Bronchoalveolar lavage characteristics correlate with HLA tag SNPs in patients with Löfgren’s syndrome and other sarcoidosis

Summary

Genetic susceptibility for sarcoidosis and Löfgren’s syndrome (LS) has been associated with prognosis. Human leukocyte antigen (HLA)‐DRB1*03 is over‐represented in LS, and is associated with a good prognosis, whereas HLA‐DRB1*15‐positive patients have a more chronic course of sarcoidosis. These HLA‐DRB1 types can be easily tagged by single nucleotide polymorphisms (SNPs). Our aim was to evaluate the association between these tag SNPs and bronchoalveolar lavage (BAL) characteristics. In 29 patients, both complete HLA‐DRB1* locus genotyping and SNP tagging was performed in parallel. HLA‐DRB1 type was inferred from the presence of *03 tag rs2040410 allele A and referred to as *03. HLA‐DRB1*15 was inferred from the presence of tag SNP rs3135388 allele A and referred to as *15. For BAL analysis, 122 patients with LS and 165 patients with non‐LS sarcoidosis were included. BAL lymphocyte subsets were analyzed by flow cytometry. The presence of tag SNPs completely corresponded with HLA‐DRB1*03/*15 genotypes in all 29 patients in whom both HLA‐DRB1* genotyping and SNP tagging was performed. In all patients together, *03⁺/*15^– patients showed a higher CD4⁺/CD8⁺ ratio than *03^–/*15⁺ (P = 0·004) and *03^–/*15^– (P = 0·001). LS patients with *03⁺/*15^– had a lower BAL lymphocyte count compared to *03^–/*15⁺ patients (P = 0·011). Non‐LS sarcoidosis patients with *03⁺/*15^– patients showed a decreased CD103⁺CD4⁺/CD4⁺ ratio compared to *03^–/*15⁺ patients (P = 0·045) and *03^–/*15^– patients (P = 0·018). We found that HLA‐DRB1*03 and HLA‐DRB1*15 can be approximated by genotyping of tag SNPs and corresponds with the degree of lymphocytosis and cell phenotypes in BAL in both LS and non‐LS sarcoidosis patients.

Introduction

Sarcoidosis is a multi‐system granulomatous disorder with a wide variation in clinical manifestation and disease outcome and is mediated primarily by CD4⁺ T helper (Th) cells. The course and prognosis may correlate with the mode of onset and extent of the disease. Löfgren’s syndrome (LS), first described by Sven Löfgren, is an acute form of sarcoidosis that is associated with a favorable prognosis 1, 2, 3, 4, 5.

Genetic variation in the human leukocyte antigen (in humans, HLA) region has been associated with the clinical course of sarcoidosis patients. HLA‐DRB1*03 allele is associated with a spontaneously resolving course of the disease and is associated with Löfgren’s syndrome 6, 7, 8, 9. In more than 90% of HLA‐DRB1*03‐positive Löfgren patients the disease resolves within 2 years 6.

By contrast, HLA‐DRB1*03‐negative patients commonly have non‐resolving disease. In line with this, non‐Löfgren sarcoidosis patients carrying HLA‐DRB1*15 have an increased risk for a chronic course of the disease 10, 11, 12. Frequencies of different HLA genotypes vary between ethnicities. For example, HLA‐DRB1*03 is extremely uncommon in Japan, whereas in a Dutch cohort DRB1*03 was found in 40% of the sarcoidosis patients 9.

Furthermore, HLA‐DRB1*15 was found to be a risk factor for sarcoidosis in a white population, but not in a black sarcoidosis population 13. HLA‐DRB1 typing can be used for risk stratification in sarcoidosis patients. However, full HLA‐DRB1 typing requires multiple steps using either sequence‐specific oligonucleotides, polymerase chain reaction (PCR) primers or even sequencing, which makes this method laborious and expensive. In the last decade studies have shown that particular HLA‐DRB1 types can be tagged by single nucleotide polymorphisms (SNPs) in linkage disequilibrium with DRB1 genotypes 14. In contrast to full HLA‐DRB1 genotyping, SNP tagging is a simple procedure. Bakker et al. found that SNP rs2040410 allele A and SNP rs3135388 have been associated with HLA‐DRB1*0301 and HLA‐DRB1*1501, respectively. Furthermore, in patients with diabetes type I SNP rs2040410 was found to identify HLA‐DRB1*0301 with great sensitivity and specificity, while SNP rs3135388 has been associated with systemic lupus erythematosus and multiple sclerosis 15. Tagging is less time‐consuming and less expensive compared with complex full HLA analysis. In other systemic diseases, tag SNPs have also been investigated in order to capture HLA genotypes. For example, in a Japanese cohort with patients with type I diabetes tag SNP rs3129888 captured haplotype HLA‐DRB1*0802 with high sensitivity and specificity. To our knowledge, we have performed the first use of these tag SNPs in sarcoidosis. For this reason, we validated the HLA‐DRB1 tag SNPs in patients with LS.

