OBJECTIVE—Perforin plays a key role in cell-mediated cytotoxicity. Mutations of its gene, PRF1, cause familial hemophagocytic lymphohistiocytosis but have also been associated with lymphomas and the autoimmune/lymphoproliferative syndrome. The aim of this work was to investigate the role of PRF1 variations in type 1 diabetes.
RESEARCH DESIGN AND METHODS—We typed for the N252S and A91V variations in an initial population of 352 type 1 diabetic patients and 816 control subjects and a second population of 365 patients and 964 control subjects. Moreover, we sequenced the coding sequence and intron-exons boundaries in 200 patients and 300 control subjects.
RESULTS—In both cohorts, allelic frequency of N252S was significantly higher in patients than in control subjects (combined cohorts: 1.5 vs. 0.4%; odds ratio 6.68 [95% CI 1.83–7.48]). Sequencing of the entire coding region detected one novel mutation in one patient, causing a P477A amino acid change not detected in 199 patients and 300 control subjects. Typing for HLA-DQA1 and DQB1 alleles showed that type 1 diabetes–predisposing DQα/DQβ heterodimers were less frequent in patients carrying N252S or P477A than in those carrying wild-type PRF1. We previously found that natural killer (NK) activity is not decreased in most N252S heterozygotes, but we detected one whose NK activity was normal at the age of 12 but strikingly low in early childhood. Here, we discovered that NK function was low in three heterozygotes in early childhood, one homozygous adult, and in the subject carrying P477A.
CONCLUSIONS—These data suggest that N252S and possibly other PRF1 variations are susceptibility factors for type 1 diabetes development.
- ALPS, autoimmune lymphoproliferative syndrome
- HLH, hemophagocytic lymphohistiocytosis
- NK, natural killer
- PBMC, peripheral blood mononuclear cells
In type 1 diabetes, autoimmune damage is mainly due to β-cell destruction by autoreactive cytotoxic T-cells by an inflammatory response organized by autoreactive TH1 cells (1,2). Ethnic variations in its incidence point to involvement of genetic and environmental factors (3,4). Susceptibility genes may include genes coding for molecules involved in immune response control and immune effector functions. Further genes may be those involved in switching off the immune response and leading to homeostatic control of the size of the peripheral lymphocyte pool and reducing the risk of autoimmunity due to cross-reactions between nonself and self antigens (3–5).
Involvement of genes participating in this switching off has been initially suggested for CTLA-4, a receptor expressed by activated T-cells that delivers negative signals upon ligation by B7.1 and B7.2. The role of these negative signals is well documented by CTLA-4–deficient mice developing severe lymphoproliferation and lymphoid infiltration of multiple organs (6). A link with type 1 diabetes has been suggested by its association with CTLA-4 gene polymorphisms associated with decreased receptor function (7–9).
A second link with defective switching off of the immune response came from our observation that a substantial proportion of type 1 diabetic patients display defective function of Fas, a death receptor triggering apoptosis of activated lymphocytes (10). The role of Fas in the immune response is shown by the finding that inherited defects of Fas function cause the autoimmune lymphoproliferative syndrome (ALPS), a rare autoimmune disease characterized by heterogeneous autoimmune manifestations, and lymphocyte accumulation in the spleen and lymph nodes (11–14).
A third mechanism involved in downmodulation of the immune response is perforin-mediated cytotoxicity (15). Cytolytic granules of CD8+ cytotoxic T-lymphocytes and natural killer (NK) cells contain perforin and granzymes and are released on the target cell upon its recognition by cytotoxic cells.
