AbstractBackground Acute graft-versus-host disease is a severe complication of allogeneic stem cell transplantation in which the functional immune cells of the donor recognize the recipient as foreign and mount an immunological attack. There is an urgent need for better diagnostic instruments for the assessment of acute graft-versus-host disease. In the present study, a novel bioinformatics framework was used to identify gene expression patterns associated with acute graft-versushost disease in patients undergoing allogeneic hematopoietic stem cell transplantation.Design and Methods Peripheral blood cells were collected prospectively from patients who did develop acute graftversus-host disease (YES) and from those who did not (NO). Gene expression profiling was performed using a panel of 47 candidate genes potentially involved in alloreactive responses. The entire population of YES/NO acute graft-versus-host disease patients formed the experimental validation set. Personalized modeling based on a gene selection technique was applied to identify the most significant mRNA transcripts, which were then used to profile individual data samples for training and testing the classification/prediction framework.Results A leave-one-out cross-validation procedure was performed to investigate the robustness of the classification framework producing the following results: 100% on the training dataset and 97% on the testing dataset. According to our integrated methodology, transcripts for FOXP3, ICOS, CD52 and CASP1, genes involved in immune alloreactive responses and participating in immune cell interactions, were identified as the most informative biomarkers in allogeneic stem cell transplant recipients experiencing acute graft-versus-host disease.Conclusions This study demonstrates that the integrated methodology proposed is useful for the selection of valid gene targets for the diagnosis of acute graft-versus-host disease, producing satisfactory accuracy over independent clinical features of the allogeneic transplanted population.
The establishment of the donor's immune system in an antigenically distinct recipient undergoing allogeneic hematopoietic stem cell transplantation (HSCT) confers a desired therapeutic graft-versus-tumor effect. It can, however, also result in detrimental acute graft-versushost disease (aGvHD) associated with a high rate of transplant-related mortality. The aGvHD phenomenon occurs classically within 100 days after transplantation and is mainly associated with the clinical consequences of an immune-mediated attack or “cytokine storm” of target tissues including the gastrointestinal tract, liver and skin.
More specifically, aGvHD is a complex disease resulting from donor T-cell recognition of a genetically disparate recipient that is unable to reject donor cells following allogeneic HSCT.1 In fact, the induction of aGvHD is dependent on the presentation of host alloantigens by antigen-presenting cells to naïve donor T cells.2 Unfortunately, the clinical diagnosis of aGvHD is not always straightforward or easy to establish, and skin or intestinal biopsies are required to confirm the diagnosis. Consequently, therapeutic strategies based on the suppression of critical molecular pathways involved in T-cell activation and the clinical manifestation of aGvHD may lead to unnecessary over-treatment by clinicians. Extensive research is currently underway to identify blood-based biomarkers of aGvHD using novel tools including proteomic and molecular methods.3,4 In this respect, proteomic analysis has identified several serum proteins as potential biomarkers of multi-organ aGvHD. One promising biomarker, Elafin protein, has been proposed to have significant diagnostic and prognostic value specifically for skin GvHD.5 Recently, a strong association was observed between soluble tumor necrosis factor receptor 1 (sTNFR1) levels and the development of aGvHD, although the low sensitivity of sTNFR1 determination reduces the clinical utility of this biomarker.6 Donor gene-expression profiles (GEP) can also be used to predict the occurrence of chronic GvHD in the recipient.7 Although only a pilot study, Buzzeo et al. identified a GEP signature of aGvHD, independent of tissue localization, in patients undergoing allogeneic HSCT.8
In the present study, we evaluated the gene expression levels of a complex panel of 47 functional candidate genes hypothesized to be involved in immune alloreactive responses in aGvHD.9 We examined the applicability of GEP analysis for supporting the diagnosis of aGvHD at presentation of symptoms with multiple targets and severity.
Although transcriptional profiles may be obtained from many human tissues, the status of the immune system can be best monitored by profiling transcripts in the blood. We, therefore, chose to use unselected peripheral blood cell (PBC) samples, which have the advantage of being easily accessible and offering a non-invasive way of monitoring molecular changes safely.
