Clinical and serologic parallels to APS-I in patients with thymomas, and autoantigen transcripts in their tumors

Abstract

Patients with the autoimmune polyendocrine syndrome type I (APS-I), caused by mutations in the autoimmune regulator (AIRE) gene, and myasthenia gravis (MG) with thymoma, show intriguing but unexplained parallels. They include uncommon manifestations like autoimmune adrenal insufficiency (AI), hypoparathyroidism (HP), and chronic mucocutaneous candidiasis (CMC) plus autoantibodies neutralizing IL-17, IL-22 and type I interferons. Thymopoiesis in the absence of AIRE is implicated in both syndromes. To test whether these parallels extend further, we screened 247 patients with MG and/or thymoma for clinical features and organ-specific autoantibodies characteristic of APS-I patients, and assayed 26 thymoma samples for transcripts for AIRE and 16 peripheral tissue-specific autoantigens (TSAgs) by quantitative PCR. We found APS-I-typical autoantibodies and clinical manifestations, including CMC, AI and asplenia, respectively in 49/121 (40%) and 10/121 (8%) thymoma patients, but clinical features seldom co-occurred with the corresponding autoantibodies. Both were rare in other MG subgroups (N=126). In 38 APS-I patients, by contrast, we observed neither autoantibodies against muscle antigens nor any neuromuscular disorders. Whereas relative transcript levels for AIRE and 7 of 16 TSAgs showed the expected under-expression in thymomas, levels were increased for 4 of the 5 TSAgs most frequently targeted by these patients’ autoAbs. Hence the clinical and serologic parallels to APS-I in patients with thymomas are not explained purely by deficient TSAg transcription in these aberrant AIRE-deficient tumors. We therefore propose additional explanations for the unusual autoimmune biases they provoke. Thymoma patients should be monitored for potentially life-threatening APS-I manifestations such as AI and HP.

Introduction

Much is being learnt about pathogenetic pathways from two human autoimmune syndromes and from the unexpected parallels between them. In autoimmune polyendocrine syndrome type I (APS-I), the autoimmunity against endocrine and ectodermal targets results from recessive mutations in the Autoimmune Regulator (AIRE) gene (1, 2). Expressed mainly in medullary thymic epithelial cells (mTECs), wild-type AIRE is one factor that normally ensures that mTECs ‘promiscuously’ express peripheral tissue-specific autoantigens (TSAgs) that then induce self-tolerance in thymocytes maturing nearby (3-5). According to current hypotheses, potentially autoreactive T-cells escape this negative selection in AIRE-deficient thymi, emigrate and cause autoimmune damage to target tissues (3, 4).

Starting in infants or young children, the typical diagnostic triad of APS-I comprises hypoparathyroidism (HP), autoimmune adrenal insufficiency (AI) and chronic mucocutaneous candidiasis (CMC). Many patients develop other autoimmune manifestations, e.g. premature ovarian insufficiency (POI), vitiligo, alopecia, autoimmune hepatitis, keratitis, enamel dysplasia, and/or intestinal malabsorption (Table I)(4, 5). Phenotypes vary widely, even within families; some patients are first recognized in adulthood (4, 5).

Table I.

APS-I-like manifestations given as number (%) in the different patient groups².

Manifestation	APS-I n=38	Thymoma		LOMG n=63	EOMG n=55
		with MG n=114	no MG n=7
Any APS-I-like manifestation	38 (100)	9 (8)	1 (14)	2 (3)	0 (0)
Hypoparathyroidism	26 (68)
Addison’s disease	25 (66)		1 (14)
Gonadal failure	8 (21)
Autoimmune thyroiditis	5 (13)	2 (2)		2 (3)
Type I diabetes	3 (8)	1 (1)
Alopecia	12 (32)	5 (4)
Vitiligo	6 (16)	2 (2)		1 (2)
Nail pitting	5 (13)	1 (1)
Keratitis	3 (8)
Enamel dysplasia	12 (32)
Malabsorption	8 (21)
Autoimmune hepatitis	3 (8)
Pernicious anemia				2 (3)
Dry eyes/mouth	2 (5)			1 (2)
Squamous carcinoma	1 (3)
Chronic mucocutaneous candidiasis	28 (74)	3 (3)		1 (2)
Hypogammaglobulinemia			1 (14)
Pure red cell aplasia		2 (2)
Asplenia	2 (5)	1 (1)

²See Tables II and Supplemental Table III for more details on the patients.