Analysis of bronchoalveolar lavage can be used to support the diagnosis of sarcoidosis by demonstrating increased total cell count, lymphocytosis and an increased CD4⁺/CD8⁺ ratio 16, 17. More recent studies have proved that a decreased CD103⁺CD4⁺/CD4⁺ ratio in the BAL is an additional reliable tool in the diagnostic work‐up of sarcoidosis patients 18, 19.

Previous reports have described characteristics of BAL in different clinical phenotypes of sarcoidosis, particularly LS versus non‐LS sarcoidosis patients 20. However, variability between patients is considerable. A few studies have investigated the role of HLA‐DRB1*03 on BAL outcomes 21, 22.

However, in most of these studies no subanalysis was performed on patients with LS. Furthermore, to our knowledge the influence of HLA‐DRB1*1501 on the BAL outcomes has not been described. This is of particular interest because HLA‐DRB1*1501 associates with a worse prognosis, and clinically it is most important to identify these patients. We investigated the CD103⁺CD4⁺/CD4⁺ ratio in BAL and compared this in different HLA genotypes. This ratio has not yet been described in different HLA genotypes.

Next to validation of the tag SNPs for HLA‐DRB1*03 and ‐DRB1*15, we investigated if these tags correlate with BAL cell phenotypes in patients with LS and non‐LS.

Material and methods

All patients were diagnosed in accordance with the American Thoracic Society/European Respiratory Society/World Association of Sarcoidosis and other Granulomatous Disorders (ATS/ERS/WASOG) consensus statement on sarcoidosis 23. Patients with LS presented with the classic symptoms of acute onset with bilateral hilar lymphadenopathy, fever, erythema nodosum (EN) and/or bilateral ankle arthritis 5.

A total of 126 LS patients were included in our cohort: 122 patients with BAL and HLA typing/tag alleles and four patients with only HLA typing/tag alleles.

In the first part of our study, the association of HLA‐DRB1*03 and *15 with the tag alleles was examined in 29 unrelated Dutch patients with LS. The tagging was confirmed with high‐resolution HLA typing, i.e. HLA‐DRB1*0301 and HLA‐DRB1*1501.

For the second part of the study, BAL and presence of the *03 and *15 tag alleles were analysed in 122 patients with LS (23 patients from the above‐mentioned LS cohort), 165 patients with non‐LS and 53 healthy controls. All included patients visited St Antonius ILD Center of Excellence, a tertiary referral center for interstitial lung disease.

We accepted a maximum duration of 4 months between diagnosis and performing BAL in LS‐patients and non‐LS patients. Data from patients were collected retrospectively from medical charts and the following parameters were recorded: gender, age at diagnosis, scadding (chest X‐ray) stage, corticosteroid use and smoking status. At the time of BAL collection 10 LS patients used corticosteroids, nine oral and one by inhalation. Regarding non‐LS patients, 17 patients used corticosteroids, four oral and 13 by inhalation at the time of BAL.

The study was approved by the Medical research Ethics Committees United (MEC‐U) of the St Antonius Hospital (R05‐08A), and all subjects gave written informed consent.

Genotyping for HLA‐DRB1, SNP tags and bronchoalveolar lavage

Genomic DNA was extracted from peripheral blood (PB) of each individual using standard methods. In 29 patients, HLA‐DRB1 locus was genotyped using PCR reverse sequence‐specific oligonucleotides (SSO#) methodology (LABType^® SSO; One Lambda Inc., Canoga Park, CA, USA).

In all subjects, tag SNPs rs2040410A and rs3135388A were used to capture HLA‐DRB1*0301 and HLA‐DRB1*1501, respectively 15. For rs2040410 genotyping, a restriction fragment length polymorphism (RFLP) assay was performed. Briefly, we amplified a 228 base pairs (bp) PCR product (forward primer: 5′‐ GTCTTTGGCTGGAGGCATTG‐3′; reverse primer: 5′‐GACTCATGGCTTGCCCCATA‐3′) and the product was digested by the restriction enzyme BsrGI (New England Biolabs, Ipswich, MA, USA) for 16 h at 37°C. The products were separated on 2% agarose gel. The band sizes were as follows for each genotype: AA = 228, AG = 49, 179 and 228 and GG = 49 and 179 bp. To identify genotype rs3135388 tag, a custom designed Taqman SNP genotyping assay was performed on an ABI 7500Fast analyser (Applied Biosystems, Foster City, CA, USA), according to standard methodology. From this point forward in our report, the tagging alleles for HLA‐DRB1*03 and HLA‐DRB1*15 are abbreviated as *03 and *15, respectively.