The major objective of our study was to determine whether GEP could be used to discriminate, at the onset of a clinical suspicion of aGvHD, those allogeneic HSCT recipients who will develop aGvHD from those who will not. A novel computational framework previously presented in the technical literature10,11 was applied in this study.
Design and Methods
Patients and graft characteristics
In this single center study we included 59 consecutive patients who received allogeneic transplants between 2007 and 2010. Hematopoietic progenitor cells were collected by apheresis and used as stem cells in 44 cases, whereas in 15 transplants the source of stem cells was bone marrow. The characteristics of the patients, transplants and grafts are presented in Table 1.
Any GvHD before day100 was classified as acute. The severity was categorized as grade I-IV according to modified Glucksberg criteria (A-D by the International Bone Marrow Transplant Registry index).12,13 The diagnosis of aGvHD was confirmed by skin or gastrointestinal biopsy in all cases. Prophylaxis for aGvHD consisted mainly of cyclosporine A (2 mg/kg on day -1, 1 mg/kg from day 0) combined with a shortterm course of methotrexate: 15 mg/m2 on day +1, and 10 mg/m2 on days +3 and +6 and also on day +11 in the case of mild-moderate mucositis. In the case of an alternative donor, anti-lymphocytic serum at a total variable dose ranging from 4.5-15 mg/kg was added to the conditioning regimens for in vivo T-cell depletion. In the case of total lymph-node irradiation plus antithymocyte globulin, methotrexate was replaced by mycophenolate mofetil at a maximum dose of 2 g daily. For patients undergoing haploidentical transplantation, aGvHD prophylaxis consisted of cyclophosphamide 50 mg/kg on days +3 and +4 followed by tacrolimus and mycophenolate mofetil.
Experimental design and sampling
The study was approved by the local Ethics Committee and the procedures followed were in accordance with the World Medical Association's Helsinki Declaration. PBC were collected from all patients +15, +30, +45, +60, +90 and +120 days after HSCT and at each event compatible with suspected aGvHD. Overall, 400 samples were collected. Because the experimental design plays a crucial role in the search for useful biomarkers, the first step was to choose the most informative specimens. To this end, we selected those samples obtained upon observation of the first clinical symptoms and which were confirmed by biopsy. Positive cases of aGvHD (YES) and controls negative for aGvHD (NO) were matched to avoid bias due to immunosuppressive treatment. This matching was best achieved using a database containing high-quality samples linked to quality controlled clinical information. We used 23 samples from aGvHD (YES) patients obtained at diagnosis and before any anti-aGvHD therapy and we selected 36 samples from patients who did not experience aGvHD (NO). Recipients were regarded as not having developed aGvHD when they remained at least 120 days without presenting any aGvHD events. The entire group of YES/NO patients formed the validation population for subsequent modeling studies.
Biological samples RNA extraction, low density gene expression assay
Total RNA was extracted from PBC samples using the RNAeasy Mini Kit (Qiagen) according to the manufacturer's instructions. Purified RNA was reverse transcribed using the High Capacity cDNA Archive kit with random primers (Applied Biosystems), and a multigene expression assay was carried out with a TaqMan® Low Density Array Fluidic card (TLDA-card) on the PRISM Real-Time 7900HT Sequence Detection system platform (Applied Biosystems), according to the manufacturer's instructions. For the TLDA-card, we selected 47 candidate genes9 (Online Supplementary Table S1) involved in the immune network and in the pathogenesis of inflammation; 18S was considered the endogenous reference gene. A pool of PBC from ten healthy volunteers was used to obtain cDNA to be used as the calibrator for the assay. Expression of each gene transcript was measured in triplicate and then normalized to the calibrator sample.
Quantitative real-time polymerase chain reaction, single assay
Total RNA was extracted from PBC samples or healthy donors and cDNA was obtained as described above. The levels of expression of FOX-P3 (ID 00203958_A1) and ICOS (ID 00359999_A1) were determined using primers and probes from Applied Biosystems (ABI Assays on Demand; http://www.appliedbiosystems.com/). The 18S pre-developed TaqMan assay (99999901_S1) was used as the endogenous control. All gene expression assays were performed on the ABI PRISM Real-Time 7900HT Sequence Detection system (Applied Biosystems). The relative expression level of target genes was determined using the ΔΔCT method.
