic studies to date, compares and contrasts those findings with other genetic effects in sarcoidosis, and offers ideas for moving the field forward.

1.1. Introduction

Sarcoidosis is a multisystem disorder characterized by granulomatous inflammation in one or more affected organs (Iannuzzi et al., 2007, Judson et al., 2012, Baughman et al., 2001). Though the etiology of sarcoidosis remains elusive, various studies have implicated both environmental and genetic components. The prevailing view is that exposure of a genetically susceptible host to one or more environmental triggers results in the chronic granulomatous inflammation that characterizes the disease (ACCESS Research Group 1999, Newman et al., 2004, Valeyre et al., 2014).

Clinically, there is great variability in the range of organs affected and the severity of sarcoidosis. The vast majority, more than 90% of patients, have lung involvement to varying degrees. Some of the more debilitating sub-phenotypes (i.e. central nervous system (CNS) and/or cardiac sarcoidosis) while less common, are still not rare, are likely more frequent than appreciated, and may even be lethal. The prevalence of nervous system involvement has been reported as 5–15% (Stern et al., 1985, Burns, 2003), but multiple case series suggest it is more frequent and one study reports it to be the second most prevalent inflammatory brain disorder after multiple sclerosis (MS) (Street et al., 2014). Tragically, despite being one of the most severe and underdiagnosed forms of sarcoidosis, neurosarcoidosis has received little attention in scientific research (Clifford, 2015). This is due to multiple factors including, but not limited to, a previous lack of consensus diagnostic criteria (something remedied only recently) (Stern et al., 2018), the difficulty in confirming the diagnosis, insufficient candidates for research studies, and the lack of awareness about the disease in the medical and general community. This lack of research in neurosarcoidosis is even more glaring in the field of genetics, despite multiple genetic linkage and genome-wide association studies (GWAS) confirming the role of genetics in susceptibility to sarcoidosis and to specific sub-phenotypes and organ manifestations (Calender et al., 2020, Spagnolo and Maier, 2021, Rybicki et al., 2007, Rybicki et al., 2001a, Adrianto et al., 2012, Iannuzzi et al., 2005, Hofmann et al., 2011, Hofmann et al., 2008, Garman et al., 2020b, Rybicki et al., 2001b, McGrath et al., 2000).

Complicating the study of the mechanisms of sarcoidosis in general and neurosarcoidosis, in particular, is the role of environmental influences that interact with or confound genetic effects. Notably, there are profound differences in the clinical course and genetic effects observed in people of different ancestries, which in turn are likely the result of shared genetic effects compounded by common environment and cultural practices of specific groups of people.(Baughman et al., 2016, Baughman et al., 2001, Rybicki et al., 2001a, Rybicki et al., 1997). Interactions between genes and the environment can lead to variability of allelic penetrance and expression, are at play in most complex diseases (Hunter, 2005, Khoury and Wacholder, 2009), and evidence supporting their role in sarcoidosis are plentiful. As examples, sarcoidosis shows similarity with other antigen-induced granulomatous or infectious diseases, (Celada et al., 2015, Esteves et al., 2016, Ronsmans et al., 2021); has an elevated prevalence in specific geographic areas, (ACCESS Research Group 1999, Newman et al., 2004, Valeyre et al., 2014, Parkes et al., 1985, Hills et al., 1987), and clusters among specific occupational exposures (Newman and Newman, 2012, Ronsmans et al., 2021, Jajosky, 1998, Smith et al., 2009, Webber et al., 2017). Gene-by-environment studies of Löfgren’s and sarcoidosis, have begun to dissect such interactions(Rivera et al., 2019, Garman et al., 2020a, Chen et al., 2019, Li et al., 2014), but unsurprisingly none have been done in neurosarcoidosis.

