Annals of Hematology

, Volume 89, Supplement 1, pp 95–103

Genetic studies in pediatric ITP: outlook, feasibility, and requirements

Authors

  • Anke K. Bergmann
    • Division of Hematology and OncologyChildren’s Hospital Boston and Harvard Medical School
  • Rachael F. Grace
    • Division of Hematology and OncologyChildren’s Hospital Boston and Harvard Medical School
    • Dana-Farber Cancer Institute
    • Division of Hematology and OncologyChildren’s Hospital Boston and Harvard Medical School
    • Dana-Farber Cancer Institute
Future Research

DOI: 10.1007/s00277-009-0865-9

Cite this article as:
Bergmann, A.K., Grace, R.F. & Neufeld, E.J. Ann Hematol (2010) 89: 95. doi:10.1007/s00277-009-0865-9

Abstract

The genomic revolution in medicine has not escaped attention of clinicians and scientists involved in medical management and research studies of immune thrombocytopenic purpura (ITP). In principle, ITP biology and care will benefit greatly from modern methods to understand the patterns of gene expression and genetic markers associated with fundamental parameters of the disease including predictors of remission, risk factors for severity, determinants of response to various therapies, and possibly biological sub-types. However, applying modern genetics to ITP carries severe challenges: (a) Achieving adequate sample sizes is a fundamental problem because ITP is rare (and in pediatric ITP, chronic cases constitute only about one fourth of the total); (b) familial transmission of childhood ITP is so rare that a convincing pedigree requires consideration of other immunologic or hematologic disorders; (iii) ITP is probably biologically heterogeneous, based on clinical observations, immunological studies, and animal models. Here we review the advantages and disadvantages of potential genetic approaches. Sufficient information is available to set reasonable bounds on which genetic analyses of ITP are feasible and how they are most likely to be accomplished. The highest priority is for accurate phenotypes to compare to genetic analyses. Several registries worldwide hold promise for accomplishing this goal.

Keywords

GWASImmune thrombocytopenic purpuraRegistriesCohort studies

Introduction

Immune thrombocytopenic purpura (ITP) is an acquired autoimmune disorder and the most common cause of isolated thrombocytopenia in children. ITP is diagnosed clinically, based upon onset of thrombocytopenia in the absence of other hematologic abnormalities, or other causes of low platelets, and with a characteristic blood smear. Degree of thrombocytopenia is highly variable, and while bleeding manifestations also vary, severe bleeding is rare. Around 50% of typical childhood acute ITP is preceded by a viral or bacterial infection and commonly resolves within weeks to months without treatment. The incidence of pediatric ITP is on the order of four to six cases per 100,000 population annually. About one fourth of these patients go on to chronic disease (which depends in part on the definition of “chronic” as discussed below). For the US pediatric population (∼70 million under age 18), this corresponds to ∼3,000–4,000 total cases annually, and around 1,000 new patients with chronic ITP each year. The prevalence of chronic ITP is somewhat higher, as patients continue to manifest their illness. Similar calculations can be done for any country or region, and an upper bound on number of cases can be set. These numbers inform the analyses that are possible.

While ITP in adults is defined as it is in children, fewer adult cases are acute and self-resolving. In young adults, particularly females, cases are more likely than in children to represent an initial manifestation of lupus, and in older adults, many cases are related to lymphoproliferative disorders. Overall though, typical adult primary ITP is very similar clinically to chronic pediatric ITP. Splenectomy remains a more common treatment in adult than pediatric disease by a wide margin. Whether all adult and pediatric chronic ITP cases are safely “lumped” together for the purpose of genetic analyses remains a matter of conjecture, but we and most treaters believe that the condition in older teens and young adults is indistinguishable clinically and biologically. Genes and pathways identified as important modifiers of lupus and lupus-like states are likely to be important candidate genes for ITP evaluation as well.

Based on the lack of familial cases in ordinary ITP, it is fair to ask, “Why is genetic analysis important at all?” In our view, a search for genes and pathways which modify ITP will improve research and future clinical practice in the field in resourceful/multi-sided ways:

Novel genes and loci identified in unbiased genetic screens (genetic association studies or “transcriptome” analysis) may help elucidate biology of ITP, and suggest the possibility of novel testing for biomarkers of disease, or targets for therapy. These may be true even if the relative contribution of a novel gene to biological variability in ITP is small.

