Amino acid sequence homology between thyroid autoantigens and central nervous system proteins: Implications for the steroid-responsive encephalopathy associated with autoimmune thyroiditis

Highlights • Alpha-enolase, aldehyde reductase-I and dimethylargininase-I are SREAT autoantigens.• Molecular mimicry between thyroid and CNS autoantigens is hypothesized in SREAT.• Homology with TSH-R, Tg and TPO exists for 6, 27 and 47 of 46,809 CNS-proteins.• The above homologies are often in epitope-containing parts of thyroid autoantigens.• Most of the above proteins are expressed in CNS regions which are altered in SREAT.

GD association, encephalopathy is very sensitive to corticosteroid therapy, another denomination is steroid-responsive encephalopathy associated with autoimmune thyroiditis (SREAT). SREAT represents a rare complication of autoimmune thyroiditis [2] and may precede it even by years, similar to thyroid eye disease in patients with Graves disease [3][4]. SREAT patients have abnormal electroencephalography and increased concentration of proteins/immunoglobulins G (IgG) in the cerebrospinal fluid, which can be observed in 90% and 80% of patients, respectively, but these findings are not specific of the disease [5]. Serum anti-thyroid antibodies (TAb) are typically elevated in SREAT patients, but their levels do not correlate with either severity or any specific clinical presentation.
Between 2002 and 2008, three autoantigens shared by the central nervous system (CNS) and the thyroid, and targeted by autoantibodies specifically present in SREAT patients, were identified: alpha-enolase, dimethylargininase-I (DDAHI) and aldehyde reductase-I (AKRIAI) [6][7][8]. This discovery led to the idea that autoimmunity against autoantigens common to CNS and thyroid could be one of the pathogenetic mechanisms of SREAT, in addition to the action of antithyroid autoantibodies on Tg, TPO and TSH-R expressed in the central nervous system [9].
In 2003, a paper described one patient with HE and reviewed the HE literature (85 patients who met their inclusion criteria out of "105 patients with brain dysfunction associated with possible Hashimoto thyroiditis") [10]. This paper reported that pathologic findings were available for only three HE patients (one based on necropsy and two based on brain biopsy) [10]. In one patient, autopsy revealed lymphocytic infiltration in the brainstem (including its veins and venules), leptomeninx of the cortex, and cerebellum [11]. In the other two patients, biopsy revealed lymphocytic infiltration of the walls of many small arterioles and venules [12], and perivascular cuffs of lymphocytic cells [10]. Quite interestingly, Chong et al. [10] wrote that it could not be excluded that the high serum levels of TAb found in HE patients were originated by reaction to proteins (viral, bacterial, or toxic) causing brain damage or brain antigens released after injury, but there were no known proteins in the above categories with structural similarity to thyroid autoantigens.
For sake of completeness, we should note that Chong et al. [10] missed three patients. One was a French patient [13], in whom postmortem neuropathology demonstrated nonspecifically activated microglia. The second was a Japanese patient [14], in whom autopsy revealed no evidence of CNS vasculitis or other brain abnormalities. The third was an American patient with a questionable 7-mm area of the left medial frontal cortex at MRI [15]. Biopsy revealed moderate gliosis, some perivascular lymphoid cells and macrophages, scattered microglia in the parenchyma, but not vasculitis or microglial nodules [15].
In subsequent years, postmortem examination in HE patients demonstrated "mild perivascular lymphocytic infiltration throughout the brain and leptomeninges plus diffuse gliosis of gray matter in the cortex, basal ganglia, thalami, hippocampi, and, to a lesser extent, the parenchymal white matter" [16]. Biopsy of other HE patients revealed: [i] "patchy myelin pallor, scant perivascular chronic inflammation, mild gliosis, and microglial activation" [17]; [ii] primary vasculitis of the CNS [18]; [iii]"diffuse gliosis and perivascular lymphocyte infiltration with CD3 + T-cell predominance, … with no signs of a brain tumor" in a patient with a tumor-like lesion of the left caudate nucleus, "suggesting cerebral vasculitis as an underlying etiology" [19]; [iv] non-vasculitic autoimmune inflammatory meningoencephalitis [20]; [v] reactive gliosis, angiogenesis, swollen vascular endothelial cells, mild lymphocyte infiltration (almost exclusively T cells) around small vessels [21].