The primary goal of this descriptive, non-systematic review is to draw attention to the sparsity of studies analyzing the molecular mechanisms of neurosarcoidosis, thereby providing a summary of the existing neurosarcoidosis genetic studies and highlighting areas of promise in future research. Our hope is that by comparing and contrasting genetic findings in sarcoidosis susceptibility, progression, and organ-specific involvement, we can briefly postulate mechanisms for the interplay of these effects, suggest strategies to overcome the current bottlenecks in the field, and offer potential future directions of research.

1.2. HLA and neurosarcoidosis

The Major Histocompatibility Complex (MHC), which includes the Human Leukocyte Antigen (HLA) region, is associated with multiple inflammatory, infectious, neoplastic, and autoimmune disorders. However, the very complex structure of this region (6p21.3) has hindered progress in fine mapping and causal variant identification (Celada et al., 2015, Dendrou et al., 2018). Central to this complexity is the high gene density, extreme sequence diversity, recombination hotspots, and broad linkage disequilibrium (LD) (Petersdorf and O’HUigin, 2019, Lokki and Paakkanen, 2019), which varies greatly by ancestry (de Bakker et al., 2006, Goodin et al., 2018).

These barriers to pinpointing causal variants have been highlighted in recent reviews of HLA associations in sarcoidosis and its sub-phenotypes (Fingerlin et al., 2015, Malkova et al., 2020). The review by Fingerlin nicely showcases the variability of HLA associations by ancestry, geographic region, and sub-phenotype (Fingerlin et al., 2015). For example, people of European ancestry in the U.S. show similar patterns of HLA allelic enrichment in sarcoidosis as those described in Europe; in contrast, American populations of non-European ancestry have distinct HLA enrichment patterns. Numerous examples of differential association not only by ancestry but also by prognosis and organ involvement have also been shown (Adrianto et al., 2012, Grunewald et al., 2010, Grunewald and Eklund, 2009, Levin et al., 2015).

The study of the role of HLA in neurosarcoidosis has been modest (Table 1). Nevertheless, there is evidence of some effects unique to CNS involvement and others held in common with sarcoidosis susceptibility or other organ involvement. For example, Sato et al. identified differential genetic associations across subphenotypes of the disease and the ethnic background of the studied population. After detailed and consistent phenotyping of three sarcoidosis cohorts from Japan, the Netherlands, and the United Kingdom, they noted that in the Japanese, HLA-DRB1*08:03 was associated with neurosarcoidosis, but not in the Dutch and British patients. These findings suggest that genetic risk factors for neurosarcoidosis, as is the case for sarcoidosis as a whole, may also be ancestry (coupled with shared environment) specific (Sato et al., 2010).

Table 1: Genetic Associations of Sarcoidosis with Involvement of the Central and Peripheral Nervous System.

HLA and non-HLA associations with neurosarcoidosis, populations affected, and other neurological diseases that share the same risk alleles.

Genetic Factor	Sarcoidosis Sub-phenotype†	Population	OR*	Other Neurological Disease	Population
HLA-DRB1*08:03	Neurosarcoidosisa (CNS)	Japanese	2.0
HLA-DRB1*04	Heerfordt syndromeb	Swedish		Multiple Sclerosisc	Japanese
HLA-DRB1*04:01	Ocular Sarcoidosisd	European- American	4.97
HLA-DQB1*06:02	Neurosarcoidosise (PNS)	Dutch		Multiple Sclerosisf.g	European
				Narcolepsy-Cataplexyh,i	African- American European- American Japanese
ZNF592 rs75652600	Neurosarcoidosisj (CNS)	African- American	4.34	Monogenic ataxia: GAMOS1 (Galloway-Mowat Sx)k***	Lebanese-Druse
chr15:85309284		European- American**	5.36

^a(Sato et al., 2010),

^b(Darlington et al., 2011),

^c(Nakamura et al., 2016),

^d(Garman et al., 2020b),

^e(Voorter et al., 2005),

^f(Zivadinov et al., 2007),

^g(Oksenberg et al., 1993),

^h(Mignot et al., 2001),

ⁱ(Nicolas et al., 2010),

^j(Lareau et al., 2015),

^k(Nicolas et al., 2010).