Numerous autoantibodies (autoAbs) recognize organ-specific autoantigens (autoAgs)(6-10), and often correlate with the clinical manifestations in APS-I patients. For example, autoAbs to NACHT leucine-rich-repeat protein 5 (NALP-5) correlate with HP (6), 21-hydroxylase (21OH) with AI (10) and side-chain cleavage enzyme (SCC) with POI (9). Remarkably, at diagnosis almost 100% of these patients have autoAbs neutralizing type I interferons (IFNs), especially IFN-α2, IFN-α8 and the related IFN-ω (11-13). Moreover, autoAbs against the Th17-mediators IL-17A, IL-17F and/ or IL-22, involved in mucous membrane defenses against Candida albicans, are prevalent (14, 15) and correlate with the CMC (14).

There are intriguing parallels and differences in the autoimmune associations with thymic epithelial neoplasms. Overall, myasthenia gravis (MG) occurs in ≥30% of all patients with thymomas, especially of the thymopoietic types B2 and B3, plus characteristic autoAbs against the acetylcholine receptor (AChR), titin and other muscle antigens (16-19). Other autoimmune features in thymoma patients (with or without MG) are also sharply focused, especially on hemopoietic targets (in ~5%), causing various bone marrow aplasias (20-23) and possibly hypogammaglobulinemia. At diagnosis, ~70% have autoAbs neutralizing type I IFNs (24, 25), similar to those in APS-I, and again with an IgG4 bias (26). Occasional thymoma patients have been reported with autoimmune endocrine (27-29) or ectodermal (30) manifestations, or even CMC plus autoAbs neutralizing IL-17 and IL-22/ deficiencies in Th17 and Th22 cells (14), very similar to those observed in APS-I (4, 5, 14). Whereas autoAbs against AChRs are clearly pathogenic in MG, the organ-specific Abs in APS-I are useful diagnostic markers, and typically directed against intracellular tissue specific autoAgs (TSAgs), as are those against titin and ryanodine receptor in many late-onset MG (LOMG; onset after age 45) and most thymoma patients (16-19).

The most obvious link between these disparate syndromes is that, in nearly all thymomas, the neoplastic TECs fail to express AIRE detectably, implying reduced expression of TSAgs like AChR, insulin and GAD65 (31, 32). The currently prevailing hypothesis is that AIRE-deficient thymi or thymomas fail to express target autoAgs, and consequently generate and export non-tolerant T-cells that eventually cause autoimmune damage in the periphery (3). An alternative explanation suggests biased selection or even active autoimmunization against certain common targets in aberrant thymic tissue (33). To distinguish between these hypotheses, we assessed whether the clinical and serologic similarities between APS-I and thymoma patients extend to APS-I-typical organ-specific autoAbs, and whether these correlate with under-expression of AIRE and TSAg transcripts in their thymomas. These tumors must hold clues to autoimmunizing mechanisms, which are otherwise very hard to study in humans.

Materials and methods

Patients

All samples were taken with informed consent and Ethics Committee approval in each referral center (Bergen, London, Oxford, and Tartu). The demographics of the patients and controls are shown in Supplemental Table I. Over 200 of the 247 patients with MG and/ or thymomas were seen (and many followed for long periods) by a single UK neurology team (34); all clinical information was from hospital and autopsy records. Another 43 were assessed similarly in Bergen. They are grouped into those with:- thymoma (with or without MG), LOMG, early-onset MG (EOMG; onset before age 45) and ocular MG. The 38 APS-I patients were from the Norwegian panel detailed previously (35).

Thymus and thymoma samples

Thymoma samples were snap-frozen as blocks from 26 patients and stored at −80 °C until use (34). Nearly all thymomas were encapsulated and could be clearly separated from any adjacent thymic remnants (n=5), which were often minimal or absent in older or steroid pre-treated cases. At least one remnant showed follicular hyperplasia, and so did ≥38 of the 48 EOMG thymi removed. Most MG thymomas resemble disorganized infant thymic cortex (17, 34), so pediatric thymi seem the most suitable and available controls.

Assays for autoAbs

All samples were tested in the same lab with radioligand binding assays against in vitro transcribed and translated proteins (Promega, Fitchburg, WI) for autoantibodies against 21OH, 17-hydroxylase (17OH), SCC, GAD65, tryptophan hydroxylase 1 (TPH-1), aromatic L-amino acid decarboxylase (AADC), tyrosine hydroxylase (TH), and NALP-5 (7, 13). The threshold for positivity was set as the mean of the indices for all healthy blood donors tested (n=57-150) + 3SD. We assayed autoAbs against thyroid peroxidase (TPO) with Immulite 2000 solid-phase chemiluminescence immunoassays (FDA Clears Siemens, Malvern, PA), against titin by ELISA (DLD diagnostika GmbH, Hamburg, Germany)), and against AChR by RIA (IBL International, Hamburg, Germany)(18). Anti-TH and anti-TPO autoAbs were only analyzed in patients from whom we had thymoma samples.

AIRE mutations

Both AIRE alleles were sequenced using standard protocols and primers as described elsewhere (35).