Immunohistochemical staining was performed on 3 μm thick serial sections cut from formalin-fixed and paraffin-embedded tissues. Conditions were optimized for each primary antibody: anti-FOXP3 (mouse monoclonal antibody, Abcam) and anti-ICOS (rabbit monoclonal antibody, Spring Bioscence). Antigen retrieval was performed in a temperature-controlled water bath at 98 °C with EDTA pH 9/citrate buffer pH 6 and antibody binding was visualized using an EnVision™ Detection System, Peroxidase (DAKO) with DAB as the chromogen. Tonsil and lymph node were used as positive controls for FOXP3 and ICOS, respectively. The negative control was included with each run by substituting the primary antibody with non-immune rabbit serum. The control sections were treated in parallel with the samples.
Personalized modeling for identification of target genes in acute graft-versus-host disease
With the aim of identifying the most informative genes able to detect aGvHD at the onset of clinical signs, we proceeded to better define our dataset by pre-processing experimental variables such as race, disease, stem cell source, HLA tissue typing, number of mismatched loci between donor/recipient, donor/recipient age, donor/recipient sex, conditioning intensity, and donor/recipient blood group. In this study, we used a novel personalized modeling based gene selection (PMGS) method,11 previously developed for GEP data tailored to the diagnosis/classification of aGvHD. Personalized modeling is a relatively new method in bioinformatics research and published information based on this method is still fairly sparse. Representative works have been published by Song and Kasabov.14,15
The main objective of personalized modeling is to create a model for each patient (sample), which is able to reveal the most important information specifically for each sample, focusing attention on the individual patient (sample) rather than simply on the global problem space.16,17
Previous works have reported that personalized modeling can produce better classification results than those obtained from classical global modeling,14,16,18 making it more appropriate to build clinical decision support systems for new patients. The framework proposed by Fiasché et al.11,19 used a “personalized” wrapping method, a PMGS described in detail in recent papers20,21 for gene expression data analysis, integrating new data with the existing models; the block diagram is reported in Figure 1. We used weighted-weighted distance K-nearest neighbor (WWKNN) as the classification algorithm15 given that the simple weighted distance K-nearest neighbor (WKNN) algorithm had returned better results than other classification techniques in previous experiments (an interesting comparison is reported by Fiasché's group).11,19,20
Finally, a compact genetic algorithm22 was used to optimize the learning function during the training process making framework performances useful in clinical applications.
Results of quantitative relative real-time polymerase chain reaction, performed as single assays, were used to calculate the area under the curve (AUC) by receiver operating characteristic (ROC) analysis for FOXP3 and ICOS expression levels. Statistical analyses were performed using the SPSS software package for Windows, release v13.0, 2004 (SPSS, UK).
Personalized modeling-based gene selection identified FOXP3, ICOS, CD52 and CASP1 as the most informative gene transcripts associated with acute graft-versus-host disease
In our attempt to determine a gene expression signature that is diagnostic of aGvHD, we first correlated global GEP data obtained from the TLDA-card with the occurrence of aGvHD in allogeneic HSCT recipients at the onset of clinical signs. A heat-map of differentially expressed genes is shown in Online Supplementary Figure S1. Since simple statistical analysis or classical data-mining techniques failed to identify a specific gene pattern, we used computational intelligence methods according to previously published algorithms.17,19,20 The PMGS approach was applied to the data obtained from the TLDA-card to identify a group of genes to represent the data set. In our framework the most important genes selected for each patient may differ (personalized subset), but during our analysis, it was evident that for each run, four genes were always present in the entire subset of patients who did develop aGvHD. The following genes were always identified with other clinical variables: FOXP3 (a member of the forkhead transcriptional factor family), ICOS (a CD28 homolog termed inducible co-stimulator), CD52 and CASP1 (caspase 1 or apoptosisrelated cysteine peptidase). Notably, the leave-one-out cross-validation (LOOCV) testing procedure system had an accuracy of 97%, specificity of 0.96, and sensitivity of 1.