All patients and healthy subjects underwent bronchoscopy and BAL procedure with a flexible bronchoscope, according to the guidelines of the ERS 24, 25, as described previously 18. To determine lymphocyte subsets in peripheral blood and BAL cellular fraction, flow cytometry was performed as described previously 18.

Statistics

spss version 24 and Graphpad prism software version 6.05 were used for the data analysis. Data are expressed as median; upper (maximum) and lower (minimum) values. The non‐parametric Mann–Whitney U‐test and Kruskal–Wallis test were computed to test for differences in medians. The χ² test was used to compare proportions.

Results

In a total of 29 patients, both full HLA‐DRB1 genotypes and tag SNPs rs2040410 and rs3135388 were determined. Characteristics of the patients are shown in Table 1.

Table 1.

Characteristics of patients with LS, non‐LS and healthy controls

		LS (n = 126)	Non‐LS (n = 165)	P	Healthy subjects (n = 53)
Age		34 (±15)	39 (±17)	0·003	22 (±21)
Gender male		48 (38%)	94 (57%)	0·001	29 (55%)
Caucasian race		124 (98%)	152 (93%)	0·025
Tag DRB1*03 positive		89 (70%)	26 (16%)	<0·001	17 (33%)
TagDRB1*15 positive		31 (24%)	48 (29%)	n.s.	10 (19%)
Smoking	Never	80/125 (63%)	87/164 (53%)	n.s.	26/49 (53%)
	Current	20/125 (16%)	34/164 (21%)		19/49 (39%)
	Ex	25/125 (20%)	43/164 (26%)		4/49 (8%)
Scadding stage	0	3/104 (3%)	6/145 (4%)	<0·001
	I	93/104 (89%)	68/145 (47%)
	II	8/104 (8%)	43/145 (30%)
	III	–	24/145 (17%)
	IV	–	4/145 (4%)

Age is median ± interquartile range (IQR); LS = Löfgren’s syndrome; non‐LS = non‐Löfgren’s syndrome. Bronchoalveolar lavage (BAL) was performed in 122 of 126 patients with LS. Furthermore, smoking history was available from 125 patients with LS, 164 non‐LS and 49 healthy subjects. Scadding stage was available from104 patients with LS and 110 patients with non‐LS.

P‐value is based on the comparison between LS and non‐LS.

Table 2 shows the validation of the association between the HLA‐DRB1 genotypes and the tags *03 and *15 in Dutch patients. The tag alleles and the corresponding HLA‐DRB1 alleles overlapped completely. HLA‐DRB1*03 positivity corresponded with presence of the A allele of the tag SNP rs2040410, while HLA‐DRB1*15 positivity corresponded with the A allele of the tag SNP rs3135388.

Table 2.

Validation of association between human leukocyte antigen (HLA)‐DRB1 genotyping and tag single nucleotide polymorphisms (SNPs) rs2040410 and rs3135388

DRB1*03	rs2040410	Allele A	DRB1*15	rs3135388	Allele A	DRB1*03	DRB1*15
	Positive	Negative		Positive	Negative		Positive	Negative
Positive	16	0	Positive	13	0	Positive	4	12
Negative	0	13	Negative	0	16	Negative	9	4

BAL outcomes in all patients

BAL was performed in a total of 122 patients with LS, 165 patients with non‐LS and 53 healthy controls. All BAL samples were obtained and analysed in our hospital. We compared the BAL results of all patients (both LS and non‐LS), subdividing patients on the basis of HLA tags *03 and *15 (Table 3 and Figs. 1 and 2).

Table 3.

Bronchoalveolar lavage (BAL) and peripheral blood (PB) findings of all patients categorized in four human leukocyte antigen (HLA)‐DRB1 genotypes: *03⁺/*15^–, *03^–/*15⁺, *03^–/*15‐, *03⁺/*15⁺ *03 and *15 were typed using tag single nucleotide polymorphisms (SNPs). Median (upper–lower value).