RNA extraction from thymomas and real-time PCR

Tissue samples were homogenized in Trizol (Thermo Scientific, Waltham, MA ) using AutoMACS with M-tubes (Miltenyi Biotech, Bergisch Gladbach, Germany), followed by RNA extraction according to the manufacturer’s protocol. RNA concentrations were measured with NanoDrop (Thermo Scientific, Waltham, MA ); 5 μg of total RNA was reverse-transcribed using Superscript III (Invitrogen), 10mM dNTP Mix, RiboLock RNase inhibitor and random hexamers (Thermo Scientific, Waltham, MA ). Real time quantitative PCR (qPCR) was performed using Applied Biosystems^® ViiA™ 7 Real-Time PCR System with 384-Well Block (Life Technologies) and Maxima SYBR Green /ROX qPCR Master Mix (Thermo Scientific, Waltham, MA). Every sample was run in 3 parallel reactions in two separate series of experiments; their results were broadly consistent and have been combined. We detected reliable signals for all transcripts tested except NALP-5.

Every transcript signal was expressed as 2^−ΔΔCt (where C_t represents the threshold cycle), and normalized relative to the value for β-actin in the same sample, and then to its (β-actin-normalized) KRT8 value. For Table V and Figs 2 and 3, the resulting AIRE or TSAg values were next expressed relative to that in one control infant thymus. Primers are listed in Supplemental Table II.

Table V.

Tissue-specific autoantigen transcript values in paired thymoma ÷ thymic remnant ⁶

Gene	Patient no.
	P24 non-MG	P17	P25	P26	P15
AIRE	0.02	0.04	0.01	0.02	0.00
21OH	4.87	0.11	0.38	0.22	13.90
17OH	7.87	0.11	0.26	0.88	1.17
SCC	0.22	0.32	0.49	1.44	0.12
AADC	0.00	1.40		0.00	0.04
TPH-1	0.40	0.15	0.19	0.11
HDC	0.02	0.01	0.04	0.37	0.65
TG	0.58	0.11	0.14	0.46
TPO	0.01	0.32	0.02	1.55	0.03
GAD65	0.01	0.06	0.00	38.30
INS	0.31	0.19	0.02	2.31	0.07
IA-2	0.16	0.10	1.54	2.71	2.63
TDRD6	0.07	0.06	0.05	0.17	0.37
H/K ATPase	0.01	0.05	0.01	0.04
SOX9					1.51
AChR-α	0.00	0.35		3.49

⁶Each number is the TSAg value in the patient’s thymoma block divided by the value in the autologous thymic remnant.

Red: transcript values >4.00 (>4x higher in thymomas than in thymic remnants)

Green: transcript values <0.25 (>4x lower in thymomas than in thymic remnants)

Yellow: intermediate transcript values (from 0.26 to 4.00).

FIGURE 2.

Relative AIRE expression in thymoma types and adjacent thymic remnants. A. AIRE expression in blocks from different WHO thymoma types using Keratin-8 (KRT-8) to adjust for epithelial cell content, and a pediatric control sample as calibrator. B. AIRE expression in paired thymomas/ thymic remnants for 5 informative patients. Expression was calculated as for A. * p < 0.05.

FIGURE 3.

Relative transcript signals for APS-I target autoAgs (normalized to KRT-8) and shown as fold change compared to one pediatric control sample. The bars for each group represent the medians; gray squares and black circles, samples with AIRE expression >0.1 or <0.1, respectively. Black horizontal lines above indicate where the thymomas differed significantly from remnants or from control groups (one-way ANOVA (Kruskal-Wallis) with Dunn’s multiple Comparison correction), and the red horizontal lines where they differed significantly from the combined remnants plus controls (Mann-Whitney U tests. ** p < 0.01

Statistics

We evaluated differences in autoAb prevalences between patients and controls using Pearson χ² tests/ program SPSS v. 15, and in transcript values between different groups using non-parametric one-way ANOVA (Kruskal-Wallis) with Dunn’s multiple comparison tests and Mann-Whitney U tests (Fig 3) (GraphPad Prism, La Jolla, CA). For TSAg transcript values, the threshold for significance was set at p= 0.01. Differences between thymoma and thymus remnant expression of AIRE were evaluated using paired t-tests. We also calculated z-scores to show the number of SDs by which each thymoma TSAg signal differed from the corresponding mean of the 5 infant control thymi, whether above it (positive z-scores) or below (with – signs).

Results

APS-I-typical clinical manifestations in patients with MG and/or thymomas

Among the 121 patients with thymomas (with or without MG), 15 have APS-I-typical manifestations, including AI in non-MG patient P2 (Tables I and II). Several of these disorders occurred together, even including asplenia or nail dystrophy (in patients P1 or P3). Both are unusual in autoimmune patients like these who had no genomic AIRE mutations (asterisked in Table II and Supplemental Table III). So is the CMC (plus IL-22 autoAbs) that severely afflicted P8-10. Many of the APS-I-typical manifestations presented long after the thymomas, for example, in 6 of 29 UK cases with intervals >5 years (21%) versus only 5/70 (7%) of patients with shorter intervals (p = 0.051).