Application of the personalized modeling-based gene selection method identified the characteristics of patients with acute graft-versus-host disease
The new model, created for each step, was trained (D0tr) with the personalized model for 59 runs and a LOOCV (97% for the integrated method) was calculated for the new dataset D, obtained by adding a new sample for each new run. A scenario of personalized feature weights is shown in Table 2. It is evident that HLA tissue typing, number of mismatched loci and age are relevant characteristics in the personalized approach making it possible to select a very low number of transcript genes with good classification results. Another important observation was made from our experimental computational analysis, as shown in the scenario in Table 2 for two samples representative of the whole dataset, which, however, differed in our personalized modeling for the sex variable. In our personalized analysis, age had a different weight in male and female subjects. Indeed, in female subjects, “age at transplantation” correlated with the features “donor age” and “blood group”, which were more important (higher weight) in female subjects than in male ones. Finally, we must highlight that the gene transcript subgroups: (FOXP3, CASP1) and (ICOS, CASP1), selected with the clinical variables above, returned a classification result with an accuracy of 86%.
An example demonstrating how PMGS can present the analysis results from a data sample is provided in Figure 2. For four sample-patients, the PMGS method selected three genes as the best subgroup (CASP1, FOXP3, ICOS) and (FOXP3, ICOS, CD52) and the classifier successfully predicted corresponding samples as diseased.19 It is easy to appreciate that a sample is more likely to be in the diseased group, since most of its nearest neighbors belong to the diseased group.
Gene expression assay confirmed low expression of FOXP3 and ICOS in acute graft-versus-host disease
All patients' samples identified as aGvHD (YES) or (NO) included in the validation setting were simultaneously quantified by gene expression assays for FOXP3 and ICOS transcripts. ROC curve analysis was used to identify the best informative transcripts associated with aGvHD resulting from the computational algorithm (Figure 3). Since we demonstrated that the levels of expression of FOXP3 and ICOS were very different between aGvHD YES and NO subsets, we also determined the minimum expression level that indicated the aGvHD NO condition. Specifically, the cut-off value of the relative expression level of target gene was 1.40 for FOXP3 (AUC=0.745, P=0.002) and 1.44 for ICOS (AUC=0.944, P<0.0001).
Immunohistochemical analysis confirmed the presence of low levels of FOXP3 and ICOS proteins in cutaneous biopsies from patients with acute graft-versus-host disease
In a subset of ten patients with histologically confirmed aGvHD we evaluated FOXP3 and ICOS protein expression by immunohistochemistry in formalin-fixed, paraffin-embedded skin tissue biopsies. In all cases, rare FOXP3 lymphocytes and scattered ICOS activated lymphocytes were observed at the dermo-epidermal junction, in the superficial dermis and around the adnexal follicular epithelium (Figure 4).
In our study we used dynamic genomic markers derived from peripheral blood gene expression profiling to develop an instrument for assessing aGvHD quickly and noninvasively. Not only is blood easier to obtain than tissue, but it also contains a number of circulating cell types that are mechanistically associated with immunological disease processes.23
It is interesting that in addition to the more well-known mechanisms involved in aGvHD,24 novel players have recently been discovered, such as CASP1 and FOXP3, supported by an intense body of research which has provided compelling evidence for their correlation with aGvHD.9,25 For example, a humanized monoclonal antibody against CD52 has been widely used for preventing aGvHD in allogeneic HSCT.26 Notably, CD52 signaling is correlated with induction of CD4 regulatory T cells,27 and FOXP3 is the major gene transcript representative of CD4/CD25/CD127 Treg. In our study, FOXP3 and ICOS were the most informative genes selected by PMGS. These findings were then validated by single gene expression assays, which showed significant AUC values for FOXP3 and ICOS and cut-off values suggestive of the occurrence of aGvHD. Notably, immunohistochemical analyses, performed on cutaneous biopsy specimens from aGvHD (YES) patients, showed small amounts of FOXP3 and ICOS proteins, in accordance with the observed low expression of their molecular transcripts.