Candidate genes based on our current understanding of pathogenesis and pharmacologic responses can be evaluated to confirm or potentially to rule out postulated contributions.

Pharmacogenomic correlates of response to treatments may have direct clinical relevance in choosing therapies for patients.

Definitions

The temporal cutoff between “acute” and chronic ITP in childhood is arbitrary, and opinion about the best time-point is evolving. While chronic ITP was traditionally defined by a platelet count of ≤150 × 10 9/L at 6 months following diagnosis, remissions after 6 months are common. The Intercontinental Childhood ITP Study Group (ICIS) has used a 12-month cutoff to define chronicity. It remains to be proven how best to apply these newer proposed criteria to pediatric ITP. Various defining criteria have been published. For further clinical review, we refer to the guidelines of The American Society of Hematology, the UK guidelines, the AIEO ITP guidelines, the references cited therein, and the recently established criteria by Rodeghiero et al. [14].

Differences in clinical ITP management approaches are substantial [5]. “Which drug treatment should be chosen in refractory patients?” and “How can we predict disease progression and treatment response?” are examples of unanswered questions.

Therapeutics in ITP vary broadly, including immunomodulatory agents, immunosuppression, and more recently thrombopoietin-receptor agonists. A detailed discussion is beyond the scope of this paper, except to say that responses to individual therapies among patients also vary widely. Biological heterogeneity in ITP course, duration, and/or in drug responsiveness are crucial “phenotypes” and may prove to have genetic bases. Although several lines of investigation have focused on the pathophysiology of this immune-mediated disease, little is understood about the underlying cause(s) of ITP (reviewed by [68]).

Importance of registries

As is true for many rare disorders, we believe that carefully designed multicenter registries are the only means to achieve sufficient level of detail in salient outcomes of ITP for comparison to genotype or gene expression results. While the extant registries (Table 1) vary by methodology, they have in common the ability to discern various patient characteristics including demographics, duration of disease, and response to treatment(s). The largest registries are on the scale of a few hundred to 1,000 to 2,000 patients. None have more than a few hundred chronic ITP patients. This is an important number in considering applicability of genome-wide approaches, discussed below.
Table 1

ITP registries

Registry

Inclusion criteria

Locale

Enrolled patients (as of July 2009)

DNA collection

Web page

North American Chronic ITP Registry (NACIR), since 2002

6 months to 18 years

North America

439

Yes

www.itpkids.org/content/research.html

Clinical diagnosis of ITP (including primary ITP (isolated immune thrombocytopenia), secondary ITP, and Evans syndrome)

Chronic ITP (duration > 6 months)

Enrollment 6 months after ITP diagnosis

Intercontinental Childhood ITP Study Group (ICIS) and Pediatric and Adult Intercontinental Registry on Chronic ITP (PARC), since 2004

Children >2 months and adults

International (30 countries)

1,886

Yes

www.itpbasel.ch/

Clinical diagnosis of ITP (acute and chronic)

UK Pediatric ITP registry, since 2006 (data are transferred to ICIS)

3 months to 16 years

UK

240

Anticipated 2009

www.uk-itp.org

Clinical diagnosis of ITP (acute and chronic)

Enrollment within 12 months of ITP diagnosis

Nordic Society of Pediatric Haematology and Oncology (NOPHO) ITP working group, 1998– 2000

0–14 years

Scandinavia

506

No

Closed (plan to continue in 2010/2011)

Clinical diagnosis of ITP (acute and chronic pl.c. below 30 × 109/L)

Enrollment at ITP diagnosis

Associazione Italiana Ematologia ed Oncologia Pediatrica (AIEOP) ITP study group, 1999–2002

6 months to 18 years

Italy

609

No

Closed

Clinical diagnosis of ITP (acute and chronic)