Sarcoidosis patients	*03+/*15–		*03–/*15+		*03–/*15–		*03+/ *15+		P
(LS and non‐LS)	n	Median (upper‐lower value)	n	Median (upper‐lower value)	n	Median (upper‐lower value)	n	Median (upper‐lower value)
Cells/ml	51	19·1 (6·7–75·6)	32	16·9 (3·5–44·0)	71	21·8 (5·6–68·0)	8	17·8 (11·6–28·6)	n.s.
Lymphocytes (%)	100	25·2 (0·0–70·3)	58	31·3 (1·0–73·0)	114	30·0 (1·6–95·1)	15	29·8 (7·7–48·8)	n.s.
Neutrophils (%)	100	1·0 (0·0–15·2)	58	1·0 (0·0–30·2)	114	1·1(0·0–34·9)	15	1·4 (0·2–4·2)	n.s.
CD4+ (%)	82	80·0 (39·0–93·0)	38	74·1 (47·0–93·0)	86	74·5 (28·0–94·0)	14	85·5 (64·0–91·0)	a0·010
									b0·001
									c0·003
									d0·004
CD8+ (%)	82	12·0 (2·0–50·0)	38	17·9 (3·0–41·0)	86	17·0 (4·0–62·0)	14	9·5 (3·0–28·0)	a0·005
									b0·002
									c0·003
									d0·002
CD4+/CD8+ ratio	82	6·6 (0·8–39·0)	38	4·26 (1·3–31·0)	87	4·5 (0·5–22·0)	14	8·9 (2·4–30·0)	a0·004
									b0·001
									c0·003
									d0·002
CD103+CD4+/CD4+ ratio	34	0·050 (0·00–0·52)	21	0·10 (0·01–0·44)	49	0·090 (0·01–0·62)	7	0·040 (0·00–0·61)	b0·024
PB CD4+/CD8 + ratio	76	2·2 (0·4–15·5)	37	1·8 (0·0–9·0)	82	1·62 (0·60–16·3)	14	1·6 (0·6–3·4)	a0·005
									b0·003

LS = Löfgren’s syndrome; non‐LS = non‐Löfgren’s syndrome; PB = peripheral blood.

^a=*03⁺/*15^– versus *03^–/*15⁺.

^b= *03⁺/*15^– versus *03^–/*15^–.

^c= *03⁺/*15⁺ versus *03^–/*15^–.

^d=*03^–/*15⁺ versus *03⁺/*15⁺.

Figure 1.

Overview of patients and healthy controls.

Figure 2.

Human leukocyte antigen (HLA) genotypes in Löfgren’s syndrome (LS), non‐LS and healthy subjects, respectively.

In all patients (LS and non‐LS), no significant differences were found in % of BAL lymphocytes and % of neutrophils between carriers and non‐carriers. Significant differences were observed in % of lymphocyte subsets and ratios (Fig. 3a). The highest BAL CD4⁺/CD8⁺ ratio was found in *03⁺/*15⁺ patients (median ratio = 8·9; 2·36–30·00) (Fig. 3b). We found no statistically significant difference between *03⁺/*15⁺ and *03⁺/*15^–. The BAL CD4⁺/CD8⁺ ratio in *03⁺/*15^– patients (median ratio = 6·6; 0·78–39·00) was significantly higher compared to *03^–/*15⁺ patients (P = 0·004) and *03^–/*15^– patients (P = 0·002). Patients with *03^–/*15⁺ had the lowest BAL CD4⁺/CD8⁺ratio (median ratio = 4·26; 1·29–31·00).

Figure 3.

Bronchoalveolar lavage (BAL) cell percentage of lymphocytes and cell phenotype ratios in all patients stratified per carriership of DRB1*03 and *15 tag. *P ≤ 0·05; **P ≤ 0·01. (a) %Lymphocytes [median ± interquartile range (IQR)]; sensitivity cut‐off value 3·0 18; (b) BAL CD4⁺/CD8⁺ ratio (median ± IQR); (c) CD103⁺CD4⁺/CD4⁺ ratio (median ± IQR) Sensitivity cut‐off value 0·2 18; (d) CD4⁺/CD8⁺ ratio (median ± IQR) peripheral blood.

The median CD103⁺CD4⁺/CD4⁺ ratio was decreased (reference value < 0·20 18) in all groups (Fig. 3c). Patients with *03^–/*15⁺ had a higher CD103⁺CD4⁺/CD4⁺ ratio compared to *03⁺/*15⁺ (P = 0·024). The CD4⁺/CD8⁺ ratio in peripheral blood was highest in the *03⁺/*15^– patients (median ratio = 2·2; 0·38–15·50), which was significantly higher compared to *03^–/*15⁺ patients (P = 0·005) and *03^–/*15^– patients (P = 0·003) (Fig. 3d).