Molecular mimicry between thyroid autoantigens and other autoantigens was mentioned by several authors as a possible clinically relevant causal mechanism of extrathyroid manifestations of thyroid autoimmunity, including some neurological and pychiatric disorders [22][23][24].
Just very recently, we demonstrated that there is striking local homology between thyroid autoantigens and the three HE/SREAT-autoantigens [25]. Particularly, Tg was homologous to 10 regions of alpha-enolase, 8 regions of AKRIAI, and 5 regions of DDAHI. TPO was homologous to 6 regions of alpha-enolase, 7 regions of AKRIAI, and 3 regions of DDAHI. Finally, TSH-R was homologous to 4 regions of alphaenolase, 5 regions of AKRIAI, and 2 regions of DDAHI. Importantly, in regard to alpha-enolase (the sole of the three HE/SREAT autoantigens for which epitopes have been characterized), a total of 5 regions homologous to Tg, one region homologous to TPO, and one region homologous to TSH-R fell within, or adjacent to, epitopes of the protein.
From the opposite perspective, a total of 4 regions of Tg, 5 of TPO and 2 of TSH-R homologous to alpha-enolase contained epitopes. Epitopes in each of the three thyroid autoantigens were present also in their regions that were homologous to regions of AKRIAI and DDAHI [25]. In brief, we provided some indirect evidence that a number of regions of homologies were relevant for the autoimmunity associated with HE/ SREAT.
We hypothesized that alpha-enolase, AKRIAI and DDAHI might be the classic "tip of the iceberg", viz. we hypothesized that there could be more proteins expressed in the CNS, not necessarily in a CNS-restricted expression mode, which share homology with at least one of the three thyroid autoantigens. Applying the same bioinformatic approach used for alpha-enolase, AKRIAI and DDAHI, we searched for such homologies.

Material and methods
We used our standard procedure, as in previous bioinformatics papers [25][26][27][28][29][30][31]. We retrieved the amino acid sequence of the precursors of the three "classical" human thyroid autoantigens, i.e. TSH-R (accession number P16473), Tg (accession number NP_003226) and TPO (accession number AAA61217) from the Entrez Protein database (https:// www.ncbi.nlm.nih.gov/protein). Next, we probed each of these three autoantigens for amino acid sequence homology with human proteins of the same database whose records contained the term "brain" or "central nervous system". Proteins labeled as "incomplete" or "hypothetical" were excluded. We also excluded alpha-enolase, AKRIAI and DDAHI, since they were investigated in our previous paper [25]. The Protein BLAST (Basic Local Alignment Search Tool) software version 2.8.0+ [32] was used to perform the comparison. Analysis was made with the standard parameters of the program, and only results with E < 10 were considered. Finally, the records of the proteins identified were manually reviewed, to exclude those not expressed in the CNS (the presence of the terms "brain" and/or "central nervous system" in the record was sometimes incidental, not related to the actual localization of the protein).
As also done previously [25][26][27][28][29][30][31], we verified the immunological relevance of the homologies selected, checking for their possible overlap (s) with known epitopes of TSH-R, Tg and TPO [26][27][28][29][30][31][33][34][35][36]. To strengthen the immunological relevance of the homologies that we found, we searched the literature for the presence of serum autoantibodies against each of the thyroid autoantigen-homologous proteins in autoimmune diseases, including thyroid autoimmune diseases. To this aim, we searched in the PubMed database using the search string "(autoanti* OR autoimm* OR autoreact*) AND" followed by the name of each protein, and manually revised the results to select only relevant original articles.