Enrollment at ITP diagnosis

In ITP, several key features of phenotype can only be determined by observation over time: Differentiation of chronic from acute disease (by any of the definitions above) necessitates either a retrospective diagnosis after the diagnostic time interval (e.g., 6, 9, or 12 months) or prospective analysis with this length of follow-up. As a practical matter for the current state of ITP registry research, it would be helpful if all registries attempt to ascertain the platelet count status at 6, 9, and 12 months from diagnosis to facilitate comparisons at a later time. Longitudinal follow-up at intervals after 12 months can also define a “remission” status in chronic ITP. For the purpose of genetic analysis, we propose that very long-term follow-up will lead to a problem of diminishing returns. The fraction of patients achieving remission after more than 2 years is certainly too small for any primary genetic analyses, but would be amenable to analysis of novel candidate genes once they are discovered. The same can be said for patients with “relapse” of ITP months or years after achieving a remission of all therapies. While overt relapse is an important phenotypic characteristic, it occurs in only a small fraction of ITP patients (perhaps 5%), and insufficient power would be available for primary analysis of this outcome in all of the world’s registries combined.

Response to therapies is a key component of the ITP phenotype, but this outcome is highly variable among ITP patients, and there is no clinical consensus for a particular treatment at a particular time in the disease course for various ITP clinical situations. Therefore, even in the largest registries, only a minority of patients would have received any particular therapy. Less severely affected patients will receive fewer pharmacologic treatments, and even severely affected, chronic ITP patients achieve experience with various therapies only over months or years. Furthermore, in clinical practice and most registries (as opposed to prospective research studies), pre- and post-therapy platelet counts will not be reliably available for each treatment in all circumstances.

Types of genetic analysis

Two ITP multicenter registries already include the collection of DNA samples, and others aim to extend their registries in similar way. Advantages and disadvantages of different genetic approaches impact the possibilities of genetic studies in ITP (Table 2).
Table 2

Relevant genetic studies in pediatric ITP

Type of study

Requirements (in addition to well described phenotypes)

Advantages

Disadvantages

Linkage studies (family-based)

Must choose statistical approaches and models of inheritance. Examples—autosomal recessive: 10–15 affected individuals (unaffected family members uninformative except with consanguinity) vs. dominant: ∼5 affected individuals (unaffected family members are informative)

Potential for adequate statistical power with a relatively small number of kindreds

Familial cases of ordinary chronic ITP are extremely rare

Genome-wide-association studies (population-based)

Power calculation of sample size based on expected effect size of a given gene

Unbiased, powerful approach for multifactorial disorders in which single genes contribute only a fraction of the variance in phenotype. Genomic DNA is not affected by current therapy or current disease status; controls may be population-based from public databases

Requires at least hundreds of patients even for relatively strong effect sizes; thousands for modest effects; ethnic diversity may be an impediment; cost

Comparative genomics

Adequate animal models

Family studies are much easier in domestic animals. Informative pedigrees may be easy to obtain. Genomes available for many species (to compare syntenic regions of human chromosomes)

Not all animal models are generalizable to human disease. Focus on conserved genetic regions can be biased.

Candidate gene approach

Pathophysiology needs to be partially understood

Several kinds of candidate gene analyses are possible—gene expression (“transcriptome” analysis or proteomics may help define candidate genes); allows many kinds of polymorphisms (not only SNPs); allele transmission studies (if parental DNA has been isolated as well)

Biased: Only known genes can be analyzed; genes of interest to investigators may or may not be the right genes. Generally, every study will require controls. In ITP, not clear which cells to study for gene expression or proteomics (e.g., subsets of lymphocytes)

Pharmacogenetic studies

Candidates in drug metabolism need to be identified, e.g., “DMET chip” analysis

Inexpensive, may account for variable response to medications, which is a key feature in ITP

Potentially biased: only known genes can be analyzed; genes of interest to investigators may or may not be the right genes

Family-based linkage studies

Family-based linkage studies will be limited since familial cases of ordinary ITP are very rare. It is unlikely that such cases will be found in significant numbers in registries. Our center’s experience is not more than a handful of convincing pedigrees of “familial” ITP in more than 1,000 consecutive probands. Other genetic disorders with thrombocytopenia, which may be (mis)diagnosed as ITP in the proband, include myosin heavy chain disorders (MYH9 gene on chromosome 22q11.2, Online Mendelian Inheritance in Man (OMIM) #605249), autoimmune lymphoproliferative syndrome, related to defects in FAS pathway (OMIM 601859), and Wiscott–Aldrich Syndrome/X-linked thrombocytopenia (OMIM 301000; http://www.ncbi.nlm.nih.gov/sites/omim queried 7/31/09)

