Research ArticleDrug Sensitivity

Drug Reaction with Eosinophilia and Systemic Symptoms (DRESS): A Multiorgan Antiviral T Cell Response

See allHide authors and affiliations

Science Translational Medicine  25 Aug 2010:
Vol. 2, Issue 46, pp. 46ra62
DOI: 10.1126/scitranslmed.3001116


Drug reaction with eosinophilia and systemic symptoms (DRESS) is a severe, drug-induced reaction that involves both the skin and the viscera. Evidence for reactivation of herpes family viruses has been seen in some DRESS patients. To understand the immunological components of DRESS and their relationship to viral reactivation, we prospectively assessed 40 patients exhibiting DRESS in response to carbamazepine, allopurinol, or sulfamethoxazole. Peripheral blood T lymphocytes from the patients were evaluated for phenotype, cytokine secretion, and repertoire of CD4+ and CD8+ and for viral reactivation. We found Epstein-Barr virus (EBV), human herpes virus 6 (HHV-6), or HHV-7 reactivation in 76% of the patients. In all patients, circulating CD8+ T lymphocytes were activated, exhibited increased cutaneous homing markers, and secreted large amounts of tumor necrosis factor–α and interferon-γ. The production of these cytokines was particularly high in patients with the most severe visceral involvement. In addition, expanded populations of CD8+ T lymphocytes sharing the same T cell receptor repertoire were detected in the blood, skin, liver, and lungs of patients. Nearly half of these expanded blood CD8+ T lymphocytes specifically recognized one of several EBV epitopes. Finally, we found that the culprit drugs triggered the production of EBV in patients’ EBV-transformed B lymphocytes. Thus, cutaneous and visceral symptoms of DRESS are mediated by activated CD8+ T lymphocytes, which are largely directed against herpes viruses such as EBV.


Drug-induced adverse reactions that manifest on the skin affect 2 to 3% of hospitalized patients who are taking medications (1). Some of these may be severe or life-threatening as a result of extensive cutaneous lesions or visceral involvement (2, 3). The pathogenesis of these drug-induced cutaneous adverse reactions remains poorly understood, hampering the development of specific treatments. They are thought to be caused by an immune mechanism, triggered when drugs or their metabolites activate T lymphocytes (46). Other causal factors such as drug metabolites, genetic factors (7, 8), and viral infections have also been suggested (912).

Drug reaction with eosinophilia and systemic symptoms (DRESS), also called drug-induced hypersensitivity syndrome (DIHS), is a drug-induced adverse reaction that is serious and well characterized. Patients exhibit skin eruptions, fever, facial edema, polyadenopathy, and visceral involvement of the liver, lungs, kidneys, and heart. Blood abnormalities include eosinophilia, lymphocytosis, and/or mononucleosis. The main drugs that cause DRESS are anticonvulsants, the anti-gout medication allopurinol, and the antibiotics minocycline and the sulfamides (6). There is usually a delay, between 2 and 6 weeks, from the beginning of drug intake to the onset of clinical symptoms; a prolonged course follows, with frequent flare-ups and relapses over weeks or even months after stopping the culprit drug. Reactivation of viruses of the herpes family [human herpes virus 6 (HHV-6), HHV-7, Epstein-Barr virus (EBV), and cytomegalovirus (CMV)] has been reported to occur at the onset of DRESS (912).

To understand the relationship between viral reactivation and the immunological mechanisms of DRESS, we prospectively evaluated the clinical and biological features of 40 patients presenting with well-characterized DRESS. We demonstrated that the cutaneous and visceral symptoms of DRESS are associated with an oligoclonal proliferation of activated CD8+ T lymphocytes that are directed against viral antigens derived from herpes viruses such as EBV and whose replication is enhanced by the culprit drug.



