The incidence of type 1 diabetes has, progressively, increased worldwide over the last decades and also in Continental Italian population. Previous studies performed in northern European countries, showed, alongside a general increase in the disease incidence, a decreasing frequency of the highest risk HLA genotype in type 1 diabetes populations, thus emphasizing the role of environmental factors. The aim of the study was to evaluate whether a decreasing trend of high risk HLA, CTLA-4 and PTPN22 genotypes would be present in type 1 diabetes subjects of Continental Italy, a country considered at low incidence of the disease compared to northern European populations. N = 765 type 1 diabetes patients diagnosed from 1980 to 2012 in Lazio region were included. For HLA, CTLA4 and PTPN22 temporal trend evaluation, subjects were subdivided into groups of years according to age at diagnosis. All subjects were typed for HLA-DRB1 and DQB1 by a reverse line blot. The CT60 polymorphism of the CTLA4 and C1858T of the PTPN22 gene were genotyped using ABI PRISM 7900HT (n = 419 and n = 364 respectively). HLA genotypes were divided in high, moderate and low risk categories. The proportion of the HLA risk categories was not statistically different over the three decades in subjects with age of onset <15 years and ≥15 years. The genotype distribution of CT60 polymorphism of CTLA4 gene did not show any change in the frequencies during time. The analysis of the PTPN22 C1858T variant revealed, instead, that the frequency of CT+TT susceptibility genotypes decreased during time (23.9% vs 13.6%, p = 0.017). We can hypothesize that the pressure of the diabetogenic environment could be milder and therefore not sufficient to reduce the need of a strong genetic background (HLA) “to precipitate” diabetes; the increased pressure of the environment could have, instead, some effects on minor susceptibility genes in our population.
Citation: Spoletini M, Zampetti S, Campagna G, Marandola L, Capizzi M, et al. (2013) Temporal Trends of HLA, CTLA-4 and PTPN22 Genotype Frequencies among Type 1 Diabetes in Continental Italy. PLoS ONE 8(4): e61331. doi:10.1371/journal.pone.0061331
Editor: Massimo Pietropaolo, University of Michigan Medical School, United States of America
Received: January 24, 2013; Accepted: March 8, 2013; Published: April 17, 2013
Copyright: © 2013 Spoletini et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by Diabetes, Endocrinology and Metabolism (DEM) Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: Raffaella Buzzetti is a PLOS ONE Editorial Board member. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
The incidence of type 1 diabetes has increased progressively over the last half century  especially in young children. In Continental Italian population (Lazio region), characterized by a low incidence of the disease, we observed that the type 1 diabetes incidence has doubled in the 2004–2009 period compared to the 1989–1993 period, considering the 0–14 year age group . The disease is a complex disorder caused by multiple genes, which interact with environmental factors. The genetic susceptibility will determine the probability of an unwanted outcome to the initial exposure but some environmental factors may influence the rate of progression . The major genes responsible for at least 50% of the type 1 diabetes genetic component are human leukocyte antigen (HLA) linked . As in the other Caucasian populations, the DRB1*03-DQB1*0201/DRB1*04-DQB1*0302 genotype confers the highest risk also in Italy where, however, only 1% of the general population carries this genotype compared to 2.5% of the general population of northern European countries . Other polymorphisms have been reported to be associated with type 1 diabetes, among those the Cytotoxic T Lymphocyte–Associated Antigen-4 (CTLA-4) and in the Protein Tyrosine Phosphatase Non-receptor type 2 (PTPN22) genes, –. Some studies, performed in northern European countries or in patients of Anglo-Saxon origin characterized by a high incidence of the disease  showed a decreasing frequency of the high risk HLA genotype in type 1 diabetes patients over the last decades –, thus emphasizing the role of environmental factors. The aim of the present study was to evaluate whether a decreasing trend of high risk HLA, CTLA-4 and PTPN22 genotype distributions would be present in Continental Italian population.
Subjects for this study include n = 765 type 1 diabetes patients diagnosed from 1980 to 2012 recruited by participating centers of the IMDIAB group in Continental Italy (Lazio region). The diagnosis of type 1 diabetes was based on the American Diabetes Association (ADA) classification criteria .