To quickly know (i) which areas of the CNS express each of the proteins that we found to be thyroid autoantigen-homologous, and (ii) whether the thyroid gland also expressed these proteins, we probed the Expression Atlas (https://www.ebi.ac.uk/gxa/home) [37]. Table 1, Table 2 and Table 3 list which of the 46,809 CNS-expressed  Table 1 Homologies between TSH-R and proteins from brain or central nervous system.  Table 2 as an example for describing the other two Tables (Table 1 and Table 3), there are proteins with a single segment of homology each, such as butyrylcholinesterase (aa 9-527 matching aa 2204-2722 of Tg), and other proteins with multiple segments of homology (which are listed from the most N-terminal to the most C-terminal position). Examples of this multiplicity are the nine segments of SPARC-1/SMOC-1 that are homologous to Tg. Close inspection of these nine segments (Table 2) shows that they fall within the long region 54-340 of SPARC-1/SMOC-1, which matches a discontinuous and much longer region of Tg comprised between aa 34 and 1212. Indeed, two long stretches of Tg (aa 359-608 and 662-880) did not match any segment of aa 54-340 of SPARC-1/SMOC-1. The extent of amino acid identity with Tg segments ranges from 22% (KIAA1366 protein) to 50% (aa 333-372 of testican-1 and aa 333-374 of testican-2), and overall homology from 36% (aa 636-752 of signal peptide, CUB and EGF-like domain-containing protein 1) to 67% (aa 333-372 of testican-1). Of interest, the group of Tg segments homologous to CNS-expressed proteins ( Table 2) and the group of Tg segments homologous to alphaenolase, AKRIAI or DDAHI [25] showed several overlaps. In detail, Tg segments of the first group fully contained a Tg segment of the second group in 59 cases (with some multiple matches) and were fully contained in a Tg segment of the second group in 3 cases, while a partial overlap of more than 10 residues was observed in 10 cases.
In the case of TSH-R (Table 1), with the only exception of LGR4, all proteins (which were cell receptors, except chondroadherin) had a single segment of homology with the thyroid autoantigen. Identity with TSH-R ranges from 19% (probable G-protein coupled receptor 34, alpha-1B adrenergic receptor and bombesin receptor subtype-3) to 55% (G protein-coupled receptor), and overall homology from 36% (Melanopsin/Opsin-4 and 5-hydroxytryptamine receptor 7) to 75 % (G proteincoupled receptor). TSH-R segments homologous to CNS-expressed proteins (Table 1) fully contained a Tg segment homologous to alphaenolase, AKRIAI or DDAHI [25] in 122 cases (with many multiple matches), while the partial overlaps of more than 10 residues were five.  The position of sequence homology within given domains of the three thyroid autoantigens can be appreciated in Figs. 2-4. Of the 46 proteins homologous to TSH-R (Fig. 2), only the first 6 (LGR4, LGR5, relaxin receptor 1, relaxin receptor 2, chondroadherin and LINGO2) match the whole length of TSH-R, while the others match the serpentine domain, most frequently for its whole length. A few proteins match the C-terminus of the extracellular domain, and a few match the intracellular domain. With the single exception of G protein-coupled receptor and C-C chemokine receptor type 7 (whose homology with TSH-R starts at aa 670 and 672, respectively, of the thyroid autoantigen), all other 44 proteins matched TSH-R regions containing at least one epitope (Fig. 2).
Concerning Tg (Fig. 3), of the 27 homologous proteins, 7 matched a long N-terminal region, 4 a very short central region, and the remaining 11 the acethylcolinesterase-like domain at the C-terminus of Tg. Noteworthy, all 27 proteins matched regions of Tg containing at least one epitope, including the short Tg segment 1470-1494 matched by Ephrin type-B receptor 6, since the aa sequence 1473-1526 of Tg is epitopic (Fig. 3).

The thyroid-autoantigen-homologous proteins are expressed in given areas of the CNS, and almost all of them are expressed in the thyroid
Supplementary Tables 1-3 summarize information from the Expression Atlas (https://www.ebi.ac.uk/gxa/home) [37] about the expression of the proteins homologous to TSH-R, Tg and TPO, respectively, in different areas of the CNS and in the thyroid. 620-676 of TSH-R are homologous to segments 89-141, 258-264, 14-22, and 268-325 of AKRIAI, respectively. Segments 141-148 and 263-292 of TSH-R are homologous to segments 242-249 and 258-283 of DDAHI, respectively [25].