We performed a subanalysis to determine the influence of smoking on the BAL outcomes. Comparable results were demonstrated in non‐smokers; however, in smokers only the BAL CD4⁺/CD8⁺ ratio was significantly higher in *03⁺/*15^– (median ratio = 6·0; 2·7–30·3) compared to *03^–/*15⁺ (median ratio = 1·84; 1·62–2·06), P = 0·022. In addition, no significant differences were found between different HLA genotypes in smokers in the BAL CD103⁺CD4⁺/CD4⁺ and PB CD4⁺/CD8⁺ ratios.

The group of patients using oral corticosteroids was too small to perform a subanalysis, therefore we excluded these patients. In this group (without patients using oral corticosteroids) the results of BAL outcomes in four different HLA‐DRB1 genotypes were comparable to the whole group.

BAL: LS versus non‐LS

Results from BAL analysis of patients with LS are shown in Table 4 and Fig. 4.

Table 4.

Bronchoalveolar lavage (BAL) and peripheral blood (PB) findings of LS and non‐LS categorized in four human leukocyte antigen (HLA)‐DRB1 genotypes: *03⁺/*15^–, *03^–/*15⁺, *03^–/*15^–, *03⁺/*15⁺ *03 and *15 were typed using tag single nucleotide polymorphisms (SNPs). Median (upper–lower value)

LS	*03+/*15–		*03–/*15+		*03– /*15–		*03+ /*15+		P
	n	Median (upper–lower value)	n	Median (upper–lower value)	n	Median (upper–lower value)	n	Median (upper–lower value)
Cells/ml	39	18·5 (7·5–75·6)	4	27·8 (12·0–30·7)	13	23·7 (7·5–53·9)	7	16·9 (11·6–24·0)	n.s.
Lymphocytes (%)	75	23·8 (0·0–70·3)	11	42·0 (11·3–71·8)	22	24·4 (4·1–69·9)	14	26·0 (7·7–48·8)	a0·011
									b0·020
Neutrophils (%)	75	1·1 (0·0–15·2)	11	0·9 (0·0–4·70)	22	1·2(0·0–31·2)	14	1·4 (0·2–4·2)	n.s.
CD4+ (%)	67	81·0 (47·0–93·0)	8	79·5 (73·0–90·0)	18	82·0 (56·0–94·0)	13	86·0 (64·0–91·0)	n.s.
CD8+ (%)	67	11·0 (2·0 36·0)	8	16·9 (4·0–23·0)	18	12·0 (5·0–22·0)	13	9·0 (3·0–16·0)	n.s.
CD4+/CD8+ ratio	67	7·6 (1·3–39·0)	8	4·7 (3·2–21·5)	19	7·0 (3·3–18·0)	13	9·3 (4·3–30·0)	n.s.
CD103+CD4+/CD4+ ratio	23	0·05 (0·0–0·5)	4	0·07 (0·03–0·10)	12	0·06 (0·01–0·20)	6	0·04 (0·00–0·10)	n.s.
PB CD4+/CD8+ ratio	62	2·1 (0·4–6·9)	8	2·1 (0·00–2·3)	16	2·0 (1·2–7·7)	13	2·0 (0·6–3·4)	n.s.
Non‐LS*	n	Median (upper–lower value)	n	Median (upper–lower value)	n	Median (upper–lower value)			n.s.
Cells/ml	12	21·3 (6·7–54·4)	28	16·7 (3·5–44·0)	58	21·6 (5·6–68·0)			n.s.
Lymphocytes (%)	25	27·6 (1·1–64·5)	47	26·5 (1·0–73·0)	92	32·8 (1·6–95·1)			n.s.
Neutrophils (%)	25	0·8 (0·10–9·20)	47	1·0 (0·1–30·2)	92	1·1 (0·0–34·9)			n.s.
CD4+ (%)	15	71·0 (39·0–93·0)	30	73·5 (47·0–93·0)	68	74·0 (28·0–93·0)			n.s.
CD8+ (%)	15	17·0 (3·0–50·0)	30	19·5 (3·0–41·0)	68	20·0 (4·0–62·0)			n.s.
CD4+/CD8+ ratio	15	3·5 (0·78–30·3)	30	3·8 (1·3–31·0)	68	3·8 (0·5–22·0)			n.s.
CD103+CD4+/CD4+	11	0·03 (0·01–0·43)	17	0·11 (0·01–0·44)	37	0·1 (0·01–0·62)			a0·045 c0·018
PB CD4+/CD8+ ratio	14	2·6 (1·3–15·5)	29	1·5 (0·3–9·0)	66	1·5 (0·6–16·3)			a0·019 c0·011

LS = Löfgren’s syndrome; non‐LS = non‐Löfgren’s syndrome; PB = peripheral blood.