Table II.

Thymoma and MG patients with both APS-I-like clinical features and autoantibodies ³

Patient (sex, MG onset-age)	MG status	Thymoma:-		Clinical features (age at onset or †death)				APS-I-type autoAbs						Other autoAbs
		WHO Type1	duration2		21OH	17OH	NALP5	SCC	GAD65	TPH-1	AADC	IFN-αs	IL-17/22
P 1 (F, 40)*	MG	B2, invasive	≥8yr	Urt (42), T (48), As (50), V (56), SLE (59), NMT (56)		++	++					++	−	IL-12, ANA
P 2 (F)	Not MG	B2 (50)	≥14yr	Hypo-γ (35), TC (41), AI, urticaria, TLP (50)					++			−	nd
P 3 (M, 46)*	MG	“malignant” recurred (56)	≥4yr ≥23yr	T1D (43), T, Alo (50), V, Nail dystr3, SPS (65)(†65)	+			+	++	++		++	nd	IL-12
P 4 (F, 41)*	MG	B2		Alo (38)			++					++	−	IL-12
P 5 (M, 28)	MG	B1	~3yr	Alo (5 and 31), myocardial infarct (†42)	++	++						++	−	IL-12
P 6 (F, 50)	MG	B1		Alo (51)						+		++	nd	IL-12
P 7 (F, 64)	MG	B3		Alo (45), RBC aplasia (74)				+	+			−	nd	IL-12±
P 8 (F, 46)*	MG	B3	≥12yr	CMC (58) (†73)	+	++			++	+		++	++	IL-12
P 9 (M, 27)*	MG	B2; metastatic	≥16yr	CMC (44), NMT (50) (†59)							++	++	++
P 10 (F, 35)	MG	B2/ B3	~9.5yr	CMC (44) (†47)							++	++		IL-12 ±
P 11 (M, 74)*	LOMG	None found4		V (60), PA (70), CMC, T, dry eyes, pulm fibr (74)4								++	++	thyroid ++ ANA
P 12 (F, 53)	LOMG	None found		PA (48), T (52), strong family history of T								++	nd	IL-12

³Common APS-I manifestations are marked in bold.

^*No genomic AIRE-mutations;

¹WHO Classification;

²delay between first signs of thymoma and of APS-I-type features;

³noted at autopsy;

⁴lost to follow-up at age 74;

⁺AutoAb positive but index <200;

⁺⁺AutoAb index >200

Abbreviations:- AI, adrenal insufficiency; Alo, alopecia; ANA, anti-nuclear antibodies, As, asplenia; CMC, chronic mucocutaneous candidiasis; hypo-γ, hypogamma-globulinemia; IFNs, autoantibodies to type 1 interferons; Nail dystr, nail dystrophy; NMT; neuromyotonia; PA, pernicious anemia; pulm.fibr, pulmonary fibrosis; RBC aplasia, red blood cell aplasia; SLE, systemic lupus erythematosus; SPS, stiff-man syndrome; T, thyroid autoimmunity; T1D, type 1 diabetes mellitus; Th17, autoantibodies to IL-17A, IL-17F or IL-22; TLP, tongue lichen planus; Urt, Urticaria; V, vitiligo.

Of the 121 MG patients without thymomas at diagnosis, only one with LOMG (P11) showed clear APS-I-typical features, while another (P12) had pernicious anemia and thyroid disease, and a strong family history of thyroid autoimmunity (Table II).

Although this study is retrospective, all patients were assessed similarly and thoroughly by the same experts in each referral center. Thus the EOMG and LOMG groups are ideally matched ‘disease controls’ for the thymoma patients – for their geographic origins as well as the assessors.

APS-I-typical organ-specific autoAbs in patients with MG and/or thymomas

In the thymoma patients – with or without MG – we found significantly higher prevalences of autoAbs against 21OH, 17OH, SCC, TPH-1 and GAD65 than in either the healthy controls or the other MG subgroups (Fig 1, Table III); against TPH-1 and GAD65, prevalences were approximately half of those found in the APS-I patients. Many of the binding signals were in the same range as in APS-I (Fig 1), as also against AADC and NALP-5 in occasional cases. AutoAbs against AADC, TPH and NALP-5 were considered APS-I-specific (7).

FIGURE 1.

APS-I-related autoantibodies in MG subgroups, APS-I patients and controls against the indicated autoAgs. Thresholds for positivity are marked by horizontal lines.