^†Comparison was made with healthy controls;

^*OR: Odds ratio – shown for the associations with available information;

^**suggestive association;

^***the majority of families diagnosed with GAMOS1 carry homozygous mutations in WDR73, a gene in the vicinity, and likely in linkage disequilibrium with ZNF592.

In a separate study of Dutch patients, the presence of HLA-DQB1*06:02 was associated with severe pulmonary disease and neurosarcoidosis manifested as small-fiber peripheral neuropathy (Voorter et al., 2005). This same allele has been associated with splenomegaly amongst Japanese sarcoidosis patients, who additionally presented with severe pulmonary involvement, disease chronicity, and higher levels of Angiotensin-Converting Enzyme (ACE) (Sato et al., 2007). Interestingly, this apparent association of peripheral neuropathy with severe pulmonary disease in sarcoidosis is the opposite of what is observed in neurosarcoidosis involving the central nervous system (Perez-Alvarez et al., 2019).

Another sarcoidosis sub-phenotype with neurological involvement, Heerfordt’s syndrome (HS), which involves cranial nerve palsy in addition to uveitis and parotid or salivary gland enlargement, has been associated with HLA-DRB1*04 in Swedish patients (Darlington et al., 2011) (Table 1).

The role of HLA as one of the most reported genetic risk factors offers some insight into, at minimum, the importance of adaptive immunity in neurosarcoidosis. Autoimmune neurology continues to be a developing field, and while autoantibodies have not been identified in any manifestation of sarcoidosis, it is not at all uncommon for patients to present with autoimmune co-morbidities or to have a family history of autoimmunity. Similarly, the fact that all of the HLA alleles identified in neurosarcoidosis thus far have also been seen in neuroimmunological disorders also increasingly considered autoimmune, gives credence to this notion. Unfortunately, however, while associations are abundant, mechanisms of HLA association to neurological disease remain poorly defined (Muñiz-Castrillo et al., 2020, Caillat-Zucman, 2017), and the functional effects of neither HLA-DRB nor HLA-DQB have been characterized in the context of any disease (Dendrou et al., 2018).

To more thoroughly investigate and interpret HLA associations with sarcoidosis in general and neurological involvement, in particular, requires functional investigations that have yet to be done. However, more easily accessible option is in silico peptide binding affinity estimation. This method use a consensus approach to identify the amino acid residues that differ between a given pair of HLA alleles and clinical phenotypes, and has been applied successfully in multiple diseases, including sarcoidosis (Wang et al., 2008). For example, an in silico binding assay experiment in patients with resolved versus persistent sarcoidosis highlighted that the multiple HLA alleles interact and that their combined effect plays a greater role in disease outcomes than any of the component single variants. In addition, the analysis pointed to the differential binding affinity to pathogen peptides of different risk alleles for sarcoidosis (Levin et al., 2015). Amongst those pathogens, Mycobacterium tuberculosis figured prominently; interestingly, a recent study found highly similar main and interaction effects at HLA-DRB1, -DQB1, and -DPB1 in participants resistant to tuberculosis and patients with resolving sarcoidosis (Dawkins et al., 2022). These and other functional and informatic investigations explaining the link between specific HLA alleles and neurosarcoidosis are still missing.

1.3. Non-HLA Associations in Neurosarcoidosis

Beyond associations to HLA genes, multiple non-HLA associations to sarcoidosis susceptibility, severity, and organ-specific manifestations have been identified. These findings are derived from linkage studies, candidate gene studies, genome-wide association studies (GWAS), and even whole-genome sequencing of a family with 13 sarcoidosis cases (Rybicki et al., 2007, Rybicki et al., 2001a, Adrianto et al., 2012, Iannuzzi et al., 2005, Hofmann et al., 2011, Hofmann et al., 2008, Garman et al., 2020b, Rybicki et al., 2001b, McGrath et al., 2000, Fritz et al., 2021). None, though, were specific to neurosarcoidosis.