Perforin polymerizes on the target cell membrane where it forms pores that allow the entry of granzymes, which trigger apoptosis of the target cell by cleaving caspases (15). This cytotoxicity is crucial to kill virus-infected cells and clear viral infections but may also be involved in downmodulation of the immune response by fratricide of effector lymphocytes and antigen-presenting cells (16–18). Biallelic loss-of-function mutations of the perforin gene (PRF1) have been associated with ∼30% of cases of familial hemophagocytic lymphohistiocytosis (HLH), a rare life-threatening immune deficiency that occurs in infants and young adults (19–23). Furthermore, 25% of patients display mutations of the MUNC 13-4 gene involved in perforin storage in the lytic granules and exocytosis (24). Intriguingly, some of these variations, themselves insufficient to cause HLH, seem to act as predisposing factors for the development of ALPS (25). This possibility was first suggested by the observation of an ALPS patient with a Fas gene mutation inherited from the father and with a PRF1 mutation inherited from the mother. Since both parents and the patient’s brother, who carried the Fas mutation only, were healthy, it appeared that both mutations contributed to the development of ALPS (25). This was confirmed in a larger group of ALPS patients, where two HLH-associated amino acid substitutions of PRF1 were detected, i.e., N252S and A91V (26). The frequency of N252S was increased in typical forms of ALPS and increased the risk of its development by about 62-fold, whereas that of A91V was increased in an incomplete variant of ALPS and increased this risk by about threefold (26).
The aim of this work was to assess involvement of PRF1 in type 1 diabetes by evaluating the frequency of N252S and A91V in two cohorts of patients and control subjects. Results showed that frequency of N252S, but not A91V, was increased in both cohorts of type 1 diabetic patients. Perforin variations may thus be involved in type 1 diabetes development in some patients.
RESEARCH DESIGN AND METHODS
We analyzed two independent cohorts of type 1 diabetic patients and randomly selected, ethnically matched, healthy control subjects. Patients and control subjects were enrolled from the Diabetes Centers of the Maggiore Hospital of Novara (Novara, Italy), the S. Giovanni Battista Hospital of Turin and the Regina Margherita Children Hospital of Turin (Turin, Italy), the Giannina Gaslini Children’s Hospital, University of Genoa (Genoa, Italy), and IRCCS Policlinico S. Matteo (Pavia, Italy). Patients were consecutive Italian patients followed by these Diabetes centers (Novara, n = 98; Turin, n = 389; Genoa, n = 155; and Pavia, n = 75); control subjects were consecutive Italian donors obtained from the transfusion services of the respective hospitals.
The first cohort consisted of 352 patients and 816 control subjects and the second of 365 patients and 964 control subjects. Patients and control subjects were unrelated, Caucasian, and Italian. Overlaps between different sites were ruled out.
All subjects gave informed consent according to the Declaration of Helsinki (International Committee of Medical Journal Editors, 1995). The research was approved by the ethical committee of the Maggiore Hospital of Novara.
Amplification of PRF1 and mutation detection.
Genomic DNA was isolated from peripheral blood mononuclear cells (PBMCs) using standard methods. In the first cohort, exons 2 and 3 of the PRF1 coding region were amplified in standard PCR conditions. The primers used for amplification have previously been described (19). PCR products were purified with the EXO/SAP kit (GE Healthcare, Piscataway, NJ). Sequencing was performed with the ABI PRISMR BigDyeTM Terminator kit (Applied Biosystems, Foster City, CA) on an automatic sequencer (Applied Biosystems 3100 Genetic Analyser) according to the manufacturer’s instructions with the amplification primers plus two internal primers (forward 5′-CAGGTCAACATAGGCATCCACG-3′; reverse 5′-GAACAGCAGGTCGTTAATGGAG-3′) for exon 3. In the second cohort, genotyping of +272 C/T and +755 A/G single nucleotide polymorphisms was performed with the TaqMan 5′ allelic discrimination assay (Applied Biosystems).
Allelic specific primers and fluorogenic probes were used for discrimination (for +272 C/T see ref. 27; for +755 A/G Cod.4351376, Applied Biosystems). Genotyping of each sample was automatically attributed by the SDS, version 1.3, software for allelic discrimination. All mutations were then confirmed by sequencing.