Table 2
Homologies between thyroglobulin (Tg) and proteins from brain or central nervous system. The same Supplementary Tables also show, highlighted in gray, which areas of CNS expressing thyroid autoantigen-homologous proteins were found to show abnormalities at diagnostic neuroimaging in patients with HE/SREAT (references about these data are available upon request). Of these areas, those with the highest number of TSH-Rhomologous proteins expressed were frontal lobe (n = 36), cerebral cortex (n = 34), frontal cortex and temporal lobe (n = 33 each); those with the highest number of Tg-homologous proteins expressed were frontal lobe and temporal lobe (n = 26 each) followed by brain, cerebral cortex and frontal cortex (n = 25 each); those with the highest number of TPO-homologous proteins expressed were brain (n = 47), temporal lobe (n = 45), cerebral cortex, frontal cortex and frontal lobe (n = 44 each).
For a few proteins homologous to TSH-R (free fatty acid receptor 3, trace amine-associated receptor 6, olfactory receptor 2A14, trace amineassociated receptor 8, olfactory receptor 2 J3), the Expression Atlas provides no details on which CNS areas express these proteins. Also for a few proteins, the same Atlas provides no details as to whether the thyroid gland expresses these proteins, or reports that their expression is below the cutoff value considered (Supplementary Tables 1-3). Supplementary Table 4 shows data reported in the Expression Atlas about the expression of the three currently known autoantigens of HE/SREAT in the thyroid and in the brain/CNS.

Autoantibodies against the thyroid-autoantigen-homologous CNSexpressed proteins have been detected in a number of autoimmune diseases
As explained under Materials and Methods, we probed the literature for articles on the presence of serum autoantibodies against each of the thyroid autoantigen-homologous proteins in autoimmune diseases, including thyroid autoimmune diseases, by performing a PubMed search with the string "(autoanti* OR autoimm* OR autoreact*) AND" followed by the name of each protein and manually selecting relevant original papers . As summarized in Table 4, of the 46 CNS proteins homologous to TSH-R, 5 (11%; LGR4, chondroadherin, alpha-1A adrenergic receptor, Mu opioid receptor, and melanin-concentrating hormone receptor 1) were reported to stimulate autoAb, and in the following conditions: CNS demyelinating disease, autoimmune hepatitis, refractory hypertension, psychiatric disorders, chronic fatigue syndrome and vitiligo [38][39][40][41][42][43][44][45][46][47][48][49][50][51]. Of the utmost interest are anti-LGR4 autoAbs, because they were detected also in patients with AIT [30]. Noteworthy is also information available on epitopes of melanin-concentrating hormone receptor 1 [27], with aa 85-98 and 254-260 being major autoantibody epitopes, aa 51-80 and 154-158 being minor autoantibody epitopes, and aa 254-260 being the target of function-blocking antibodies. Thus, the segment 119-393 of melanin-concentrating hormone receptor 1, which we found to be homologous to the segment 424-691 of TSH-R contains epitopes (aa 154-158 and 254-260), as does the homologous TSH-R segment (epitope at aa 441-661).

Table 3
Homologies between thyroid peroxidase (TPO) and proteins from brain or central nervous system. As summarized in Table 6, of the 47 CNS proteins homologous to TPO, 7 (15%; fibrillin-1/asprosin, fibrillin-3, LRP-2, LRP-4, P-selectin/ CD62P/granule membrane protein 140/leukocyte-endothelial cell adhesion molecule 3, and the aforesaid nidogen-1/entactin) were Of interest, it was found that the random peptide TNRRGRGSPGAL, which Dolcino et al. found to be recognized by nearly all sera of patients with psoriatic arthritis, shows amino acid sequence homology and crossreacts with some skin autoantigens, including fibrillin-3 [86]. The 17 TPO-homologous segments of fibrillin-3 contained an epitope in four cases, while their TPO counterparts had four complete and one partial overlap with an epitope. The only match between two epitope- Concerning LRP-2, three TPO-homologous segments were found, of which only one (aa 1388-1428) contained autoepitopes (aa 1397-1412, citrullinated in R12 and R16, and aa 1401-1416, citrullinated in R8 and R12); their TPO counterparts had insignificant or no overlap with known epitopes. Upon describing one patient with AIT and membranous nephropathy, the authors report that low-density lipoprotein receptorrelated protein 2 (megalin) is expressed on thyroid cells in a TSHdependent manner and could be the link between the two diseases [91]. The single local homology found between LRP-4 and TPO involved aa 359-433 of LRP-4, which contain the epitopes 361-376 (citrullinated in R13) and 365-380 (citrullinated in R9), and aa 747-838 of TPO, which contain the epitope 766-775.