Our research involved human participants and has been approved by the “Sapienza” University ethical committee and a written informed consent was obtained from all participating subjects. Written inform consent has been obtained from parents or legal guardian for all the children participating to the study. All clinical investigation have been conducted according to the principles expressed in the Declaration of Helsinki.
All subjects were Caucasians with parents of Italian origin. The age at diagnosis ranged from 1 to 49 years (12.7±8.9). Subjects were sub-divided into two categories: <15 years of age at diagnosis and ≥15 of age at diagnosis. For HLA temporal trend evaluation, all subjects were grouped by three decades of diagnosis: 1980–1989, 1990–1999, 2000–2012; for PTPN22 and CTLA4 they were divided into two groups of years of diagnosis: 1980–1995, 1996–2012 due to the smaller number of subjects evaluated, n = 419 and n = 364 respectively.
The HLA genotypes were classified in three risk categories based on the absolute risk values (AR) previously estimated in Continental Italian population (Lazio region) : (1) High risk (AR = 1:23) for DRB1*03-DQB1*0201/DRB1*04-DQB1*0302 genotype (DRB1*04 different from 0403, 06, 11); (2) Moderate risk (AR = 1:150) for DRB1*04-DQB1*0302/DRB1*04-DQB1*0302, DRB1*03-DQB1*0201/DRB1*03-DQB1*0201, DRB1*04-DQB1*0302/X (X different from DRB1*02, 03, DRB1*04-DQB1*0302 (DRB1*04 not 0403, 06, 11) or DQB1*0602) and DRB1*03/X (X different from DRB1*02, 03, DRB1*04-DQB1*0302 (DRB1*04 not 0403, 06, 11) or DQB1*0602 genotypes; (3) Low risk (AR = 1:1100) for the remaining genotypes.
Blood samples were collected and were stored at −20°C until used for genomic extraction of DNA. Genomic DNA was extracted using QIAamp DNA Blood Kit (QIAGEN Genomics Inc., Bothell, WA). All subjects were typed for HLA-DRB1 and DQB1 loci by polymerase chain reaction (PCR) followed by a reverse line blot assay using an array of immobilized sequence–specific oligonucleotide probes, as previously described . Probes were kindly provided by Dr H. A. Erlich and T. Bugawan (not commercial kit; Roche Molecular System, Alameda USA). The CT60 polymorphism of the CTLA4 gene and the missense SNP C1858T of the PTPN22 gene were genotyped using the fluorogenic 5′ nuclease assay application of the ABI PRISM 7900HT Sequence Detection System (ABI, Foster City, CA) in , .
Comparisons of HLA, PTPN22 and CTLA-4 genotype frequencies between the cohorts over time were assessed by the Chi square or by Fischer test. A probability value of 0.05 or less was considered to be statistically significant. Based on the genotype frequencies of HLA (DRB1 and DQB1), CTLA4 and PTPN22 in Italian type 1 diabetes population, we should be able to identify statistical differences between the groups analysed with a power of 96%, 93% and 90% respectively and a p value <0.05.
The mean age at diagnosis of type 1 diabetes was 12.7±8.9 years (mean ± SD), and did not differ across decades of diagnosis. The distribution of HLA genotype categories in subjects with age of onset <15 and ≥15 years, according to the three groups of decades is shown in Table 1. We observed that the proportion of the three HLA risk genotype categories, was not statistically different between the three cohorts in subjects with age of onset <15 years and in subjects with age of onset ≥15 years. Table 2 shows the genotype frequencies of the CT60 polymorphism of CTLA-4 and PTPN22 C1858T variant in type 1 diabetes subjects sub-divided according to the year of diagnosis. The genotype distribution of CT60 polymorphism of CTLA4 gene did not show any change in the frequencies during time. Instead, the analysis of the PTPN22 C1858T variant revealed that the frequency of CT+TT susceptibility genotypes significantly decreased during time (23.9% vs 13.6%, p = 0.017).