NK activity of PBMC was assessed by a standard 4-h 51Cr release assay with K562 cells as the target. Results are expressed as specific lysis percentage, calculated as follows: (sample 51Cr release − spontaneous release)/(maximal release − spontaneous release) × 100.
Analysis of lymphocyte subset in PBMCs was performed by direct immunofluorescence and flow cytometry. Perforin expression was evaluated in fixed and permeabilized cells (Cytofix-Cytoperm, BD PharMingen, San Diego, CA) using a phycoerythrin-conjugated anti-perforin antibody (BD PharMingen) and flow cytometry.
HLA-DQA1 AND DQB1 genomic typing.
HLA-DQA1 and DQB1 genes were typed at a high resolution level using the reverse PCR-SSO technique, as previously described (28), in 23 patients carrying N252S or P477A, 81 random patients carrying wild-type PRF1, and 677 healthy children recruited from the cord blood bank of the Pavia center. The control subjects were ethnically matched with the patients and checked for absence of diabetes in their families; their ages ranged from 6 months to 4 years (mean age 2.6 years).
Statistical analysis was performed with GraphPad Instat (GraphPad Software, San Diego, CA). Allelic frequencies were compared with the χ2 test or Fisher’s exact test, as indicated. The Mann-Whitney test was used for NK activity. All P values are two tailed, and the significance cutoff was P < 0.05.
Search for the N252S and A91V variations of perforin.
In PRF1, the C/T substitution in position 272 (rs35947132) of the cDNA (numerations are referred to the GenBank cDNA clone M28393, ATG = +1) and an A/G substitution in position 755 (rs28933375) cause an A91V and a N252S amino acid variation at the protein level, respectively, and have been associated with both HLH and ALPS. By sequencing the genomic DNA, we initially assessed the frequency of these substitutions in 352 type 1 diabetic patients and 816 control subjects. The N252S variation was found in nine type 1 diabetic patients and two control subjects (all heterozygotes). Its allelic frequency was significantly higher in type 1 diabetic (1.3%) than in control (0.1%) subjects (P = 0.0006) and conferred an odds ratio (OR) of 10.55 (95% CI 2.13–70.82) for type 1 diabetes development. The A91V variation was carried by 19 type 1 diabetic patients (18 heterozygotes and one homozygous) and 72 control subjects (69 heterozygotes and three homozygotes), and its allelic frequency was not significantly different in the two groups (type 1 diabetic vs. control subjects: 2.8 vs. 4.6%) (Table 1). The frequency of both variations in the control subjects was similar to that reported in other studies, and their genotypic distributions did not deviate significantly from Hardy-Weinberg equilibrium in any group.
These data indicate that N252S, but not A91V, may be a predisposing factor for type 1 diabetes development. To confirm this observation, we evaluated both variations in a second cohort of type 1 diabetic patients (n = 365) and control subjects (n = 964) using the TaqMan 5′ allelic discrimination assay. N252S was found in 13 type 1 diabetic patients and 13 control subjects (all heterozygotes). Its allelic frequency was higher in type 1 diabetic patients (1.8%) than in the control subjects (0.7%) (P = 0.0179) and conferred an OR of 2.67 (95% CI 1.16–6.15) for type 1 diabetes development. A91V was carried by 36 type 1 diabetic patients (35 heterozygotes and one homozygotes) and 118 control subjects (115 heterozygotes and three homozygotes), and its allelic frequency was not significantly different in the two groups (type 1 diabetic vs. control subjects 5 vs. 6.3%, respectively) (Table 1). The genotypic distributions of these variations did not deviate significantly from Hardy-Weinberg equilibrium in either group. No difference was found between the N252S carriers and the other patients with regard to sex distribution, age at diagnosis of type 1 diabetes, or presence of a second concomitant autoimmune disease (Table 2).
Analysis of the PRF1 gene.