A single local homology was found also between P-selectin and TPO, but in this case neither segment (aa 531-621 and 759-843, respectively) contained epitopes (there was only an overlap of few residues in the case of the TPO segment).

Discussion
Expanding our previous data [25], we have provided some evidence for molecular mimicry between thyroid autoantigens and CNSexpressed proteins being a reasonable mechanism for HE/SREAT. First, a limited number of CNS-expressed proteins match relatively short to relatively long sequences of the thyroid autoantigens. Second, the homologous sequences of the three thyroid autoantigens almost always contain at least one epitope. Third, the CNS areas where the thyroidautoantigen homologous proteins are expressed match CNS areas where abnormalities were detected at biopsy/necropsy and/or by neuroimaging in patients with HE/SREAT. Fourth, the literature associated a number of the homologous CNS-expressed proteins with a number of autoimmune disorders (not necessarily CNS-restricted), in which corresponding serum autoAb were detected. TSH-R belongs to the superfamily of the rhodopsin-like G proteincoupled receptors (GPCR), whose ectodomain belongs, in turn, to the family of proteins with leucine-rich repeats (LRR) [94]. Thus, many of the homologies found (Table 1, Fig. 2) were not unexpected.
Interestingly, the TSH-R regions of homology involve its nine LRR repeats, the serpentine domain (aa 414-682, with seven transmembrane helices) and most of the cytoplasmic tail (aa 683-764). Further to the last 20 residues (aa 745-764), two other TSH-R regions are spared by homologies: the signal peptide (first 20 residues) and, upon ignoring LGR4, LGR5, LGR7 and LGR8, the region 255-369. This last region encompasses the LRR9 repeat at 250-271 and most of the hinge region (aa 272-413) with its TSH-R specific sequence at aa 317-366. This segment 317-366 (also called the 50-residue long C-peptide of TSH-R), that is deleted following an intramolecular cleavage, is TSH-R specific because it is absent in the cognate gonadotropin receptors (FSH-R, LH-R) [95].
Assuming that the CNS-expressed TSH-R undergoes the same intramolecular cleavage as the thyrocyte-expressed TSH-R, then the CNS cell will continue to have a cell-attached TSH-R, so called B subunit, with a few extracellular residues distal to the cleaved 317-366 segment, the whole serpentine domain and the intracellular C-terminus. This approximately 400-residue long portion of TSH-R will retain zones of homology with alpha-enolase, AKRIA and several CNS-expressed proteins, as well as a number of epitopes. Most of these epitopes bind TSH-R Ab that inhibit the TSH-R signaling. Thus, it is possible that, whatever the function(s) of TSH-R may be in the CNS, binding to these Ab might inhibit such function(s).
Also not surprising is the presence of esterases in the list of proteins omologous to the C-terminal part of Tg, because the segment starting at aa 2188 is the acetylcholinesterase domain of this thyroid autoantigen. As reported by Veneziani et al. [96] "type I repeats of Tg share varying degrees of homology with a six-residue cysteine motif found in a variety of proteins. These include: …. the cell-adhesion protein nidogen/entactin, the insulin-like growth factor binding protein (IGFBP), … the proteoglycan tes-tican…". Moreover, "The cysteine-rich units of Tg share limited structural analogy with the epidermal growth factor (EGF-) homologous repeats found, in single or multiple copies, in a variety of proteins… The homology between Table 4 Involvement in autoimmune disorders, as resulting from a PubMed search, of the proteins that we found share local homology with thyrotropin receptor (TSH-R).

Table 5
Involvement in autoimmune disorders, as resulting from a PubMed search, of the proteins that we found share local homology with thyroglobulin.

Protein
No. of articles

Table 6
Involvement of the proteins that we found share local homology with thyroperoxidase in autoimmune disorders, as resulting from a PubMed search.