Table III.

Prevalences (%) of APS-I-type organ-specific autoantibodies in the different cohorts 4 a.

Autoantibody (Ab)	APS-I	Thymoma ± MG	LOMG	EOMG	Healthy controls	AI/APS-IIb
21OH	55**	13*	1.6	3.6	3.5	86
17OH	24**	8.8*	0	1.8	0	11
SCC	40**	11**	3.2*	7.3*	0	6.5
AADC	41**	6.2	0	3.6	4.4	2.2
NALP-5	35**	3.5	1.8	1.8	1.6	NA
TPH-1	28**	14**	0	1.8	0	NA
GAD65	29**	15*	4.8	1.8	3.4	21

^{4 a}↓Numbers of sera assayed for the various autoAbs varied slightly; we tested sera from ≥37 patients with APS-I, ≥113 with thymomas ± MG, ≥56 with LOMG, and ≥55 with EOMG, and ≥57 controls. We tried to test the earliest bleeds from each patient, but many with thymomas were already taking immunosuppressive drugs for their MG.

^bData from 426 Norwegian AI/APS-II patients fully characterized by MM Erichsen et al (1). APS II is defined as AI plus thyroid disease and/or type I diabetes.

^*p < 0.05, patient group versus control;

^**p < 0.01, patient group versus control. NA, not analyzed.

1. Erichsen, M. M., K. Lovas, B. Skinningsrud, A. B. Wolff, D. E. Undlien, J. Svartberg, K. J. Fougner, T. J. Berg, J. Bollerslev, B. Mella, J. A. Carlson, H. Erlich, and E. S. Husebye. 2009. Clinical, immunological, and genetic features of autoimmune primary adrenal insufficiency: observations from a Norwegian registry. J Clin Endocrinol Metab 94:4882-4890.

In total, 49/121 thymoma patients (40%) had ≥1 APS-I-related organ-specific autoAbs, and they tended to co-occur with each other (Table IV), though often not with the corresponding clinical manifestations (Table II). For example, the one patient with AI (P2) tested negative against 17OH or 21OH, but positive against GAD-65. She was one of the 7 thymoma patients without MG, of whom 2 others also had GAD65 autoAbs, and one gave low signals vs 17OH and 21OH. Notably, APS-I-type autoAbs were found in more patients with disease durations >5 years (13/25 (52%) than with more recent onset (20/65 (31%), although not quite significantly (p = 0.061).

Table IV.

Percentages of patients without and from 1 to ≥3 APS-I-type organ-specific autoantibodies ⁵

Number of autoantibodies	APS-I n=38	Thymoma		LOMG n= 63	EOMG n= 55
		with MG n=114	no MG n=7
None	8	57***	57	89***	82***
1	29	21	29	11	15
2	11	16	14	0	4
≥ 3	52	6	0	0	0

⁵p < 0.0001 when comparing all thymoma patients (± MG) with MG patients without thymoma (EOMG+LOMG) using χ2 test.

^***p < 0.0001 when comparing all thymoma patients (± MG) with MG patients without thymoma (EOMG+LOMG) using χ2 test.

In the serial samples available from these patients, the autoAbs showed minor variations at different time-points (Supplemental Fig. 1); some of them might reflect changes in immuno-suppressive drug doses, though similar variations occur in APS-I where these drugs are used only rarely (35).The autoAbs in the thymoma patients showed no significant correlations with MG status or thymoma histology (where the proportions of thymocytes/TECs varied greatly). WHO typing was available for 71 MG thymomas; we found ≥1 APS-I-like autoAb in 2 of the 12 (17%) patients with Types A or AB versus 20 of the 56 (36%) with Types B2 and B3 (p = 0.20). Type B1 is the only subtype reported to express AIRE (in ~50%)(32). Among 4 such patients, P5 had high-index autoAbs against 21OH and 17OH (Table II), but, unfortunately, we did not have access to his thymoma.

In the LOMG, EOMG and ocular MG subgroups, by contrast, none of these autoAbs were more common than in the controls, even in P11 who had several APS-I-like manifestations (Table II). Since the above clinical and serologic associations thus appear to be with the thymomas rather than the MG, although we had only 7 non-MG thymoma patients, we group all the thymoma patients together from here on.

MG-specific autoAbs in APS-I patients

None of the APS-I patients had detectable autoAbs against either AChR (0/38 versus 114/114 of the MG/thymoma patients; p < 0.001) or titin (p < 0.001; not shown). Nor did any of them have any obvious MG-like neurologic features.