To assess whether type 1 diabetic patients carry other variations of PRF1, the entire coding sequence plus intron-exon boundaries were sequenced in 200 type 1 diabetic patients and 300 control subjects. Besides A91V and N252S, four other known nucleotide variations were detected but were not further evaluated because they did not change the amino acid or influence the splicing sites. Two, C822T (rs885821) and T900C (rs885822), had been previously reported as common polymorphisms not associated with HLH. Their frequency was similar in the patients and the control subjects (C822T, 15 vs. 12%; T900C, 43 vs. 40%). The other two (G435A and A462G) are known to be in perfect linkage disequilibrium with N252S and were in fact only detected in all subjects carrying this variation (25). Moreover, we detected a novel variation in one type 1 diabetic patient, a C/G substitution in position 1,429 (C1429G) causing a P477A amino acid change at the protein level. This variation was not found in any other subject.
Analysis of NK function and HLA-DQ typing.
We had previously found, like other workers, that NK activity is not decreased in subjects carrying N252S. However, we also described an ALPS patient heterozygous for N252S whose NK activity was normal at age 12 years but had been almost undetectable at age 3 years and extremely low at the age 5 years (26). We suggested that, in this patient, N252S was associated with factors decreasing NK function in early childhood, followed by normalization on the part of unknown compensatory mechanisms. Because in this study we detected three N252S heterozygous subjects the early childhood, we assessed whether their N252S was associated with defective NK function by evaluating NK cell count, NK activity, and perforin expression in their PBMCs (Table 3). Two (patients 1 and 2) had type 1 diabetes, whereas the third was the healthy sister of patient 1.
Moreover, we performed the same analyses in patient 1’s healthy father, who was homozygous for N252S; in his wild-type mother; and in the type 1 diabetic patient heterozygous for the novel P477A mutation (patient 3). Results showed that NK activity was defective in patient 1, his sister, and father, whereas it was low but in the normal range in the mother not carrying the mutation; all these subjects displayed normal perforin expression. In patient 2, NK activity was borderline and perforin expression was decreased. In patient 3 (P477A), both NK activity and perforin expression were defective. NK cell counts were normal in all subjects. Specificity of the NK function defect was assessed by evaluating NK function in 10 children in early childhood (age <5 years) and lacking PRF1 variations; 5 were healthy, whereas 5 had type 1 diabetes. Results showed that NK activity of these children was similar to that displayed by our random control subjects (reported in Table 3) and significantly higher (P < 0.01) than that displayed by patients 1, 2, and 3 and by patient 1’s sister (Fig. 1).
Finally, we evaluated the frequencies of DQαβ diabetogenic heterodimers in type 1 diabetic patients carrying or not carrying the N252S or P477A variations and in 677 healthy control subjects. In this analysis, susceptible heterodimers were those comprising a DQα chain with an arginine at position 52 and a DQβ chain with a nonaspartic acid at position 57; a subject can have one, two, or four susceptible heterodimers; the higher the number, the higher the risk for type 1 diabetes (29). Table 4 shows that the distribution of subjects carrying zero, one, two, or four susceptible heterodimers was significantly different in the two patient groups (overall P = 0.028). The proportion of carriers of four predisposing heterodimers was significantly lower in the N252S/P477A group than in the wild-type group (34.8 vs. 65.4%, P = 0.015), but it was higher in both patient groups than in the control subjects (1.6%, P < 0.0001 vs. both patient groups).
This paper stems from our finding that variations of PRF1 may be a predisposing factor for development of ALPS, a rare inherited autoimmune disease (26). It shows that PRF1 variations may also predispose to development of type 1 diabetes.
Of the two PRF1 variations associated with ALPS, i.e., N252S and A91V, only N252S was associated with type 1 diabetes, since its frequency was significantly higher in two independent groups of type 1 diabetic patients than in the respective control subjects. The OR calculated in the combined cohorts (717 patients and 1,780 control subjects) was 3.68 (95% CI 1.83–7.48; P = 0.00007). PRF1 is located on chromosome 10q22, far from the known type 1 diabetes susceptibility loci located on this chromosome, i.e., IDDM10 (10p11-q11) and IDDM17 (10q25). However, other genes involved in cell death (PPIF) or cell-mediated cytotoxicity (PRG1) are located nearby PRF1, and we cannot rule out linkage disequilibrium between the PRF1 variations and those that may play a role in the effector phase of β-cell destruction.