The first GWAS of sarcoidosis conducted in the United States analyzed a family-based discovery cohort of African Americans and two independent replication case-control cohorts, one African American, and the other of European Americans. Previously known associations derived from European GWAS, including HLA-DRA, HLA-DRB5, HLA-DRB1, BTNL2, and ANXA11, were replicated in both ancestries. In African-Americans, a novel sarcoidosis-associated locus, NOTCH4, reached genome-wide significance independently from other associations in the MHC region. Multiple other potential associations were identified that will require further fine-mapping including a promising candidate in 15q25 (Adrianto et al., 2012).

In 2015, Lareau et al. presented the results of a follow-up fine-mapping study of the chromosome 15 candidate variants (Lareau et al., 2015). The phenotype of choice was neurosarcoidosis based on a previous linkage mapping study that pinpointed neurological and lymphatic involvement to this region (Rybicki et al., 2007). In that study, neurosarcoidosis cases were compared to both healthy controls and to non-neurological sarcoidosis. The most significant association was to rs75652600 in ZNF592, a zinc finger protein. This SNP was unique to the African American patients, but additional suggestive SNPs within ZNF592 not in LD with rs75652600 were found to be significant in the European American (EA) cases.

The association with ZNF592 is of particular interest because rare homozygous mutations in ZNF592 have been postulated as the cause of a monogenic variant of autosomal recessive spinocerebellar ataxia with cognitive disability in a Lebanese Druze family (Nicolas et al., 2010). However, follow-up studies cast doubt on whether the causal gene is ZNF592 or the neighboring WDR73 (Vodopiutz et al., 2015) in which homozygous mutations have been reported as a cause of GAMOS1 in multiple families (Jinks et al., 2015, Ben-Omran et al., 2015, Colin et al., 2014). Nevertheless, ZNF592 is widely expressed in the brain, including the cerebellum and cerebellar vermis, with higher expression in the substantia nigra (Nicolas et al., 2010), and has multiple relevant associations with other genes of interest.

One such connection is a potential interaction between ZNF592 and USF1, which has a role in MHC Class II activity (Lareau et al., 2015). In turn, USF1 and STAT1 mediate the activation of CIITA (an MHC Class II transactivator) by IFN-gamma, (Muhlethaler-Mottet et al., 1998) suggesting that USF1 has a defined role in MHC Class II activity. Furthermore, ZNF592 is also part of a multi-protein complex with ZMYND8, which displays chromatin modification patterns in CD4+ T cells and is a marker of cutaneous lymphoma and T-cell lymphoma (Malovannaya et al., 2010). Combined, these findings indicate a role of ZNF592 in T-cell immune responses, particularly CD4+ T Cells and their interaction with MHC Class II activity, both mechanisms implicated in sarcoidosis.

Another potential role for ZNF592 in sarcoidosis could be related to its strong interaction with SMAD9 (Colland et al., 2004, Massaous and Hata, 1997). SMAD9 encodes a downstream modulator of the bone morphogenetic protein (BMP) signaling pathway, which is part of the TGF-beta (TGFB1) superfamily that regulates growth, differentiation, apoptosis, and development (Davis et al., 2008). Interestingly, truncating mutations in SMAD9 have been identified in patients with primary pulmonary hypertension (Shintani et al., 2009, Drake et al., 2011), which albeit secondary, is a finding in some patients with severe pulmonary sarcoidosis. TGF-b itself has also been associated with the severity of pulmonary sarcoidosis. Additionally, SMAD9 and multiple other members of the SMAD family of proteins promote the processing and activation of miR-21 (Davis et al., 2008). miR-21 is one of the most abundant and ubiquitous microRNAs and is deeply involved in the regulation of inflammatory responses, including the resolution of inflammation and down-regulating pro-inflammatory responses, in particular, in macrophages (Sheedy, 2015). However, how these very intriguing processes are linked specifically to neurosarcoidosis remains to be determined (Fig. 1).