^*Our cohort included only one *03⁺/*15 ⁺ non‐LS patient, therefore this group was too small and was excluded from analysis.

^a=*03⁺/*15‐ versus *03^–/*15⁺.

^b= *03^–/*15 ⁺ versus *03⁺/*15⁺.

^c= *03⁺/*15^– versus *03^–/*15^–.

Figure 4.

Bronchoalveolar lavage (BAL) cell percentage of lymphocytes and cell phenotype ratios in Löfgren’s syndrome (LS) stratified per carriership of DRB1*03 and *15 tag. *P ≤ 0·05. (a) %Lymphocytes [median ± interquartile range (IQR)]; sensitivity cut‐off value 3.0 18; (b) BAL CD4⁺/CD8⁺ ratio (median ± IQR); (c) CD103⁺CD4⁺/CD4⁺ ratio (median ± IQR); sensitivity cut‐off value 0·2 18; (d) CD4⁺/CD8⁺ ratio (median ± IQR) peripheral blood.

LS patients with *03⁺/*15^– showed a significantly lower percentage of lymphocytes (median ratio = 23·8; 0·0–70·3) compared to *03^–/*15⁺ patients, who showed the highest percentage of lymphocytes (median ratio = 42·0; 11·3–71·8; P = 0·011) (Fig. 4a).

Patients with *03⁺/*15^– and *03⁺/*15⁺ showed a higher CD4⁺/CD8⁺ ratio in lavages, a higher CD4⁺/CD8⁺ ratio in PB and a lower CD103⁺CD4⁺/CD4⁺ ratio in lavages compared to *03^–/*15⁺ and *03^–/*15^– patients, but this difference was not statistically significant (Fig. 4b–d).

BAL outcomes of non‐LS patients comparing distinct genotypes are shown in Table 4 and Fig. 5.

Figure 5.

Bronchoalveolar lavage (BAL) cell percentage of lymphocytes and cell phenotype ratios [median ± interquartile range (IQR)] in non‐Löfgren’s syndrome (LS) stratified per carriership of DRB1*03 and *15 tag. *P ≤ 0·05. (a) %Lymphocytes (median ± IQR); sensitivity cut‐off value 3·0 18; (b) BAL CD4⁺/CD8⁺ ratio (median ± IQR); (c) CD103⁺CD4⁺/CD4⁺ ratio (median ± IQR); sensitivity cut‐off value 0·2 18; (d) CD4⁺/CD8⁺ ratio (median ± IQR) peripheral blood.

No significant differences were found when comparing the percentages of lymphocytes or neutrophils in the BAL of different HLA‐DRB1 genotypes in non‐LS patients.

Patients with*03⁺/*15^– had a lower CD103⁺CD4⁺/CD4⁺ ratio (median ratio = 0·030; 0·01–0·43) than *03^–/*15⁺ patients (median ratio = 0·11; 0·01–0·44) and *03^–/*15^– patients (median ratio = 0·10; 0·01–0·62), P = 0·045 and P = 0·011, respectively (Fig. 5c). Furthermore, the CD4⁺/CD8⁺ ratio in the peripheral blood of *03⁺/*15^– patients was higher (median ratio = 2·6; 1·3–15·5) than in the *03^–/*15^– patients (median ratio = 1·5; 0·6–16·3; P = 0·011) (Fig. 5d).

Discussion

Most sarcoidosis patients have a good prognosis, although a clear proportion of patients develop chronic and/or progressive disease. It is a clinical challenge to distinguish patients who will develop a chronic course from patients who are more likely to spontaneously resolve disease. Usually, the only way is to monitor all patients on a regular basis to identify the few who develop chronic disease. Detailed analysis of BAL in combination with HLA‐DRB1*03 and *15 typing may aid in the early identification of these patients. We studied whether clinically informative HLA types associate with BAL cell characteristics.