Table 1. Proportion of high, moderate and low risk HLA genotypes of type 1 diabetes subjects with age of onset <15 and ≥15 years, subdivided according to year of diagnosis.doi:10.1371/journal.pone.0061331.t001
Table 2. Distribution of genotype frequencies of CT60 polymorphism of CTLA4 gene and of C1858T polymorphism of PTPN22 gene in type 1 diabetes subjects subdivided according to year of diagnosis.doi:10.1371/journal.pone.0061331.t002
In the present study we did not observe any significant temporal change in HLA class II genotype distributions, including the highest risk DRB1*03-DQB1*0201/DRB1*04-DQB1*0302, over the last decades, both in type 1 diabetes children <15 years and in adult subjects, alongside a general increase in the disease incidence. In Continental Italy (Lazio region) type 1 diabetes incidence doubled in the years 2004–2009 (15.68 new cases per 100,000 per year <15 years of age with a peak in 2004:17.3 for 100,000) compared to 1989–1993 and to 1990–1999 periods of time (7.9 and 8.8 new cases per 100,000 per year <15 years of age respectively) .
According to our results, a recent study showed no changes in HLA genotype frequencies in a large sample of type 1 diabetes subjects with age at onset ≤20 from Germany and Austria . They observed that the highest risk HLA genotype was associated with a lower age at onset, as previously described in our population .
This finding is, apparently, in contrast with a decreasing trend of the highest risk HLA genotype observed, over the last decades, in northern European countries , , , , ,  as well as in Australia, where a “predominantly Anglo-Celtic European population” was evaluated, and in United States  (Table 3).
Table 3. Temporal trends of high risk HLA genotype distribution in type 1 diabetes subjects from different countries.doi:10.1371/journal.pone.0061331.t003
Studies performed in northern European populations demonstrated the same decreasing trend of the highest HLA susceptible genotype in children <18 years, although this frequency decreased from 35.6% in the decade 1980–1989 to 19.1% in 2000- in Swedish population , while decreased from 25.3% to 18.2%, in the same period of time, in the Finnish population . Gillespie et al.  observed a significant decreasing frequency of the highest risk HLA genotype DR3-DQ2/DR4-DQ8 over decades in children ≤15 years of age, stressing the concept that a major environmental effect could have been accelerated the type 1 diabetes onset, thus diluting the concentration of high risk HLA susceptible genotypes in type 1 diabetes population. Fourlanos et al. showing a significant decreasing trend of the highest risk HLA genotype from 79% in 1950–1969 to 28% in 2000–2005 , in presence of an increase in type 1 diabetes incidence, concluded that changing environmental conditions have increased the chance to get the disease.
Vehik K et al , showed that the frequency of the DRB1*03-DQB1*02/DRB1*04-DQB1*03 was higher (39%) in type 1 diabetes children both from Hispanic and non Hispanic origin diagnosed during the 1978–1988 period than in those diagnosed during 2002–2004 (28%). The Authors concluded that the increasing environmental exposure is now able to trigger type 1 diabetes in subjects who are less genetically susceptible. This result was confirmed by two larger data sets of mainly Caucasian subjects (86.4%) with a minority of Hispanic and African American, still from USA, where there appeared a significant decrease of the highest risk DR3/DR4-DQB1*0302, over time, together with an increased percent of other HLA genotypes without HLA-DR3 or DR4 .
The reason why the highest HLA susceptible genotype tends to be constant in some populations while decreasing in others is not clear.
The two groups of populations (Italian and German vs the others) mainly differ for a) type 1 diabetes incidence, b) “ab initio” (early 80′) frequency of the HLA class II genotypes in type 1 diabetes populations, both evidences being related, at least in part.
The main difference between the two groups of populations regards the incidence of the disease, Italian  and German  populations being characterized by a moderate/low incidence (approximately 15 new cases per 100,000 per year <15 years) compared to the other countries characterized by a high incidence of the disease (new cases per 100,000 per year <15 years: Finland 64.2, Sweden 39.6, UK 26.4, USA 26.4 and Australia 23.1 , , , , .
Then, in Italian and German populations, the DRB1*03-DQB1*0201/DRB1*04-DQB1*0302 frequency, in type 1 diabetes population, was around 25% in the 80′ compared to >35% of that reported in the northern European countries at that time, except Finland. This finding reflected the low frequency of DRB1*03-DQB1*0201/DRB1*04-DQB1*0302 in the general populations of Italy and German and could be, at least in part, the reason why the incidence of the disease has been low so far in these countries, assuming that the Odds Ratio (ORs) for every specific genotype tended, at that time, to be constant in all Caucasian populations.