Figure 1.

Potential roles of ZNF592 in sarcoidosis and neurosarcoidosis.

Recently, we completed a genome-wide association analysis of the cohort used in Lareau et al. (unpublished data). The analyzed sample had insufficient power to detect novel findings at the genome-wide significance level but yielded additional suggestive associations that can serve as the basis for additional confirmatory analyses. Specifically, using logistic regression as implemented in PLINK in the previously described 83 African-American neurosarcoidosis cases and 1654 race-matched healthy controls, we confirmed genome-wide significance for the effect on 15q25, including ZNF592 as well as identified three novel regions of interest on chromosomes 1, 10, and 12, each with multiple SNPs reaching suggestive significance (p < 1×10⁻⁵) (Fig. 2). These additional suggestive multi-SNP regions, while not as well characterized as ZNF592, contain genes with interesting links to the CNS. For example, F5, the gene contained within the peak on chromosome 1, is a well-known coagulation factor expressed in the brain that has been associated with, among other things, ischemic stroke (Casas et al., 2004). The signal on chromosome 10 falls within a long intergenic non-coding (linc) RNA (RP11–462L8.1). While little is known about this particular lincRNA, the role of these molecules in disease in general, and specifically neurologic disease is widely accepted (Esteller, 2011, Pastori and Wahlestedt, 2012). The region on chromosome 12 contains AMN1, a protein-coding gene (Antagonist of Mitotic Exit Network 1), that is expressed in the brain, cortex, and cerebellum, and, while not previously associated with neurologic disease, has been linked to corneal astigmatism (Yazar et al., 2013). Several other single SNPs that reached suggestive significance in African Americans are in genes previously associated with neurological disease (OLR1, ANO2, CAMTA1, TRAF5, and CAMK1D) and warrant further investigation. Finally, our reanalysis of the Lareau cohort (Lareau et al., 2015) also included a GWAS comparison of 32 EA neurosarcoidosis cases to 2284 healthy EA controls, which yielded very few SNPs that met GWAS threshold of significance. However, three of the top 10 most significant SNPs were within the gene TMEM132D, a gene previously associated with multiple neurological disorders including Joubert syndrome, transient tic disorder, major depressive disorder, and panic disorder GeneCards_RRID:SCR_002773 (Safran et al., 2021).

Figure 2.

Manhattan Plot of GWAS of neurosarcoidosis. The red line indicates genome-wide significance (p<5 × 10⁻⁸) and the blue line is suggestive significance (p<1 × 10⁻⁵). The blue circles correspond to regions containing F5, RP11–462L8.1, and AMN1 on Chromosomes 1, 10, and 12, respectively, and the red circle corresponds to the 15q25 region containing ZNF592.

1.4. Co-occurrence of Neurosarcoidosis and other manifestations

Patterns of association of neurosarcoidosis with other organ manifestations for which associated genes or pathways have been identified could provide candidate genes for follow-up studies of neurological involvement. For example, Schupp et al. used multidimensional correspondence analysis and cluster analysis to stratify patients of European ancestry into distinct sub-groups based on shared organ involvement. One such group is characterized by ocular–cardiac–cutaneous–central nervous system disease involvement (Schupp et al., 2018). Similarly, a large study in Spain replicated the described association of CNS sarcoidosis with extrapulmonary disease, in particular, ocular involvement (Perez-Alvarez et al., 2019).

As noted, there are significant differences in disease severity between Americans of European and African ancestry. Multiple studies show that African Americans tend to have more severe disease measured across multiple indicators including mortality (Mirsaeidi et al., 2015), earlier age of onset (Rybicki et al., 1997), extrapulmonary involvement (Baughman et al., 2001), clinical outcomes (Judson et al., 2003), and hospitalization rate (Foreman et al., 2006). Thus, studies analyzing the patterns of organ involvement in sarcoidosis amongst Europeans are likely not fully replicable in the American population.