Genome-wide association studies

In these times of ready single nucleotide polymorphism (SNP) chip analysis and whole genome sequencing, genome-wide association studies (GWAS) are technically feasible and could reveal novel insights especially in the pathophysiology and disease susceptibility. However, the most challenging requirement is to achieve adequate sample size to conduct association studies with reliable and conclusive results. Three major factors affect the sample size, “n” required to have significant results. First, the smaller the effect of any given gene, the larger a population will be required to see the effect. GWAS studies on populations of many thousand individuals can detect relative risks much less that 1.5-fold, but these will be impossible in chronic ITP. Second, rare but important genes responsible for familial ITP would be very difficult to ascertain in population studies. For all practical purposes, alleles with less than 5% or 10% prevalence in the population would need enormous “risk” to be discovered by unbiased approaches. Third, when the number of loci tested becomes very large (Affymetrix 6.0 SNP-chips allow testing of nearly 106 loci and up to 2.5 million SNPs can be inferred), an enormous statistical price is paid for multiple comparisons. For the traditional accepted type I error of 5%, to test one million loci means that p values need to be less than 5 × 10−8 to be statistically significant. This fact alone drives sample sizes into the thousands for unbiased analyses. Table 3 demonstrates how effect size and allele frequency drive sample size requirements for recessive, co-dominant, or dominant inheritance
Table 3

Sample size estimates for 80% power

Mode of inheritance

Relative risk

Allele frequency 0.05

Allele frequency 0.1

Allele frequency 0.3

Dominant

1.3

2,500

1,250

750

2

300

150

110

Recessive

1.3

8 × 104

2 × 104

2,500

2

1 × 104

2,500

300

Co-dominant (2df)a

1.3

3,000

1,500

1,000

2

350

200

150

Number of cases calculated with the “Power for Genetic Association analyses 1” program [13]. Parameters used: incidence 5/100,000 (0.00005), type I error (alpha) = 0.05, effective degree of freedom = 1, control to case ratio = 1. Only numbers in italics are sample sizes likely to be achieved in the pediatric ITP population

aCo-dominant (2 degrees of freedom): This model tests the effect of each genotype independently; therefore, it is assumption free

Candidate gene approach

Candidate gene studies do not carry the same severe statistical penalty as GWAS techniques, although corrections for multiple comparisons are often not mentioned at all in published reports. A few association studies in ITP have been published (Table 4). Results of these studies have to be confirmed in larger cohorts to make conclusions about pathogenetic implications. These reports include diverse groups of patients and populations, which can make the analyses difficult to generalize. Several are of borderline statistical significance (p values not much less than 0.05), perhaps because of small sample size, chance associations, or biological heterogeneity. Only by replication in additional populations will true significance become clear. Investigators will achieve most meaningful results if they are able to assign patients definitively to acute and chronic ITP phenotypes (in case these are biologically distinct). It is imperative to have controls which match the ethnic and racial background of the case group: Frequencies of polymorphisms differ among groups, so that overrepresentation of different ethnic groups between cases and controls by ethnicity complicates analysis and accurate interpretation of genetic data [9].
Table 4

Reported DNA variations in ITP

Gene/gene product

SNP

Outcome

Number and characteristics of patients

Ethnicity

Reference

B cell-associated

TNFSF13B (BAFF)a/TNF13B

−871 T/C

No significant association (no p value, chi-square calculated for this review not significant)

53 adult cITP (serum)

Caucasian (German)

[14]

17 adult cITP (genetics)

10 controls

Ig variable (v) gene (hv3005)

Deletion

Deletion in 14/44 cITP patients, compared to 7/88 in controls (p = 0.002)

44 adult cITP

Caucasian

[15]

88 controls

T cell-associated

IL-4, IL-6, IL-10

IL-4: 70-bp VNTR in intron 3

IL-4 (RP1/RP2) genotype (p = 0.04), IL-4 (RP2) allele (p = 0.03), and IL-10 (A/C) genotype (p = 0.01) were less frequently detected in children with chronic ITP than in controls

50 pediatric aITP

Asian (Chinese)

[16]