Over the past few decades we assisted at a decreasing frequency of the highest HLA susceptible genotype in the diabetes population of countries with a high incidence of the disease, but not in those with a low incidence. Based on these evidences we can speculate that different environmental factors in various populations could differently influence the penetrance of HLA genes. Thus, populations with a lower incidence could be at an earlier stage of the natural evolution of diabetes history; the pressure of the diabetogenic environment could be milder and still “polarized” to the highest HLA genotype, as not sufficient to reduce the need of a strong genetic background (HLA) “to precipitate” diabetes. A logical consequence is that the HLA class II genotypes ORs appear now to be different in the different populations.
Viral infections , physical inactivity, excess of food intake or excessive hygiene may contribute to the increase incidence of type 1 diabetes . It is conceivable that a decreased microbial load in early life may have a major impact on the programming of the immune system, particularly the gut-associated lymphoid tissue .
Environmental factors could either modify the penetrance of susceptibility genes, or, as triggering factors, could contribute directly to the incidence. It has been hypothesized that changes in penetrance might be linked to patterns of childhood immunization, but this has yet to be confirmed . Environmental exposures to dietary antigens and microbes could be implicated in the increasing incidence of type 1 diabetes. However, no single pathogenic environmental agent has been identified that explains all cases .
Due to the limited sample size of subjects, we sub-divided the genotype distribution of CTLA4 and PTPN22 into two groups: 1980–1995 and 1996–2012 according to the year of diagnosis of type 1 diabetes. As well as for HLA genotypes, also the distribution of CT60 polymorphism of CTLA4 gene did not show any change in the frequencies during time, as previously demonstrated .
Conversely, we observed that the frequency of CT+TT susceptibility genotypes of PTPN22 gene decreased during time. For our knowledge this is the first study to analyze the PTPN22 genotype frequency in type 1 diabetes during decades. The risk conferred by this genotype to the disease is quite small (OR = 2.48)  compared to that conferred by HLA DRB1*03-DQB1*0201/DRB1*4-DQB1*0302 in our population OR = 24.7 . The increased pressure of the environment could have some effects on minor susceptibility genes and still not on the major HLA susceptibility component in our population.
Our study has some limitations, the most important one is due to the small number of subjects who took part to this study in the last decade compared to the previous ones. Nevertheless our data contribute to clarify that the increase in type 1 diabetes over the last decades might be explained by a complex interactions between genes and environmental risk factors which may differ in the different populations. The epidemiology of type 1 diabetes suggests that varying gene–environment interactions are likely triggering and/or accelerating the autoimmune destruction of β-cells leading to complete insulin deficiency . When the T susceptibility allele of PTPN22 gene is not present the environmental factors could have a predominant role in the type 1 diabetes pathogenesis.
We can hypothesize that also epigenetic regulation could be one way to explain the rapid increase in incidence and could be a central mechanism by which environmental factors can influence the development of type 1 diabetes .
Conceived and designed the experiments: RB. Performed the experiments: MS SZ. Analyzed the data: GC. Contributed reagents/materials/analysis tools: LM MC. Wrote the paper: RB MS. Recruitment of patients: IMDIAB Study Group RB.
- 1. Gale EA (2002) The rise of childhood type 1 diabetes in the 20th century. Diabetes 51: 3353–61. doi: 10.2337/diabetes.51.12.3353
- 2. Bizzarri C, Patera IP, Arnaldi C, Petrucci S, Bitti ML, et al.. (2010) Incidence of type 1 diabetes has doubled in Rome and the Lazio Region in the 0- to 14-year age-group: a 6-year prospective study (2004–2009). Diabetes Care 33:e 140.
- 3. Noble JA, Valdes AM, Cook M, Klitz W, Thomson G, et al. (1996) The role of HLA class II genes in insulin-dependent diabetes mellitus: molecular analysis of 180 Caucasian, multiplex families. Am J Hum Genet 59: 1134–1148.