Our own, still unpublished, clustering analysis of a cohort of 900 African American and 242 European American patients recruited in the United States supports this notion of distinct clinical subgroups with racial background-driven differences. We performed Multiple Correspondence Analysis (MCA), akin to Principal Component Analysis (PCA), using the FactoMineR package in R version 3.6.3 (2020-02-29) to generate hierarchical clusters using Ward’s method on the first 5 principal MCA dimensions (R Core Team 2020, Le et al., 2008). The clustering in the European Americans replicates the patterns identified by Schupp et al., (Schupp et al., 2018) identifying only one cluster enriched for neurological and ocular involvement. In contrast, the clustering in African Americans is substantially different, with significantly more overlap between clusters than what was observed in the Europeans and, interestingly, enrichment of the neurological and ocular manifestations in separate clusters (Fig. 3).

Figure 3.

Clustering of organ involvement in 242 European American (Panel A) and 900 African American (Panel B) sarcoidosis patients. Sample coordinates represent the first two MCA principal dimensions (Dim1 and Dim2). The proportion of the variance explained by each principal dimension is given in parentheses as a percentage. The colors and shapes represent cluster membership for the various organs involved. The clusters including neurological involvement are highlighted (green circle) and labeled. The primary, distinguishing organ systems in each of the other clusters are as follows: Panel A, 1-muscle and joint, 2-skin, 3-neurological and ocular, 4-renal, 5-cardiac, 6-liver and spleen; Panel B, 1-extra-thoracic lymph nodes, 2-skin, 3-eye, 4-neurological, 5-renal and cardiac, 6-liver and spleen, 7-muscle.

While clinical clustering alone does not give insight into the exact genetic underpinnings of neurosarcoidosis, it is plausible that genetic effects linked to the involvement of specific organs or prognosis also influence susceptibility to neurosarcoidosis. It is noteworthy that of the known genetic associations of neurosarcoidosis, most have also been identified in at least one other study of sarcoidosis severity or organ involvement (Table 1). Specifically, while the ZNF592 effect appears to be specific to neurosarcoidosis, the HLA-DRB1*04 neurosarcoidosis finding in a Swedish cohort has also been associated with other extrapulmonary manifestations in the same population and with ocular sarcoidosis in multiple populations (Garman et al., 2020b). This supports the presence of pleiotropy or shared genetic effects across organs and provides a rationale for studying in neurosarcoidosis the role of genes associated with ocular involvement, such as MAGI1, the most significant candidate gene for ocular sarcoidosis in European Americans (Garman et al., 2020b). Conversely, AMN1, which showed suggestive association in our neurosarcoidosis GWAS, is perhaps also associated to ocular sarcoidosis –given its role in corneal astigmatism (Yazar et al., 2013).

1.5. Summary and Future Directions

As mentioned above and previously reported, research in neurosarcoidosis has been largely neglected and the genetics of neurosarcoidosis even more so. Years without consensus diagnostic criteria certainly played a role in cohorts of patients either not including neurosarcoidosis cases or underdiagnosis of neurological involvement. The creation of a multi-centric registry of harmonized clinical data derived from patients comprehensively and consistently classified would go far in overcoming these current limitations. This would also allow the collection of the number of cases needed to perform the population-based studies that are often difficult with rare diseases. Certainly, the genetic effects reported here need to be replicated in independent cohorts, and, again, having an international or national registry of these patients would allow for the confirmation of genetic effects that is so desperately needed.

Finally, while genetic studies can be incredibly insightful as to the underlying mechanism of disease, translating genetic findings into actionable strategies or improved diagnostic and prognostic tools requires linking genetics to function through transcriptomics and proteomics. Doing so will require the availability of not only serum and plasma, but also circulating cells, cerebrospinal fluid, and granulomatous tissue. By linking what is observable in easily accessible samples like the blood with what is happening within the CNS we can make great strides in the diagnosis and management of the disease.