Protein
No. of articles

Citations Results
Fibrillin-1/ Asprosin/ Epididymis secretory sperm binding protein 11 Atanasova MA et al. 2011 [63] Increased anti-fibrillin- Anti-fibrillin-1 autoantibodies in systemic sclerosis patients correlate with specific ethnic groups but not HLA alleles Morse JH et al. 2000 [70] Anti-fibrillin-1 autoantibodies are present in primary pulmonary hypertension, other than in systemic sclerosis, CREST (calcinosis, Raynaud's esophageal dysmotility, sclerodactyly, and telangiectasia) syndrome, mixed connective tissue disease. Lundberg I et al. 2000 [71] Anti-fibrillin-1 autoantibodies are present in CREST (calcinosis, Raynaud's esophageal dysmotility, sclerodactyly, and telangiectasia) syndrome and mixed connective tissue disease. Arnett FC et al. 1999 [72] Anti-fibrillin-1 autoantibodies are present in patients with linear scleroderma or morphea. Tan FK et al. 1999 [73] Anti-fibrillin-1 autoantibodies may be found in patients with systemic sclerosis, CREST (calcinosis, Raynaud's esophageal dysmotility, sclerodactyly, and (continued on next page)  12 Inoue H et al. 2020 [74] Case report of a patient affected by myasthenia gravis and autoimmune polyglandular syndrome type 3, with autoantibodies to both acetylcholine receptor and low-density lipoprotein receptorrelated protein 4 antibody Park KH et al. 2018 [75] Analysis of multiple autoantibodies (including those to low-density lipoprotein receptorrelated protein 4) in patients with myasthenia gravis Ohnari K et al. 2018 [76] Report of a case of myasthenia gravis and amyotrophic lateral sclerosis, with autoantibodies to acetylcholine receptor and low-density lipoprotein receptor-related protein 4 Kruger JM et al. 2017 [77] Report of a case of myasthenia gravis with autoantibodies to lowdensity lipoprotein receptor-related protein 4, but not to acetylcholine receptor nor to musclespecific kinase Ishikawa H et al. 2017 [78] Report of two cases of myasthenia gravis and invasive thymoma, with autoantibodies to acetylcholine receptor and low-density lipoprotein receptor-related protein 4 Li Y et al. 2017 [79] Identification of autoantibodies to lowdensity lipoprotein receptor-related protein 4 in Chinese patients with myasthenia gravis Takahashi H et al. 2016 [80] Report of two cases of amyotrophic lateral sclerosis with autoantibodies to lowdensity lipoprotein receptor-related protein 4, who showed myasthenic symptoms Marino M et al. 2015 [81] Analysis of the presence of autoantibodies to lowdensity lipoprotein receptor-related protein 4 in an Italian cohort of 101 myasthenic patients, 45 healthy blood donors and 40 patients with other neurological diseases Zisimopoulou P et al. 2014 [82] Autoantibodies to lowdensity lipoprotein receptor-related protein 4 were found in 18.7% of about 800 patients with myasthenia gravis from 10 countries Zouvelou V et al. 2013 [83] Report of two cases of myasthenia gravis with autoantibodies to low-   [89] Report of a case of anti-LRP2 nephropathy/antibrush border antibody disease Yu X et al. 2001 [90] Detection of amino acid sequence homology and cross-reactivity between CD69 and low-density lipoprotein receptorrelated protein 2 Illies F et al. 2004 [91] Report of a patient with autoimmune thyroiditis and membranous nephropathy; low-density lipoprotein receptorrelated protein 2 (megalin) is expressed on thyroid cells in a TSH-dependent manner and could be a link between the two diseases P-selectin (CD62P)/ Granule membrane protein 140/Leukocyteendothelial cell 2 Jiang H et al. 1993 [92] Autoantibodies to granule membrane protein 140 were found in 13/46 patients with severe  [53] Entactin is a possible autoantigen of the glomerular basement membrane, which could be involved in some types of human autoimmune glomerulonephritis (non-Goodpasture) Saxena R et al. 1991 [54] Anti-entactin antibodies were found in extracapillary glomerulonephritis patients, although very few. Saxena R et al. 1991 [55] Circulating anti-entactin antibodies are present in specific types of glomerulonephritis, but not in others nor in healthy subjects. Wang J et al. 1994 [56] In the iris of rats with experimental autoimmune uveoretinitis, there is an increase in immunoreactivity of several proteins, including entactin Saxena R et al. 1994 [57] Patients with systemic lupus erythematosus often have anti-entactin antibodies, which are more common in case of severe disease. Saxena R et al. 1995 [58] Two of 40 patients with pulmonary renal syndrome had anti-entactin autoantibodies Li QZ et al. 2005 [59] Autoantibodies to entactin are frequent in patients with lupus but not associated with disease activity Cuadrado E et al. 2015 [60] IgG antibodies to several autoantigens, including entactin, are present in patients with Aicardi-Goutières syndrome, an autoimmune disorder with some similarities to systemic lupus erythematous which particularly targets the cerebral white matter.