30 pediatric cITP

IL-6: −572 G/C

100 controls

Il-10: −627 C/A

 

Interleukin (IL)-1β and IL-1 receptor antagonist (IL-1 Ra)

IL-1β: +3953

IL-1 Ra (I/II) genotype is underrepresented in patients with aITP (p = 0.02)

50 pediatric aITP

Asian (Chinese)

[17]

IL-1 Ra: 86-bp VNTR in intron 2 (I = 410 bp; II = 240 bp)

30 pediatric cITP

100 controls

TNF-α

TNF-α: −238 G/A, −308 G/A

TNF-β +252 (G/G) genotype significantly higher in patients with ITP (p = 0.04)

84 adult cITP

Asian (Japanese)

[18]

TNF-β

TNF-β: +252 G/A

56 controls

IL-1β

IL-1β: −511 C/T, +3953 T/C

CTLA-4

49 A/G (exon 1)

No significant association

60 ITP patientsb

Albanian (Macedonian)

[19]

100 controls

T cell immunoglobulin- and mucin-domain-containing molecule-3 (TIM-3)

−1516 G/T, −574 T/G, 4259 G/T

No significant association

187 adult/pediatric, a/cITPb

Asian (Chinese)

[20]

PTPN22

1858 C/T

10 patients (22%) were heterozygous and 2 (4.4%) homozygous for this SNP compared to frequency of 8.6% observed in a published population study of 960 controls (p < 0.0001)

45 adult a/cITP

Caucasian

[21]

SOX13

SOX13 coding region

Two sequence variations 1603 C/T—Pro534Ser and 1836 C/T—synonymous were detected in 4 patients. The rare heterozygous genotypes are 3.5 (1603 C>T) respectively 2.3 times (1836 C>T) more common than in published non-ITP controls

34 patients

 

[22]

B/T cell-associated

CD72

VNTR in intron 8

No significant association

206 adult/pediatric, a/cITPb

Asian (Chinese)

[23]

169 controls

Fcγ receptor-associated

TNF, LTA (lymphotoxinα = TNF superfamily member1), IL-1 RN, IL-1A, IL-1B, IL-4, IL6, IL10, FcγR2A, FcγR3A, FcγR3B

TNF: −308 1/2, LTA: NcoI 1/2; IL-1 RN: VNTR in intron 2 (2–5)

FcγR3A 158 (V/F) increased in cITP patients (p = 0.017)

37 pediatric cITP

Caucasian

[24]

IL-1A: −889 C/T, IL-1B: 3953 C/T, IL-4: −590 G/T, IL6: −174 C/G, IL10: −1082 A/G, FcγR2A: 131 H/R

FcγR3B (NA1/NA1) genotype overrepresented in cITP patients (p = 0.0022)

218 controls

FcγR3A: 158 V/F, FcγR3B: NA1/NA2

LTA NcoI (2/2) genotype overrepresented in cITP patients (p = 0.0070)

TNF −308 (1/2) and (2/2) genotypes were observed less in cITP patients (p = 0.027)

FcγR2A, FcγR3A

FcγR2A: 131 H/R, FcγR3A: 158 V/F,

Fcγ3A 158 (F/F) is less frequently in ITP (p < 0.05); in patients with Fcγ3A 158 (V/V) was the remission rate with medications higher (p < 0.05)

104 adult cITP

Asian (Japanese)

[25]

59 controls

FcγR2A,FcγR3A

FcγR2A: 131 H/R

FcγR2A 131 H and

101 pediatric a/c ITP

Unknown

[26]

FcγR3A: 158 V/F

FcγR3A 158 V allelic frequencies are associated with childhood ITP (p = 0.03)

130 controls

FcγR2C, FcγR3A, FcγR3B

CNV

Overrepresentation of FcγR3A 158 V allele in pediatric ITP (p < 0.001)

116 adult/pediatric, a/cITPb

Caucasian

[27]

FcγR2A: 131 H/R

FCGR2B/C: −386 C/C rare in cITP patients (p = 0.03 in adult and 0.05 in pediatric ITP patients)

100 controls

FcγR2B: 232 I/T FcγR3A: 158 V/F

Overrepresentation of ORF FcγR2C gene in ITP patients (p < 0.009)