Transcript levels for AIRE and TSAgs in thymomas and adjacent thymic remnants

To test the hypothesis that these autoAbs correlate with decreased AIRE and TSAg expression, we assayed AIRE and 16 TSAg transcripts in the available thymoma, remnant or control thymus blocks. To adjust for content of any TEC subtype (36, 37), which varies greatly between thymomas (17, 34), we normalized the qPCR by the keratin-8 (KRT8) signals; these correlated strongly, but inversely, with the thymocyte content estimated when the tumors were first processed (34) (not shown). They were broadly consistent in the duplicate blocks available from 5 thymomas and one remnant, which have each therefore been averaged.

Relative AIRE transcript levels were low in almost all thymoma subtypes, but there were large individual differences (Fig 2A). Values were high in one of the two available type B1 thymomas, in line with previous reports (31, 32). As expected, AIRE expression was much lower in the thymomas than in the adjacent autologous thymic remnants in all 5 available pairs (one was non-MG; Fig 2B).

Data from these paired thymomas/thymic remnants also illustrate the variability between different patients and different TSAgs (Table V). TDRD6 and H+/K+ ATPase transcript values were lower in most of the thymomas than in the remnants, as expected. In contrast, the steroidogenic enzyme transcripts mostly showed similar or even higher values in the thymomas.

The 26 thymomas (including 2 non-MGs) showed the most striking variability in TSAg transcripts, even when AIRE expression was very low (black circles in Fig 3). Values for AADC, H+/K+ ATPase and AChR-α clearly followed AIRE’s expression pattern; highest in control thymi, intermediate in remnants, and low in thymomas (Fig 3). We also noted significantly lower values for insulin, HDC and TDRD6 in thymomas (p < 0.01; Kruskal-Wallis test), and for GAD65 when compared with the pooled control and remnant thymi (p < 0.01; Mann-Whitney test) (Fig 3).

Surprisingly, even when AIRE values were very low, we found higher TSAg transcript values per epithelial cell in many thymomas [numbers with z-scores >3 are shown in brackets] than in any of the control thymi (Fig 3) for:- 21OH [10], 17OH [1], TG [3], TPO [5], TH [1], HDC [2], TDRD6 [2], IA-2 [1], SOX9 [11] and TPH-1 [3] (Supplemental Table III). Thus, these TsAgs appear AIRE-independent, despite their frequent recognition by autoAbs in thymoma and APS-I patients (Fig 1, Table III). Transcript values showed no obvious correlations with MG status or thymoma histology.

Correlating APS-I type features and autoAbs with TSAg transcript values in thymomas

For manifestations that could be correlated with TSAg transcripts, there were only 6 informative patients (Supplemental Table III). In the two with alopecia (P6 and P7), TH transcript levels were similar or even higher than in control thymi (z-scores 13.4 and 0.8). In 3 of the 4 patients with autoimmune thyroid disease (P1, P13 and P14), it presented later than their MG and their thymomas. Interestingly, two of them gave elevated transcript values for TPO (z-scores 2.3 and 3.4), the target that correlates best with autoimmune thyroiditis, whereas values were low for both TPO and TG in P15, whose thyroiditis had presented several years earlier.

The autoAbs likewise failed to show consistent negative correlations with TSAg transcript values. In 14/26 patients, we detected a total of 31 autoAbs against informative targets (excluding AChR) (Supplemental Table III). Transcript values gave positive z-scores in 14 of these 31 instances (for 21OH, TPO and SOX9 they were >3 in >5 instances). Even though AIRE values overlapped the controls in P19 and P20, together they had autoAbs to 5 TSAgs, despite positive z-scores for 3 of these TSAgs (and AChR-α). Conversely, AIRE was very low in P21, but she had adrenal autoAbs despite z-scores >2.5 for 17OH and 21OH.

Overall, these findings provide no clear support for general TSAg under-expression in thymomas as the main cause of the associated autoimmunity.

Discussion

This wide-ranging study unexpectedly shows APS-I-typical organ-specific autoAbs in over 40% of thymoma patients, and APS-I-typical clinical manifestations in 8%, but often not in the same individuals. It confirms and extends an earlier observation of ‘thymocopying’ of APS-I development in a patient with a type B thymoma (27). Despite having no AIRE mutations, some thymoma patients developed major APS-I manifestations like AI, CMC or even asplenia, but at much lower frequencies than in APS-I. However, APS-I patients showed no clinical or serologic signs of MG. Surprisingly too, while some of the autoAgs targeted in thymoma patients showed the expected under-expression in their AIRE-deficient tumors, many did not – including several adrenocortical, gonadal and neuro-ectodermal targets, some of which are considered APSI-specific. Thus, our results highlight differences as well as similarities in pathogenesis in these two syndromes, and question current hypotheses that lack of AIRE-regulated TSAg expression in thymomas is the major cause of the unusually biased autoimmunity.