We assessed 40 consecutive patients (19 males and 21 females) with DRESS, whose characteristics are indicated in Table 1. Median age was 56 years (range, 18.3 to 81.7 years). No association was found among human leukocyte antigen (HLA) I typing, ethnicity, and the culprit drug (table S1). The median time between the beginning of the culprit drug intake and the onset of DRESS was 26.5 days (range, 7 to 75 days). Median time between the onset of DRESS and the inclusion in the study (day 0) was 10 days (range, 2 to 86 days). Twenty-three patients (57%) showed a deterioration of their general condition, as defined by a Karnofsky score lower than 50%. All but one patient had an increase in alanine aminotransferase and/or aspartate aminotransferase serum concentrations, and 10 patients (25%) had renal problems, including proteinuria and/or increased creatinine serum concentrations. The most frequent abnormalities in blood cell counts were eosinophilia in 32 cases (80%), monocytosis in 18 cases (45%), and mononucleosis in 14 cases (35%).

Table 1

Clinical and biological characteristics of the 40 DRESS patients at baseline and during follow-up. Numbers (%) are given except where indicated.

View this table:

Eleven patients exhibited severe visceral involvement, including nine patients with interstitial pneumonitis and two patients with severe hepatitis and hepatic failure, one of whom underwent a liver transplantation. Three patients died at days 15, 20, and 30, respectively. Causes of death were severe infections (endocarditis or septicemia) in two patients, considered a DRESS complication, and stroke in the third case, considered unrelated to DRESS (3).

A prolonged course of DRESS after stopping the culprit drug was frequently observed; abnormalities were still present up to 180 days (Table 1).

Immunological and molecular analysis

Baseline analyses CD8+ blood T cell numbers were greater in DRESS patients than in healthy controls by a factor of 2.5 (P = 0.01) (fig. S1A). At day 0, CD8+ blood T lymphocytes had an activated phenotype with overexpression of the CD25 (P = 0.0002), CD69 (P < 0.0001), and HLA-DR (P < 0.0001) activation markers when compared with lymphocytes from healthy controls (fig. S1B). Only CD69 (P = 0.004) and HLA-DR (P < 0.001) were overexpressed in CD4+ T cells from DRESS patients (fig. S1D). In addition, CD8+ T lymphocytes from the patients overexpressed the cutaneous homing receptors CLA (cutaneous lymphocyte–associated antigen) (P < 0.0001) and CCR4 (C-C chemokine receptor type 4) (P < 0.0001), and cytotoxic markers such as Fas ligand (FasL) (P = 0.0003), when compared to healthy controls (fig. S1B). CD8+ T lymphocytes also contained more of the cytokines tumor necrosis factor–α (TNF-α) (P = 0.0287), interferon-γ (IFN-γ) (P = 0.0165), and interleukin-2 (IL-2) (P = 0.0283) when compared to healthy controls (fig. S1C). In contrast, blood CD4+ T lymphocytes from DRESS patients did not secrete more cytokines or express more homing markers than controls (fig. S1, D and E). IL-5–secreting T lymphocyte number was not increased in DRESS patients, a result we found unexpected because eosinophilia observed in DRESS has been reported to be related to IL-5 (13).

Next, we used the immunoscope method, in which BV [variable domain of the β chain of the T cell receptor (TCR)] usage and the CDR3 (complementary-determining region 3) length in the TCR β chain were assessed, to characterize expanded T cell populations in CD4+ and CD8+ blood T lymphocytes. T cell repertoire analysis showed oligoclonal expansions in 39 of the 40 DRESS patients, preferentially among CD8+ T lymphocytes (table S1). We saw no common repertoire bias, as defined by common BV usage and CDR3 length among patients with the same culprit drug (table S1). Using the results of the immunoscope analysis, we then determined the phenotype and cytokine secretion of expanded (corresponding to the BV expansions in table S1) and nonexpanded CD8+ T lymphocytes using antibodies specific to BV (Fig. 1). The proportion of IFN-γ– and TNF-α–secreting cells was larger in the expanded CD8+ populations than in the nonexpanded CD8+ populations by a factor of 1.7, with a Gaussian repertoire (P = 0.048) (Fig. 1). Secretion of larger amounts of TNF-α was confirmed by enzyme-linked immunosorbent assays (ELISAs) of patients’ serum [median, 92 pg/ml (11 to 230 pg/ml)]; TNF-α was not detected in serum from healthy controls.