Several works have shown that A91V decreases perforin function by altering its conformation, decreasing its cleavage to the active form, and increasing its degradation (30–33). By contrast, the functional significance of N252S has been debated because it occurs within the membrane-attack complex, a region critically involved in the pore-forming activity of perforin, but several works have associated it with normal NK function and perforin expression (19,31–33). However, these studies were performed on cells from N252S heterozygotes or artificial systems where mutated forms of perforin were transfected in reporter cell lines to assess their expression. We have recently described an ALPS patient heterozygous for N252S who displayed a striking deficiency of NK activity when he was aged 3–5 years followed by normalization when he was aged 12 years (26). Therefore, we suggested that N252S heterozygosis per se or other factors associated with it may decrease NK function in early childhood.
Our present findings substantiate this possibility, since it describes three new children heterozygous for N252S with low NK function in their early childhood. A follow-up will show whether compensatory mechanisms eventually normalize their NK function. A second point is that we also detected low NK function in an adult homozygous for N252S. In the presence of homozygosity, therefore, these mechanisms may not be sufficient. However, N252S homozygosity is not sufficient to induce development of type 1 diabetes, ALPS, or HLH. It is noteworthy that variations altering NK function in early childhood may be particularly significant for type 1 diabetes since it is the outcome of a transient autoimmune aggression that destroys β-cells, generally in childhood.
Intriguingly, one type 1 diabetic patient displayed a novel PFR1 mutation causing the P477A substitution. This mutation was only found in this subject and has never been detected in HLH patients. Its location within the carboxy-terminal C2 domain of perforin suggests that it may have functional significance, since this domain plays a key role in Ca2+-dependent binding of perforin to membranes, the first step of perforin-mediated lytic activity (34). In line with this possibility, our analysis of PBMC from this patient detected defective NK activity and normal perforin expression. In addition to the recurrent N252S mutation, therefore, other sporadic perforin mutations might favor type 1 diabetes development, and the global predisposing effect of perforin variation in type 1 diabetes development may be higher than that calculated for N252S.
Nevertheless, perforin variations are a rare predisposing factor: N252S was carried by 1.5% patients and P477A by 0.5%. N252S-mediated predisposition seems to require concurrence of HLA-predisposing alleles, since about 80% of patients carrying N252S also carried HLA-DQ heterodimers involved in type 1 diabetes susceptibility. However, susceptibility dependent on HLA-DQ seems to be lower in patients carrying the PRF1 variations than in those carrying wild-type PRF1, which suggests that PRF1 contributes to the onset of type 1 diabetes additionally and independently from HLA. PRF1 variations may thus be a predisposing factor for type 1 diabetes. Perforin-mediated cytotoxicity is the main effector system in clearance of virus-infected cells but may also be involved in downregulation of the immune response due to its involvement in fratricide of effector lymphocytes and antigen-presenting cells (15–18,35–39). Defects of both of these functions may predispose to autoimmunity by prolonging the immune response and increasing the risk of cross-reactions between viral and self antigens by molecular mimicry.
This work was partially supported by Compagnia di San Paolo (Turin, Italy), Regione Piemonte (Turin, Italy), Telethon Grant E1170 (Rome, Italy), the PRIN Project (MIUR, Rome, Italy), and AIRC (Milan, Italy).
We are grateful to Dr. Rita Maccario (Department of Pediatric Sciences, University of Pavia, Fondazione IRCCS Policlinico San Matteo, Pavia, Italy) for helpful discussion.
Published ahead of print at http://diabetes.diabetesjournals.org on 15 January 2008. DOI: 10.2337/db07-0947.
E.O. and G.C. contributed equally to this work.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received July 10, 2007.
- Accepted January 9, 2008.