Clinical and serologic parallels implicating thymic aberrations in autoimmunization

The overlapping autoAb reactivities were largely confined to patients with thymomas rather than LOMG or EOMG, even though thymoma and LOMG patients are well known to share several other autoAbs, especially against internal muscle autoAgs (18), type I IFNs and/ or IL-12 (25). Thymoma patients without MG are very hard to collect; though their numbers here are too small to exclude differences in prevalences of APS-I-type autoAbs or manifestations definitively, the present results – and the contrasts with LOMG – suggest that the thymoma alone – rather than the associated MG – is responsible for this serological and clinical overlap. Thymic aberrations are further implicated by the type I IFN-, IL-17- and IL-22-neutralizing autoAbs – not only by their prevalence in both APS-I and thymoma patients but also by their rarity in numerous other autoimmune diseases, even where peripheral type I IFNs, dendritic or Th17/ Th22 cells are involved in pathogenesis (14-15, 25, 39). Conversely, most of the ‘standard autoAbs’ commonly found in sporadic and systemic autoimmune diseases are uncommon in thymoma patients (22, 32, 40, 41).

Thymic aberrations are also implicated in autoimmunization by our observations in patients with the 22q11.2 deletion/Di George Syndrome, who have primary thymus aplasia. About 15% have similar autoAbs against SCC, 17OH and 21OH, again with no signs of AI or POI (42, 43). Interestingly too, MG has been reported in rare patients with hypomorphic RAG1 mutations who also have disorganized AIRE-deficient thymi (44), and/ or IFN-α autoAbs (45), which again implicates aberrant AIRE-deficient thymopoiesis in autoimmunization in these disparate syndromes.

Clinical and immunologic differences between APS-I and thymoma patients

It is widely accepted that the autoimmune endocrine tissue damage in APS-I is T-cell-mediated (46). Whereas autoAbs against AChR are directly pathogenic in MG, those against intracellular targets are valuable as diagnostic markers. In APS-I, the autoAbs against steroidogenic enzymes correlate well with AI and POI, against TH with alopecia (47), and against-TPH-1 with malabsorption (7); curiously, GAD65 autoAbs correlate with malabsorption in APS-I instead of type 1 diabetes (7).

Although the APS-I-typical manifestations appear to be uncoupled from the autoAbs in our thymoma patients, both were probably under-estimated here for several clinical reasons. There are often delays of many years between detection of autoAbs and onset of the corresponding clinical feature in APS-I (35). Similarly, further manifestations appeared ≥15 years after initial presentation in several thymoma patients (eg, P1-P3, P8 and P9; Tables II and Supplemental Table SIII), but they might have been missed in many others where pre-thymomectomy follow-up was much shorter, as possibly in a previous study (where only 28 patients were tested for autoAbs (32)). Asplenia might well have escaped notice since it requires imaging. Moreover, the immunosuppressive therapy often used for their MG may have repressed autoimmune T-cell-mediated tissue destruction, and their glucocorticoid treatment might have compensated for unsuspected AI. There may also be genetic differences in responsiveness to certain TSAgs.

If there is some true uncoupling of autoimmunization of the B-cell and pathogenic T-cell responses in patients with thymomas, it might reflect several immunologic differences from APS-I. Normally, thymic AIRE plays its key roles when the T-cell repertoire is being established in the fetus and infant. Indeed, its deletion in mice after post-natal day 3 does not cause autoimmunity (48); nor does thymectomy, even in children. Whereas APS-I patients have genomic AIRE mutations from conception, the AIRE-deficiency arises only in the tumors (31, 32) and decades later in life in thymoma patients. Second, APS-I patients also lack functional AIRE expression in spleen and lymph nodes, where it could play important roles in maintaining peripheral tolerance (49, 50). Those checkpoints presumably remain intact in patients with thymomas, and may be sufficiently potent to restrain some of the potentially autoaggressive T-cells exported by these tumors, again making the observed parallels seem the more remarkable.

Moreover, most thymomas show additional aberrations not to be expected in APS-I thymi (which are not available for study). These might also contribute to loss of tolerance, as they include absence of HLA-class II on the neoplastic TECs and of muscle-like thymic myoid cells (17, 34). Interestingly, in both thymomas and Aire^−/− mouse thymi, there are few (if any) Hassall’s corpuscles, around which AIRE⁺ TECs are normally frequent and FoxP3⁺ regulatory T cells (Tregs) are positively selected (17, 51-53). There are also changes in Tregs in both syndromes (52, 54-57).

These clinical factors, plus the late onset of AIRE-deficiency localized in aberrant thymoma microenvironments, probably contribute substantially to the apparent differences from APS-I.

TSAg transcripts in thymomas

Currently favored hypotheses assume that minimal or undetectable TSAg expression by the AIRE-deficient neoplastic TECs (31, 32) is the main cause of the associated autoimmunity. While it is much easier to test TSAgs for Aire-dependence in Aire^−/− mouse TECs (3, 58) than in humans, their AIRE-deficient thymomas seem a practical alternative. Although we recognize that their TSAg transcript values alone may not fully reflect protein expression levels, other options were not available to us.