Fig. 1

Production of IFN-γ and TNF-α by expanded and nonexpanded CD8+ T lymphocytes in DRESS patients. Using the immunoscope technique, we analyzed the baseline CD8+ expanded T lymphocytes for those patients who showed a peak in their repertoire (corresponding to the BV expansions shown in table S1) and the nonexpanded T lymphocyte populations, which showed a Gaussian repertoire. (A and B) The percentage of IFN-γ–positive (A) and TNF-α–positive (B) CD8+ T lymphocytes in expanded (n = 5) and nonexpanded (n = 5) T lymphocyte populations was determined by FACS with the corresponding antibodies to BV. Data from patients 8, 9, 10, 12, and 14 are shown. Controls included 30 patients taking drugs or have taken drugs without side effects and without inflammatory disease at study time (control group A) (n = 30). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.005.

To further investigate the specificity and the functional abnormalities of T lymphocytes at the onset of DRESS, we analyzed, using microarrays, gene expression in peripheral blood mononuclear cells (PBMCs) and CD4+ and CD8+ blood T lymphocytes from DRESS patients (14). CD8+ cells from DRESS patients exhibited increased transcription of TNF-α and IFN-γ genes, in accordance with the results of fluorescence-activated cell sorting (FACS) analysis (table S2A). In addition, we noted an up-regulation of genes implicated in apoptosis, lymphocyte activation, proinflammatory cytokine production, and cutaneous homing (table S2A). IL-17 transcription was higher in patients with DRESS than in healthy controls treated with carbamazepine by a factor of ~6 (table S2B). IL-5 gene transcription did not increase, raising the possibility that the IL-17 family member IL-17E, rather than IL-5, could be contributing to the patients’ blood eosinophilia.

Follow-up analyses Because 52% of patients still had clinical or biological abnormalities 90 days after drug withdrawal, we studied cytokine production at these later times. TNF-α– and IFN-γ–producing cells increased from control values until day 90 and then progressively decreased until day 360 (fig. S1C). Nevertheless, microarray analysis showed a down-regulation of TNF-α and IFN-γ receptor gene expression in PBMCs at day 90 compared to day 0 (table S2B), possibly explaining the clinical improvement in patients despite a persistent high production of TNF-α and IFN-γ by CD8+ cells after drug withdrawal.

Clinical correlations We then compared the cytokine production profile between patients with nonsevere (without pulmonary involvement or hepatic failure) or severe visceral involvement, as defined by the presence of interstitial pneumonitis or hepatic failure. At day 0, the numbers of CD8+ T lymphocytes producing IL-2 (P = 0.014), TNF-α (P = 0.011), and IFN-γ (P = 0.014) were significantly higher in patients with severe visceral involvement than in those with nonsevere visceral involvement (Fig. 2).

Fig. 2

Cytokine production by CD8+ lymphocytes from DRESS patients according to severity of visceral involvement. (A to C) The number of CD8+ T lymphocytes producing TNF-α (A), IL-2 (B), and IFN-γ (C) was compared in DRESS patients with severe visceral involvement (n = 7, solid lines) and DRESS patients without pulmonary involvement and with a moderate liver cytolysis without hepatic failure (n = 12, dashed lines). Data are expressed as the number of positive cells per cubic millimeter of blood. *P ≤ 0.05, **P ≤ 0.01.