- 4. Hermann R, Bartsocas CS, Soltész G, Vazeou A, Paschou P, et al. (2004) Genetic screening for individuals at high risk for type 1 diabetes in the general population using HLA Class II alleles as disease markers. A comparison between three European populations with variable rates of disease incidence. Diabetes Metab Res Rev 20: 322–329. doi: 10.1002/dmrr.455
- 5. Nisticò L, Buzzetti R, Pritchard LE, Van der Auwera B, Giovannini C, et al. (1996) The CTLA-4 gene region of chromosome 2q33 is linked to, associated with, type 1 diabetes. Belgian Diabetes Registry. Hum Mol Genet 5: 1075–1080. doi: 10.1093/hmg/5.7.1075
- 6. Ueda H, Howson JM, Esposito L, Heward J, Snook H, et al. (2003) Association of the T-cell regulatory gene CTLA4 with susceptibility to autoimmune disease. Nature 423: 506–511. doi: 10.1038/nature01621
- 7. Bottini N, Musumeci L, Alonso A, Rahmouni S, Nika K, et al. (2004) A functional variant of lymphoid tyrosine phosphatase is associated with type I diabetes. Nat Genet 36: 337–338. doi: 10.1038/ng1323
- 8. Patterson CC, Dahlquist GG, Gyürüs E, Green A, Soltész G, et al. (2009) Incidence trends for childhood type 1 diabetes in Europe during 1989–2003 and predicted new cases 2005–20: a multicentre prospective registration study. Lancet 373: 2027–2033. doi: 10.1016/s0140-6736(09)60568-7
- 9. Gillespie KM, Bain SC, Barnett AH, Bingley PJ, Christie MR, et al. (2004) The rising incidence of childhood type 1 diabetes and reduced contribution of high-risk HLA haplotypes. Lancet 364: 1699–1700. doi: 10.1016/s0140-6736(04)17357-1
- 10. Hermann R, Knip M, Veijola R, Simell O, Laine AP, et al. (2003) Temporal changes in the frequencies of HLA genotypes in patients with Type 1 diabetes–indication of an increased environmental pressure? Diabetologia 46: 420–425.
- 11. Vehik K, Hamman RF, Lezotte D, Norris JM, Klingensmith GJ, et al. (2008) Trends in high-risk HLA susceptibility genes among Colorado youth with type 1 diabetes. Diabetes Care 31: 1392–6. doi: 10.2337/dc07-2210
- 12. Fourlanos S, Varney MD, Tait BD, Morahan G, Honeyman MC, et al. (2008) The rising incidence of type 1 diabetes is accounted for by cases with lower-risk human leukocyte antigen genotypes. Diabetes Care 31: 1546–1549. doi: 10.2337/dc08-0239
- 13. Palmer JP, Fleming GA, Greenbaum CJ, Herold KC, Jansa LD, et al. (2004) C-peptide is the appropriate outcome measure for type 1 diabetes clinical trials to preserve beta-cell function: report of an ADA workshop, 21–22 October 2001. Diabetes 53: 250–264. doi: 10.2337/diabetes.53.1.250
- 14. Buzzetti R, Galgani A, Petrone A, Del Buono ML, Erlich HA, et al. (2004) Genetic prediction of type 1 diabetes in a population with low frequency of HLA risk genotypes and low incidence of the disease (the DIABFIN study). Diabetes Metab Res Rev 20: 137–143. doi: 10.1002/dmrr.426
- 15. Petrone A, Giorgi G, Galgani A, Alemanno I, Corsello SM, et al. (2005) CT60 single nucleotide polymorphisms of the cytotoxic T-lymphocyte-associated antigen-4 gene region is associated with Graves' disease in an Italian population. Thyroid 15: 232–238. doi: 10.1089/thy.2005.15.232
- 16. Petrone A, Spoletini M, Zampetti S, Capizzi M, Zavarella S, et al. (2008) The PTPN22 1858T gene variant in type 1 diabetes is associated with reduced residual beta-cell function and worse metabolic control. Diabetes Care 31: 1214–1218. doi: 10.2337/dc07-1158
- 17. Awa WL, Boehm BO, Kapellen T, Rami B, Rupprath P, et al. (2010) HLA-DR genotypes influence age at disease onset in children and juveniles with type 1 diabetes mellitus. Eur J Endocrinol 163: 97–104. doi: 10.1530/eje-09-0921