Regarding the goals of our study, we confirmed the association of HLA‐DRB1*03 with the A allele of rs2040410 and ‐DRB1*15 with the A allele of rs3135388 15. The SNPs tagging HLA‐DRB1*03 and HLA‐DRB1*15 have each been used before independently in other diseases, diabetes and multiple sclerosis, respectively 26, 27. To our knowledge, this is the first study that validated a tag SNP with HLA typing in sarcoidosis patients. Several studies have shown that HLA‐DRB1*03‐positive LS patients have a favorable prognosis, with recovery within 2 years in 95% of the cases. In contrast, only half of the HLA‐DRB1*03‐negative LS patients experience a resolving course of disease. Furthermore, more than half of the non‐resolving HLA‐DRB1*03 patients are HLA‐DRB1*15‐positive 6. This association of HLA‐DRB1*15 with chronic disease has been confirmed by others 10, 28.

Due to its complexity, complete HLA typing is not suitable for daily clinical practice. However, the use of SNP tagging is simple and affordable, and allows easy identification of patients with a good prognosis and patients who will develop a chronic course of disease.

Over decades various studies have investigated the characteristics of BAL and its relation to prognosis of disease in different phenotypes of sarcoidosis. Drent et al. demonstrated that patients with an acute presentation of sarcoidosis with arthritis and erythema nodosum had significantly higher proportions of lymphocytes and CD4⁺/CD8⁺ ratio than patients with respiratory and general constitutional symptoms 20. This study and others have implied a possible beneficial role of CD4⁺ T lymphocytes 29, 30.

In the whole group, the CD4⁺/CD8⁺ ratio was higher in *03⁺ patients compared to *03^– patients in our combined cohort. In LS patients a similar difference was observed, which did not reach statistical significance, due probably to the small group of patients in the analyses. In our LS cohort a higher lymphocyte percentage was found in the patients with *03⁺/*15^– versus *03^–/*15⁺.

A few papers have described BAL characteristics comparing HLA‐DRB1*03‐positive and ‐negative sarcoidosis patients. Idali et al. 31 did not observe differences in the BAL lymphocyte percentage between HLA‐DRB1*03‐positive and ‐negative patients. However, they did not make a distinction between LS and non‐LS patients, nor did they include other HLA genotypes than HLA‐DRB1*03. In line with their findings, in our combined cohort there are also no differences in lymphocyte percentages between *03⁺ and *03^– patients. A striking finding is that in LS patients the *03^–/*15⁺ patients had significantly higher lymphocyte percentages compared to *03⁺/*15^– patients.

In a cohort of 118 sarcoidosis patients, Planck et al. 22 also observed a decreased lymphocyte percentage and an increased CD4⁺/CD8⁺ ratio in BAL from HLA‐DRB1*03‐positive patients compared to those negative for HLA‐DRB1*03. Although their cohort included 43% LS patients, LS and non‐LS patients were not analysed separately, nor was the contribution of the HLA‐DRB1*15 allele studied. Similar findings regarding lymphocyte percentage and CD4⁺/CD8⁺ ratio in HLA‐DRB1*03‐positive patients were reported in a more recent study that included LS patients 32. Unfortunately, other HLA genotypes were not studied. Our results show that besides a low lymphocyte percentage, a high CD4⁺/CD8⁺ ratio provides a good prognosis.

In addition, a more recent paper published by Kinloch et al. found lower lymphocyte counts in combination with a higher CD4⁺/CD8⁺ ratio in BAL fluid of HLA‐DRB1*03‐negative patients 21. However, none of the above‐mentioned studies had investigated the CD4⁺/CD8⁺ ratio in the peripheral blood.

In our cohort, a total of 15 patients were positive for both *03 and *15. The effect of carriage of HLA‐DRB1*15 in ‐DRB1*03⁺ patients is not known; one allele could be dominant over the other or the combined carriage could cancel the effect. No hard conclusions can be made due to low patient numbers in this group. However, a significantly higher percentage of lymphocytes was found in *03^–/*15⁺ patients compared to *03⁺/*15⁺ patients. This finding suggests that the influence of *03 is dominant over the influence of *15 in patients with LS. In essence, this is implicated by the studies from Grunewald, and is therefore most likely 12, 22.

Braun et al., who analysed CD4⁺ T cells from the BAL of a variety of fibrotic lung diseases, suggested that CD103⁺ cells are terminally differentiated effector T cells that might be involved in the process of lung fibrosis 33.

Several authors have reported that patients with a more advanced radiological stage of sarcoidosis show a higher proportion of CD4⁺ T lymphocytes expressing CD103 and have a higher CD103⁺CD4⁺/CD4⁺ ratio 18, 19. Our current findings support this by showing that sarcoidosis patients with *03⁺/*15^–, who are generally known as having a favorable prognosis, have a decreased CD103⁺CD4⁺/CD4⁺ ratio compared with *03^–/15⁺ and *03^–/*15^– patients. Our data suggest that a decreased CD103⁺CD4⁺/CD4⁺ ratio predicts a benign course of disease.