In clinical practice the low expression levels of these genes could be helpful for discriminating patients who are developing aGvHD from those who are not. In addition, we observed that, in patients with aGvHD who underwent anti-aGvHD treatments, the number of copies of FOXP3 and ICOS increased in patients responding to treatment (data not shown). This phenomenon also occurred for genes regulated by ICOS, including T-helper (Th)2 cytokines (i.e.IL-4, STAT-6, IL-18). This finding is supported by the fact that Th2 cell therapy can rapidly improve severe aGvHD via IL-4- and IL-10-mediated mechanisms.28 Thus, determining the levels of FOXP3 and ICOS gene expression may represent a molecular approach to monitoring the efficacy of therapy in allogeneic HSCT recipients with aGVHD. This could be of clinical value, since the method used in this study may identify patients with a deteriorating acute syndrome, despite the administration of corticosteroids. We found that the ICOS gene was down-regulated during aGvHD in allogeneic HSCT recipients, which is a novel observation in humans. Previously, Buzzeo et al.8 observed that the human ICOS ligand was down-regulated during the onset of aGvHD, but changes in ICOS expression have never been reported.
Great attention is placed on the role of ICOS in the pathophysiology of aGvHD. ICOS is a CD28/cytotoxic T lymphocyte antigen 4 (CTLA-4) family member expressed on activated T cells.29 The role of ICOS in aGvHD is controversial. In experimental mice models, ICOS exacerbated Th1-mediated aGvHD30 while another animal study demonstrated that ICOS blockade had positive effects on inhibiting aGvHD. In contrast, Witsch showed that the ICOS signal inhibits the development of aGvHD mediated by CD8 effector cells in myeloablative bone marrow transplantation.31,32
From a clinical standpoint, the best transcript gene targets prioritized by computational systems for the assessment of immunological alloreactions in HSCT patients are also very robust since they exhibit the same behavior even in the presence of confounding factors related to sex or HLA donor-disparity, transplant conditioning and co-morbidity such as viral infection. In fact, the expression levels of the most informative genes were not influenced in the subset of patients who were positive for DNA cytomegalovirus (data not shown). We observed that this model was applicable to all cases of aGvHD including four cases of grade III aGvHD. Moreover, the level of expression of the four most informative genes showed a slight trend in correlating with non-relapse related mortality (data not shown).
The transferability of the aGvHD-linked gene-expression profiles might suggest a major epigenetic influence rather than an environmental one. However, higher-order predictive variable combinations do require the support of many more samples in order to prevent over-fitting of the model. Convincing assessment of this issue will, therefore, require expression profiling of the genes identified here in larger cohorts of participants and validation by other centers before it can be widely used to guide clinical decisionmaking processes.
In conclusion, our results could lay the basis for a novel approach in transplantation medicine by contributing to the differentiation of an immune reaction from other morbidities and to tailoring immunosuppressive regimens.
Funding: this study was supported by a grant from the Italian Government Research Program of the Ministry of the Health to PI and FM entitled: “Project of Integrated Program: Allogeneic Hemopoietic Stem Cell Transplantation in Malignant Hemopathy and Solid Neoplasia Therapy - Predictive and prognostic value for graft vs. host disease of chimerism and gene expression”.
- The online version of this article has a Supplementary Appendix.
- Authorship and Disclosures: The information provided by the authors about contributions from persons listed as authors and in acknowledgments is available with the full text of this paper at www.haematologica.org. Financial and other disclosures provided by the authors using the ICMJE (www.icmje.org) Uniform Format for Disclosure of Competing Interests are also available at www.haematologica.org.
- Received December 6, 2011.
- Revision received February 21, 2012.