- 18. Petrone A, Galgani A, Spoletini M, Alemanno I, Di Cola S, et al. (2005) Residual insulin secretion at diagnosis of type 1 diabetes is independently associated with both, age of onset and HLA genotype. Diabetes Metab Res Rev 21: 271–275. doi: 10.1002/dmrr.549
- 19. Steck AK, Armstrong TK, Babu SR, Eisenbarth GS (2011) Type 1 Diabetes Genetics Consortium. Stepwise or linear decrease in penetrance of type 1 diabetes with lower-risk HLA genotypes over the past 40 years. Diabetes 60: 1045–1049. doi: 10.2337/db10-1419
- 20. Resic-Lindehammer S, Larsson K, Ortqvist E, Carlsson A, Cederwall E, et al. (2008) Temporal trends of HLA genotype frequencies of type 1 diabetes patients in Sweden from 1986 to 2005 suggest altered risk. Acta Diabetol 45: 231–235. doi: 10.1007/s00592-008-0048-5
- 21. Ehehalt S, Neu A, Michaelis D, Heinke P, Willasch AM, et al. (2012) Incidence of type 1 diabetes in childhood before and after the reunification of Germany–an analysis of epidemiological data, 1960–2006. Exp Clin Endocrinol Diabetes 120: 441–444. doi: 10.1055/s-0032-1309045
- 22. Harjutsalo V, Podar T, Tuomilehto J (2005) Cumulative incidence of type 1 diabetes in 10,168 siblings of Finnish young-onset type 1 diabetic patients. Diabetes 54: 563–569. doi: 10.2337/diabetes.54.2.563
- 23. Dahlquist GG, Nyström L, Patterson CC (2011) Swedish Childhood Diabetes Study Group, Diabetes Incidence in Sweden Study Group (2011) Incidence of type 1 diabetes in Sweden among individuals aged 0–34 years, 1983–2007: an analysis of time trends. Diabetes Care 34: 1754–1759. doi: 10.2337/dc11-0056
- 24. Washington RE, Orchard TJ, Arena VC, Laporte RE, Tull ES (2012) Incidence of type 1 and type 2 diabetes in youth in the US Virgin Islands, 2001–2010. Pediatr Diabetes 2012 Aug 28.
- 25. Catanzariti L, Faulks K, Moon L, Waters AM, Flack J, et al. (2009) Australia's national trends in the incidence of Type 1 diabetes in 0–14-year-olds, 2000–2006. Diabet Med 26: 596–601. doi: 10.1111/j.1464-5491.2009.02737.x
- 26. Ghazarian L, Diana J, Simoni Y, Beaudoin L, Lehuen A (2012) Prevention or acceleration of type 1 diabetes by viruses. Cell Mol Life Sci. Epub ahead of print.
- 27. Gale EA (2002) A missing link in the hygiene hypothesis? Diabetologia 45: 588–594. doi: 10.1007/s00125-002-0801-1
- 28. Knip M (2011) Pathogenesis of type 1 diabetes: implications for incidence trends. Horm Res Paediatr 76 Suppl 157–64. doi: 10.1159/000329169
- 29. Pitkäniemi J, Onkamo P, Tuomilehto J, Arjas E (2004) Increasing incidence of Type 1 diabetes–role for genes? BMC Genet 5: 5. doi: 10.1186/1471-2156-5-5
- 30. Knip M, Veijola R, Virtanen SM, Hyoty H, Vaarala O, et al.. (2005) Environmental triggers and determinants of type 1 diabetes. Diabetes (Suppl 2):S125–S136.
- 31. Petrone A, Suraci C, Capizzi M, Giaccari A, Bosi E, et al. (2008) The protein tyrosine phosphatase nonreceptor 22 (PTPN22) is associated with high GAD antibody titer in latent autoimmune diabetes in adults: Non Insulin Requiring Autoimmune Diabetes (NIRAD) Study 3. Diabetes Care 31: 534–538. doi: 10.2337/dc07-1457
- 32. Vehik K, Dabelea D (2011) The changing epidemiology of type 1 diabetes: why is it going through the roof? Diabetes Metab Res Rev 1: 3–13. doi: 10.1002/dmrr.1141
- 33. Dang M, Buzzetti R, Pozzilli P (2012) Epigenetics in autoimmune diseases with focus on type 1 diabetes. Diabetes Diabetes Metab Res Rev. 2012 Nov 23. doi: 10.1002/dmrr.2375.