Sarcoidosis has traditionally been regarded as a Th1‐driven disease, characterized by excessive interferon (IFN)‐γ, interleukin (IL)‐12 and tumour necrosis factor (TNF)‐α production in the lungs 34. By separating patients into LS and non‐LS or stratifying by HLA type, differences in effector T cell subsets were seen by Moller and co‐workers 34.

Our data show that the type of lymphocytes is important in redirecting the inflammation towards a self‐limiting or chronic disease. Further studies need to be conducted to elucidate the complex interactions between genetics, T cell function and clinical behavior of the disease.

Whether an inflammatory immune response results in a self‐limiting or a chronic relapsing–remitting type of disease is dependent on multiple factors. Important factors are type and quantity of the antigens, the extent of antigen presentation in terms of tissue and duration, the context of antigen presentation and genetic composition. The net balance determines whether the CD4 response is balanced with a beneficial CD4 T regulatory component or dominated by proinflammatory Th17.1 cells.

Regarding genetic composition, DR3 is generally a good prognostic factor for LS, irrespective of the second DR allele. Detailed immunological analysis of LS cohorts revealed that Vα2.3⁺Vβ22⁺CD4⁺ T cells are expanded in these patients 32. Their phenotype and cytokine profiles in lavages of LS patients demonstrate profiles to be less skewed towards proinflammatory Th17.1 cells 35, and it is speculated these cells may recognize autoantigens such as vimentin 21.

The current study points out that DR3 positivity does not guarantee an inflammatory process to become self‐limiting, and a proportion of the patients develop sarcoidosis. Considering the lower CD4⁺/CD8⁺ ratios particularly observed in sarcoidosis patients, the capacity to establish a more Th1‐like response and involve CD8 T cells in the inflammatory response may mark a relapsing–remitting type of response.

Whether the DR3 non‐LS patients are incapable of producing Vα2.3⁺Vβ22⁺CD4 T cells, or vimentin is not involved in the inflammatory response, remains to be determined. Furthermore, studies aiming to identify the triggers in sarcoidosis 36 will help to identify the Th phenotype and establish their role in the immunological puzzle that results in a self‐limiting or a relapsing–remitting type of disease, along with DR type.

Due to its retrospective design, not all parameters were available for all patients. Furthermore, 23% of the patients were current smokers. Smoking increases the total cell count and reduces the lymphocyte percentage and CD4⁺/CD8⁺ ratio in BAL 20. However, the percentage of smokers was similar in both groups, LS and non‐LS. Also, the duration between BAL and diagnosis in LS patients, a maximum of 4 months, could have possibly influenced the BAL outcomes, because LS is known for its acute inflammation.

The strength of our study is that for confirmation of tag SNPs full HLA‐DRB1* typing was performed in a laboratory that regularly participates in external proficiency testing in order to ensure the quality of the laboratory. Furthermore, we separated the patients into four groups (*03⁺/*15^–, *03^–/*15⁺, *03^–/*15^–, *03⁺/*15⁺) in order to study the influence of both clinical course and associated tag SNPs.

In conclusion, in this study we show that HLA‐DRB1*03 and DRB1*15 can be perfectly approximated by genotyping tag SNPs, which can be used easily in daily clinical practice to distinguish between patients who will develop a chronic course or not. Secondly, we found a significantly higher CD4⁺/CD8⁺ ratio in *03⁺/*15^– patients in the whole group. We also found a lower lymphocyte percentage in *03⁺/*15^– LS patients and a decreased CD103⁺CD4⁺/CD4⁺ ratio in *03⁺/*15^– non‐LS patients.

Our results indicate that a phenotype and HLA markers of favorable disease associate with low a lymphocyte percentage, high CD4⁺/CD8⁺ ratio and low CD103⁺CD4⁺/CD4⁺ ratio in BAL.

Author contributions

B. K. collected and studied patients’ characteristics, performed the statistical analysis and drafted the manuscript. M. S. performed the statistical analysis and drafted the manuscript. L. K. carried out DNA isolation and HLA typing. J. v. d. V. carried out DNA isolation and genotyping. B. M. carried out BAL analysis. J. G. participated in the design and co‐ordination of the study. M. P. participated in the study design and HLA typing and substantially contributed to drafting of relevant parts of the manuscript. C. v. M. conceived the study, participated in its design and co‐ordination and genotyping of samples as guarantor. All authors read and approved the final manuscript.

Disclosures

The authors declare no potential conflicts of interest.