- Ferrara JL. Advances in the clinical management of GVHD. Best Pract Res Clin Haematol. 2008; 21(4):677-82. PubMedhttps://doi.org/10.1016/j.beha.2008.07.003Google Scholar
- Choi SW, Levine JE, Ferrara JL. Pathogenesis and management of graft-versus-host disease. Immunol Allergy Clin North Am. 2010; 30(1):75-101. PubMedhttps://doi.org/10.1016/j.iac.2009.10.001Google Scholar
- Paczesny S, Levine JE, Braun TM, Ferrara JL. Plasma biomarkers in graft-versus-host disease: a new era¿. Biol Blood Marrow Transplant. 2009; 15(Suppl 1):33-8. PubMedGoogle Scholar
- Paczesny S, Krijanovski OI, Braun TM, Choi SW, Clouthier SG, Kuick R. A biomarker panel for acute graft-versus-host disease. Blood. 2009; 113(2):273-8. PubMedhttps://doi.org/10.1182/blood-2008-07-167098Google Scholar
- Paczesny S, Braun TM, Levine JE, Hogan J, Crawford J, Coffing B. Elafin is a biomarker of graft-versus-host disease of the skin. Sci Transl Med. 2010; 2(13):13ra2. PubMedhttps://doi.org/10.1126/scitranslmed.3000406Google Scholar
- August KJ, Chiang KY, Bostick RM, Flanders WD, Waller EK, Langston A. Biomarkers of immune activation to screen for severe, acute GVHD. Bone Marrow Transplant. 2011; 46(4):601-4. PubMedhttps://doi.org/10.1038/bmt.2010.165Google Scholar
- Baron C, Somogyi R, Greller LD, Rineau V, Wilkinson P, Cho CR. Prediction of graft-versus-host disease in humans by donor gene-expression profiling. PloS Med. 2007; 4(1):e23. PubMedhttps://doi.org/10.1371/journal.pmed.0040023Google Scholar
- Buzzeo MP, Yang J, Casella G, Reddy V. A preliminary gene expression profile of acute graft-versus-host disease. Cell Transplant. 2008; 17(5):489-94. PubMedhttps://doi.org/10.3727/096368908785096042Google Scholar
- Paczesny S, Hanauer D, Sun Y, Reddy P. New perspectives on the biology of acute GVHD. Bone Marrow Transplant. 2010; 45(1):1-11. PubMedhttps://doi.org/10.1038/bmt.2009.328Google Scholar
- IFIP Advances in Information and Communication Technology, Engineering Applications of Neural Networks. Springer-Verlag: Berlin Heidelberg; 2011. Google Scholar
- Fiasché M, Cuzzola M, Iacopino P, Kasabov N, Morabito FC. Personalized modeling based gene selection for acute GvHD gene expression data analysis: a computational framework proposed. Australian Journal of Intelligent Information Processing Systems, Machine Learning Applications. 2010; 12(4 Part II):13-8. Google Scholar
- Przepiorka D, Weisdorf D, Martin P, Klingemann HG, Beatty P, Hows J. Consensus conference on acute GVHD grading. Bone Marrow Transplant. 1994; 15(6):825-8. Google Scholar
- Rowlings PA, Przepiorka D, Klein JP, Gale RP, Passweg JR, Henslee-Downey PJ. IBMTR severity index for grading acute graft-versus-host disease: retrospective comparison with Gluckberg grade. Br J Haematol. 1997; 97(4):855-64. PubMedhttps://doi.org/10.1046/j.1365-2141.1997.1112925.xGoogle Scholar
- Song Q, Kasabov N. TWNFI–Transductive weighted neuro-fuzzy inference system and applications for personalized modelling. Neural Netw. 2006; 19(10):159-96. Google Scholar
- Kasabov N. Evolving Connectionist Systems: The Knowledge Engineering Approach. Springer: London; 2007. Google Scholar
- Kasabov N. Global, local and personalized modelling and profile discovery in bioinformatics: an integrated approach. Pattern Recognit Lett. 2007; 28(6):673-85. https://doi.org/10.1016/j.patrec.2006.08.007Google Scholar
- IFIP Advances in Information and Communication Technology, Engineering Applications of Neural Networks. Springer-Verlag: Berlin Heidelberg; 2011. https://doi.org/10.1007/978-3-642-23957-1_31Google Scholar
- LNCS, Neural Information Processing. Springer-Verlag: Berlin Heidelberg; 2009. https://doi.org/1007/978-3-642-10684-2_40Google Scholar