We also recognize that other biologic factors might have affected the AIRE and TSAg signals we detected. For example, thymoma TECs apparently derive from progenitors with combined cortical and medullary markers (34); the normal counterparts of these progenitors are rare, and have scarcely been studied. Since thymoma TECs appear clonal (59), AIRE and TSAg expression might vary between tumors, as they both do between individual mTECs (60, 61). They might vary within thymomas too, though that was not obvious in the available duplicate samples. Expression might also change as thymomas evolve over time/react to changing blood supply or hormonal influences (34), which might explain the correlations we noted with thymoma durations. Some of these issues might be clarified by testing single TEC from thymomas and pediatric thymi.

Nevertheless, some TSAgs were clearly under-expressed in thymomas, notably AChR-α, H+/K+ ATPase, AADC, insulin, GAD65, HDC and TDRD6. These data partially fit with previous reports of AIRE-dependent, but very variable, expression of insulin, AChR-α, IA-2 and H+/K+ ATPase (32, 33, 61, 62).

Notably, AChR-α transcript levels were high in some thymoma patients, implying that under-expression is not the sole cause of their MG. That also seems unlikely because none of our APS-I patients had MG or detectable autoAbs against AChR or titin. Other AChR subunits may be important targets too (63), and so may pre-existing peripheral tolerance to any of them.

In striking contrast, transcript values for 21OH, TPO and SOX9 were higher (with positive z-scores) in 40 - 65% of our thymomas than in the control thymi, even in tumors where AIRE transcripts were almost undetectable. Our data also suggest AIRE-independence of 17OH, SCC, TG, TH and TPH-1. Intriguingly, these include 4 of the 5 TSAgs most frequently recognized by autoAbs in the thymoma group (Fig.1; p < 0.05 in Table III). Likewise, AIRE-independence has been reported for TG, TPO and GAD67 in control TECs (61, 64), and for 17OH and 21OH in AIRE-negative thymomas (32) . Moreover, TG, steroidogenic enzymes and type I IFNs and Th17/Th22 cytokines are expressed by thymic cell types other than mTEC (14, 64, 65), and should therefore be available for negative selection even in the absence of AIRE, again suggesting that additional mechanisms are operating.

Towards a unifying hypothesis

Among APS-I-typical TSAgs, 21OH, 17OH, SCC and TPO stand out because of their apparent AIRE-independence and the significantly increased frequencies of autoAbs against the first three in thymoma patients. Moreover, these autoAbs – and others against IFNs, IL-17s and IL-22 – were among the first to appear in APS-I infants (66). By contrast, only autoAbs against AADC and GAD65 clearly conformed to current thinking. Overall, therefore, our data question the generality of current hypotheses that the similar AIRE-deficiency in APS-I and thymomas predisposes purely by impairing their expression of AIRE-dependent TSAgs/self-tolerance induction. We therefore propose that AIRE-deficiency might additionally create ‘dangerous’/Treg-deficient microenvironments where available AIRE-independent autoAgs bias selection, or even actively autoimmunize (33), as in paraneoplastic syndromes. To us and others (67), that explains these unusual early dominant responses more neatly than current hypotheses, which apply better to those against other TSAgs like GAD65. Once again, key evidence has come from observations in patients.

Our findings also have implications for patient management. They argue for reconsideration of thymectomy in APS-I children, to halt the continuing supply of autoaggressive T-cells. Further, in thymoma patients, subclinical adrenal insufficiency can mimic MG, with muscle weakness and fatigue (68). If not recognized, it could lead to an acute or even fatal Addisonian crisis.

List of Abbreviations

AI

adrenal insufficiency

AADC

aromatic L-amino acid decarboxylase

AChR

acetylcholine receptor

AIRE

Autoimmune Regulator

APS-I

Autoimmune polyendocrine syndrome type I

AutoAb

auto-antibody

autoAg

autoantigen

CMC

chronic mucocutaneous candidiasis

EOMG

early-onset myasthenia gravis (onset before 45 years)

GAD65

glutamic acid decarboxylase 65

HP

hypoparathyroidism

HC

healthy controls

IA-2

insulinoma-associated tyrosine phosphatase-like protein

KRT-8

keratin-8

LOMG

late-onset MG (onset from 45 years and older)

mTEC

medullary thymic epithelial cells

NALP-5

NACHT leucine-rich-repeat protein 5

POI

premature ovarian insufficiency

SCC

side-chain cleavage enzyme

17OH

17-hydroxylase

21OH

21-hydroxylase

TPH-1

tryptophan hydroxylase type 1

TSAgs

tissue-specific autoantigens