Among CNS proteins homologous to TPO, low-density lipoprotein receptor-related protein 4 (LRP4) deserves particular attention. LRP4 has a central role in synaptic development and maintenance, and acts as the muscle receptor for neural agrin, propagating the signal to muscular tyrosin kinase receptors (MuSK) for acetylcholine receptors (AChR) clustering at the neuromuscular junction (NMJ), a peripheral cholinergic synapse between motor neurons and skeletal muscle fibers [97]. LRP4 autoantibodies are detected in some patients with myasthenia gravis (MG), and the inhibition of the LRP4-agrin interaction appears to be responsible, at least in part, for their pathogenicity [98]. In a systematic review, autoimmune thyroid disease was the most frequent of 23 associated autoimmune disorders, occurring in 10% of MG patients [99]. LRP4 antibodies have also been detected in 10-23% of amyotrophic lateral sclerosis (ALS) patients [100][101] As to the NMJ, neurotransmission in the CNS requires precise control of neurotransmitter release from presynaptic terminals and responsiveness of neurotransmitter receptors on postsynaptic membrane, and this process is regulated by glial cells; however, underlying mechanisms are not fully understood. Being expressed in the brain, LRP4 has been implicated in hippocampal synaptic plasticity [102][103]. It has been demonstrated that glutamate release in the hippocampal regions of the brain is impaired in LRP4-defective mice, revealing a critical role of the LRP4-agrin signaling in modulating astrocytic ATP release and synaptic glutamatergic transmission [103][104]. More recently, it has been demonstrated that LRP4 is reduced in the brain of patients with Alzheimer disease (AD), paralleling the reduced levels in an AD mouse model that are associated with exacerbation of cognitive impairment and increases in the amount of amyloid aggregates [105]. Impaired synaptogenesis and altered synaptic transmission at the temporal regions are commonly associated with cognitive disturbances, behavioural alterations, memory reduction. All the above-mentioned disturbances are likewise described in HE/SREAT. Hence, in the light of the of homology between LRP4 and TPO, and considering that presence of LRP4 in temporal areas of the brain has been described, we could speculate on a possible cross reactivity between anti-TPO antibodies and LRP4, explaining the cognitive and behavioural manifestations of HE/SREAT.
The multiform clinical symptomatology of SREAT and its dramatic responsiveness to the corticosteroid therapy (as also supported by the disappearance of abnormalities detected at neuroimaging and electroencephalography, in parallel with a fall both in the serum and CSF levels of the pre-therapy markedly elevated thyroid Ab levels), is better explained by the following scenario. Prior to therapy, elevated levels of thyroid Ab (viz. any of TgAb, TPOAb and TSH-R-Ab) would gain access to the CSF through a damaged blood-brain-barrier. Not only, as we explained previously [25], any of these thyroid Ab can attack CNS cells that express the corresponding autoantigen (viz. any of Tg, TPO, TSH-R) but it/they may attack cells that express one or more of CNS-expressed proteins described here and previously [25]. A requisite for this last attack and associated Ab binding with at least one of these proteins is that the thyroid Ab has/have been elicited by one or more epitopes contained in regions of the thyroid autoantigen that share homology with such CNS protein(s). As shown in this paper, a given region of a given thyroid autoantigen can share homology with only one, a few or several CNS-expressed proteins. Hence, it would be hard to find two HE/ SREAT patients with the same panel of symptoms. Once steroid therapy has knocked-down thyroid Ab levels and thyroid Ab passage into the CNS, then attacks to the above CNS cells would terminate and symptomatology, neuroimaging and electroencephalography abnormalities disappear.