Viral reactivation and antiviral CD8+ T cell responses To assess viral reactivation in patients with DRESS, we first studied viral replication during the onset of DRESS at day 0 and during follow-up. Because previous studies had shown HHV-6 replication during the acute phase of DRESS (912), we replaced HHV-6 and other herpes-related viruses, such as EBV, CMV, and HHV-7, in serum and PBMCs from DRESS patients. Among the 38 patients we analyzed, a reactivation of EBV, HHV-6, or HHV-7 was detected between day 0 and day 30 in 16 cases (42%), 17 cases (45%), and 12 cases (32%), respectively (table S3). Thirteen (34%) had multiple viral reactivations, 12 of them with EBV and HHV-6 or HHV-7. Overall, 29 patients (76%) showed a reactivation of HHV-6, HHV-7, or EBV, whereas no EBV, HHV-6, or HHV-7 reactivation was detected in 50 healthy control subjects (table S3). A recall of immunoglobulin M (IgM) immune response against HHV-6 was observed in 16 patients (42%). On the contrary, no reactivation of CMV was seen in DRESS patients at day 0 or during follow-up. EBV viral DNA was detected in the PBMCs of all DRESS patients and of 40% of the control subjects.

Because many clinical features of DRESS patients were reminiscent of those found in viral infections, we grouped the DRESS patients according to HHV-6 and/or EBV reactivation. Patients with viral reactivation of HHV-6 or EBV at day 0 (n = 14, patients 1, 2, 6, 10, 12, 13, 17, 19, 21, 22, 23, 30, 31, and 38) more frequently had facial edema (100% versus 70%, P = 0.03), lymphadenopathies (100% versus 58%, P = 0.01), and monocytosis (1740 cells/mm3 versus 730 cells/mm3, P = 0.01) than did patients without (table S1).

To further characterize the antigens recognized by T lymphocytes from DRESS patients, we sequenced the BV CDR3 region of the expanded CD8+ T lymphocytes previously identified by the immunoscope technique. Four hundred and nine sequences from 19 CD8+ T lymphocyte expansions were obtained from 10 patients selected according to the main culprit drugs. Comparison of these nucleic acid sequences with those available in databases or published in the literature (1522) showed that 17 of the 19 populations of expanded CD8+ T lymphocyte contained cells with BV CDR3 sequences homologous to EBV-specific TCRs (table S4B). On average, 47% (range, 10 to 90%) of the sequences obtained from CD8+ expanded lymphocytes from DRESS patients had homology with CDR3 regions of EBV-specific CD8+ T lymphocytes (table S4A), suggesting that there is an EBV-driven selection of the CD8+ T lymphocyte response in many patients with DRESS. CDR3 homologies with EBV-specific CD8+ T lymphocytes were also observed in DRESS patients without evidence of EBV reactivation. Comparison of CDR3 region sequences with sequences of HHV-6– and HHV-7–specific T lymphocytes was not possible because the latter are still unknown.

To further investigate the hypothesis that DRESS is a result of an anti-EBV T lymphocyte response, we tested by FACS the specificity of blood T lymphocytes from DRESS patients against EBV antigens with HLA class I tetramers loaded with EBV immunodominant peptides. Because the well-characterized immunodominant EBV peptides are restricted to particular HLA molecules, namely, HLA-A*0201 and HLA-B*0702, tetramer analysis was only possible on patients with HLA-A*0201 and HLA-B*0702. Tetramer analysis was performed in six patients in the HLA-A*0201 group and in two patients in the HLA-B*0702 group. T lymphocytes specific for BMFL1, LMP2, and EBNA peptides of EBV represented between 1 and 21% of total blood CD8+ T lymphocytes from DRESS patients compared with <0.1% in 50 healthy controls with HLA-A*0201 and HLA-B*0702 (Fig. 3 and table S5). These EBV-specific T lymphocytes were found in patients not only with verified EBV reactivation but also without EBV reactivation (table S5).