FcγR3B: HNA1a/b/c

FcγR2C: stop (in exon 3)

FcγR2B/C: −386 G/C

FcγR2A

131 H/R

No significant association

29 post-splenectomy patients, refractory ITP

Caucasian (Irish)

[28]

61 controls

Cytokine

Transforming growth factor-β1 (TGF β1)

509 C/T

No significant association

35 pediatric aITP

Turkish

[29]

Codon 10 (L/P)

40 pediatric cITP

Codon 25 (R/P)

97 controls

DNA methylation

DNMT3B

−149 C/T

No significant association

201 adult/pediatric, a/cITPb

Asian (Chinese)

[30]

136 controls

Antigens

Human platelet antigen 1, 2, 3, and 5 systems

HPA 1a−/+/1b−/+

Only HPA2-a was detected in ITP patients not HPA-2b (p = 0.017)

33 adult chronic refractory ITP patients

Caucasian

[31]

HPA 2a−/+/2b−/+

80 controls

HPA 3a−/+/3b−/+

HPA 4a−/+/4b−/+

HLA-A, -B, and -C antigen and -DR DNA typingc

HLA-A, -B, and -C antigen and -DR DNA typing

DRB1*0410 was significantly increased in ITP patients (p < 0.05)

111 adult aITP

Asian (Japanese)

[32]

71 controls

aITP acute ITP, cITP chronic ITP

aSynonymous: BLYS, CD257, TALL-1, TALL1, THANK, TNFSF20

bNot reported in more detail

cOther HLA typing studies [3336] have discrepant results and used the non-genetic two-stage microlymphocytotoxicity test according to the NIH Tissue Typing Manual. Therefore these reports are not reported in this table

Comparative genomics

Convincing familial or breed-specific ITP is common in some dog groups, including old English sheep dogs, poodles, and American cocker spaniels. The fact that ITP is more common in some breeds (33% of all canine cases are seen in cocker spaniels) supports the hypothesis that genetic variation is a key factor in the development of ITP. Comparative studies of identified candidate genes in canine models is feasible due to the availability of the complete dog genome genomic sequences and the fact that linkage studies based on affected and unaffected littermates and pedigrees could be generated.

Pharmacogenomics

Hematologists routinely practice pharmacogenetics with any ITP patient under consideration for Anti-D therapy. The Rh blood group locus on chromosome 1 determines RhD expression, and d/d individuals do not respond to this drug because they do not express the D antigen on erythrocytes. Although treaters do not routinely think of blood type as a genetic analysis, we see this as a nontrivial example; on the contrary, the Rh gene (which has nothing to do with ITP pathogenesis) demonstrates how specific genomic loci may have powerful genetic effects on environmental (drug treatment) responses to ITP.

It is controversial if a gene, which may be a key receptor protein for the chimeric anti-CD20 antibody, rituximab, namely the FCGR3A V/V allele, influences the response to rituximab [10, 11]. The identification of further candidates in the metabolism of rituximab would make pharmacogenetic studies in ITP feasible.

6-Mercaptopurine and the related azathioprine must be converted to active 6-thioguanine, for biological activity; these compounds are then cleared in part by thiopurine-methyltransferase (TPMT). A common allelic variant of the TPMT gene (approximately 10% in Caucasians) causes slow elimination of the active drug, and homozygotes for this variant (about 1% in the population) have high risk of severe myelosuppression and other toxicities [12].

Conclusion

Logistic and statistical considerations set strong boundaries on how genetic methods may be used for analysis of pediatric ITP pathophysiology and phenotypic expression. To the extent that ITP is biologically heterogeneous (T vs. B cell origin, for example), these restrictions will be even more severe.

However, the several ITP registries worldwide will play an important role in defining phenotype and identifying subjects for feasible genetic studies. We propose that if the registries can have common phenotypic data elements, they will be much more powerful tools to assist basic scientists studying mechanisms of disease. If registries and samples were pooled collaboratively, larger sample sizes would be possible but replication of analyses would be hard. If registries are kept distinct, replication data sets will be easy to come by, but sample sizes will be small. Pharmacogenetics and comparative genomics have not yet been optimally tested. Unbiased GWAS studies will be particularly challenging due to sample size problems.

Copyright information

© Springer-Verlag 2009