CNS proteins that share a series of homologies with antibodies associated to HE/SREAT have a prevalent distribution in areas correlated with the limbic system and temporal regions in general, as also supported by the literature data on neuroradiological alterations which are prevalent in these regions in HE/SREAT patients (Supplementary  Tables 1-3). This could justify some symptoms, such as confusion, behavior and memory disorders, and epilepsy. In our study, homologies are also detectable among some proteins located in the blood-brain barrier (BBB) (i.e. proteins of the Notch families) and HE/SREAT associated antibodies target, determining a BBB damage and suggesting a possible mechanism of brain aggression by autoantibodies and immunocompetent cells.
short-term plasticity between parallel fiber-Purkinje cell transmissions and defective glutamate release have been postulated as potential pathological mechanisms in some patients with HE/SREAT [106][107].
The spatial position of the homologous segments in relation to cell compartments (extracellular, transmembrane, intracellular) and in the context of the three-dimensional structure of the respective proteins (conformation and chemical characteristics of protein surface, degree of solvent exposure) may be important in the pathogenesis of autoimmune diseases. As a general rule, autoantibodies against extracellular, solvent exposed parts of a molecule are more often directly pathogenetic, while the role of autoimmunity against parts of the molecules that are normally "hidden" from the immune system is less straightforward.
Some authors compared autoimmune conditions characterized by intracellular and extracellular target autoantigens, pointing out differences and possible implications of this difference in clinical, monitoring, diagnostic and therapeutic terms [108]. By analogy, these considerations could be applied to autoimmunity against exposed and hidden parts of molecules, and this aspect, currently unexplored, could be an intriguing line of future research in the field of SREAT.
In conclusion, we support our idea of HE/SREAT being ignited by thyroid autoantigens that, after having gained access to CS, bind to Tg, TPO and TSH-R expressed in cells of the CNS [25] forming immune complexes. In addition to this mechanism, it is well possible that TgAb, TPOAb, TSH-R-Ab may cross-react with CNS-expressed proteins that share local homology with the corresponding thyroid autoantigens. Depending on the prevalent thyroid Ab that forms the immune complex, the homologous protein(s) that cross-react(s), the CNS area(s) and cell (s) expressing such homologous protein(s) and the resulting impairment of its/their function(s), given symptoms will appear, thus explaining the notoriously multiform clinical presentation and neuroradiological abnormalities of HE/SREAT. If one admits that pathogenicity ensues from TgAb, TPOAb, TSH-R-Ab that gain access to the CNS and the necessity of such Ab to be directed against epitopes that are shared with the corresponding CNS-expressed homologous protein(s), then the probability of occurrence of such epitope requisite would be relatively rare. This rarity fits with the knowledge of HE/SREAT being a rare event compared with the high frequency of HT, the most prevalent autoimmune disease.
At the very minimum, we believe that our data will prompt a number of investigations to directly prove the involvement in HE/SREAT of at least some of the CNS-proteins having homology with the thyroid autoantigens. For instance, one straighforward implication is to check serum thyroid Ab (any of TSH-RAb, TgAb, TPOAb) detected in patients with SREAT for cross-reactivity with the corresponding homogous CNSproteins (TSH-R-homologous, Tg-homologous, TPO-homologous). Another straightforward translational implication of our data is to characterize epitopically serum thyroid autoantibodies in patients with HT and GD (both TgAb and TPOAb in HT, and at least TSH-RAb in GD). If any of these serum Ab recognize epitopes of the corresponding thyroid autoantigen that fall in regions sharing homology with any of the known HT/SREAT autoantigens (alpha-enolase, AKRIA, DDAHI) and/or any of the CNS-expressed proteins we report here, once it/they has/have been proved as autoantigens associated with HT/SREAT, then HT or GD patients can be sorted out in terms of risk for HT/SREAT.