- Fiasché M, Cuzzola M, Irrera G, Iacopino P, Morabito FC. Advances in medical decision support systems for acute graft-versus-host disease: molecular and computational intelligence joint approaches. Front Biol. 2011; 6(4):263-73. Google Scholar
- Frontiers in Artificial Intelligence and Applications, Neural Nets WIRN11. Amsterdam Netherlands; 2011. Google Scholar
- LNCS Artificial Neural Networks – ICANN 2010. Springer-Verlag: Berlin Heidelberg; 2010. Google Scholar
- Fiasché M, Verma A, Cuzzola M, Iacopino P, Kasabov N, Morabito FC. Discovering diagnostic gene targets for early diagnosis of acute GvHD using methods of computational intelligence on gene expression data. Journal of Artificial Intelligence and Soft Computing Research. 2011; 1(1):81-9. Google Scholar
- Singh MK, Scott TF, La Framboise WA, Hu FZ, Post JC, Ehrlich GD. Gene expression changes in peripheral blood mononuclear cells from multiple sclerosis patients undergoing beta-interferon therapy. J Neurol Sci. 2007; 258(1-2):52-9. PubMedhttps://doi.org/10.1016/j.jns.2007.02.034Google Scholar
- Socié G, Blazar BR. Acute graft-versus-host disease: from the bench to the bedside. Blood. 2009; 114(20):4327-36. PubMedhttps://doi.org/10.1182/blood-2009-06-204669Google Scholar
- Spivey TL, Uccellini L, Ascierto ML, Zoppoli G, De Giorgi V, Lucia Gemma Delogu LG. Gene expression profiling in acute allograft rejection: challenging the immunologic constant of rejection hypothesis. J Transl Med. 2011; 9:174. PubMedhttps://doi.org/10.1186/1479-5876-9-174Google Scholar
- Busca A. The use of monoclonal antibodies for the treatment of graft-versus-host disease following allogeneic stem cell transplantation. Expert Opin Biol Ther. 2011; 11(6):687-97. PubMedhttps://doi.org/10.1517/14712598.2011.566852Google Scholar
- Watanabe T, Masuyama J, Sohma Y, Inazawa H, Horie K, Kojima K. CD52 is a novel costimulatory molecule for induction of CD4+ regulatory T cells. Clin Immunol. 2006; 120(3):247-59. PubMedhttps://doi.org/10.1016/j.clim.2006.05.006Google Scholar
- Foley JE, Mariotti J, Ryan K, Eckhaus M, Fowler DH. Th2 cell therapy of established acute graft-versus-host disease requires IL-4 and IL-10 and is abrogated by IL-2 or hosttype antigen-presenting cells. Biol Blood Marrow Transplant. 2008; 14(9):959-72. PubMedhttps://doi.org/10.1016/j.bbmt.2008.06.007Google Scholar
- Hutloff A, Dittrich AM, Beier KC, Eljaschewitsch B, Kraft R, Anagnostopoulos I, Kroczek RA. ICOS is an inducible T-cell co-stimulator structurally and functionally related to CD28. Nature. 1999; 397(6716):263-6. PubMedhttps://doi.org/10.1038/16717Google Scholar
- Ogawa S, Nagamatsu G, Watanabe M, Watanabe S, Hayashi T, Horita S. Opposing effects of anti-activation-inducible lymphocyte-immunomodulatory molecule/inducible costimulator antibody on the development of acute versus chronic graftversus-host disease. J Immunol. 2001; 167(10):5741-8. PubMedhttps://doi.org/10.4049/jimmunol.167.10.5741Google Scholar
- Witsch EJ, Peiser M, Hutloff A, Büchner K, Dorner BG, Jonuleit H. ICOS and CD28 reversely regulate IL-10 on re-activation of human effector T cells with mature dendritic cells. Eur J Immunol. 2002; 32(9):2680-6. PubMedhttps://doi.org/10.1002/1521-4141(200209)32:9<2680::AID-IMMU2680>3.0.CO;2-6Google Scholar
- Taylor PA, Panoskaltsis-Mortari A, Freeman GJ, Sharpe AH, Noelle RJ, Rudensky AY. Targeting of inducible costimulator (ICOS) expressed on alloreactive T cells down-regulates graft-versus-host disease (GVHD) and facilitates engraftment of allogeneic bone marrow (BM). Blood. 2005; 105(8):3372-80. PubMedhttps://doi.org/10.1182/blood-2004-10-3869Google Scholar