Fig. 3

Recognition of EBV-derived peptides by CD8+ T lymphocytes from representative DRESS patients. Tetramers were used to detect EBV-specific CD8+ T cells. HLA-A*0201 and HLA-B*0702 were folded with synthetic peptides derived from EBV antigens (GLCTLVAML from BMLF1, CLGGLLTMV from LMP2, and RPPIFIRRL from EBNA3A). Patients 12, 17, and 27 with DRESS were selected because they shared HLA-A2 and/or HLA-B7. An example from an HLA-A2/HLA-B7 healthy control patient is shown.

Anti-EBV CD8+ T lymphocytes in skin lesions and organs of patients with DRESS To confirm the presence of blood EBV-specific expanded CD8+ T lymphocytes in the skin and liver from DRESS patients, we analyzed the immunoscope profile and sequenced the CDR3 region of T lymphocytes from the skin of three patients and the liver of one patient (patient 37) who underwent liver transplantation for hepatic failure. CD8+ T lymphocytes with the same immunoscope profile and an identical CDR3 sequence sharing homology with EBV-specific T lymphocytes were found in the blood, skin, and liver from DRESS patients (Fig. 4, A and B, and table S4B). In addition, we confirmed the presence of these T lymphocytes in the liver with immunochemistry using monoclonal antibodies specific to BV (Fig. 4B). In an additional DRESS patient with pulmonary involvement, we confirmed the presence of the blood-expanded EBV-specific CD8+ T lymphocytes in blood, skin, liver, and lung with immunohistochemistry using monoclonal antibodies specific to BV (Fig. 4C).

Fig. 4

EBV-specific CD8+ T lymphocytes in blood, skin lesions, liver, and lungs of DRESS patients. We analyzed the immunoscope profile and then sequenced the CDR3 region of expanded T lymphocyte in blood, skin, and involved organs of three DRESS patients: (A) in blood and skin of patient 38 and (B) in blood and liver of patient 37. (Far right panel) Immunohistochemical staining of liver from patient 37 with monoclonal antibody to BV9. (C) Immunoscope profile (left panels) and immunohistochemical staining of skin, lung, and bronchoalveolar lavage with monoclonal antibodies to BV17, BV18, and BV20 (right panels) for patient 22, who presented with a severe pulmonary involvement. We performed immunochemistry on EBV-expanded, EBV-specific CD8+ lymphocytes with monoclonal antibodies to BV9, BV17, BV18, and BV20. No T lymphocytes were detected in skin, liver, or lungs from patients without DRESS using immunochemistry with monoclonal antibodies to BV9, BV17, BV18, and BV20.

Involvement of culprit drugs

Because individual culprit drugs in DRESS are not associated with particular T lymphocyte expansions or immunological and clinical features, we hypothesized that the culprit drugs play a direct role in viral reactivation. We then used an in vitro model based on immortalized B lymphocytes obtained by EBV transformation. EBV-transformed B lymphocytes derived from DRESS patients and controls were used to study the specific effect of drugs on EBV reactivation. We incubated EBV-transformed B lymphocytes from four patients with carbamazepine-, sulfamethoxazole-, and allopurinol-induced DRESS with their corresponding culprit drugs and with the three other major DRESS inducers, namely, carbamazepine, sulfamethoxazole, and allopurinol. An increase in EBV production in the supernatant and in the intracellular compartment of these cells was observed not only with the culprit drug but also with the three other main DRESS inducers (Fig. 5A and fig. S2). On the contrary, no change in EBV production was observed when EBV-transformed B lymphocytes from DRESS patients were incubated with gentamicin, an irrelevant drug that has not been involved in the occurrence of DRESS (Fig. 5A). Carbamazepine, sulfamethoxazole, and allopurinol did not induce an increase in EBV production in EBV-transformed B lymphocytes from four healthy individuals (Fig. 5B). Because switching to a structurally unrelated anticonvulsant may induce flare-ups of DRESS, we also incubated EBV-transformed B lymphocytes from two patients with carbamazepine-induced DRESS with valproic acid, which is not structurally related to carbamazepine (23, 24). Valproic acid induced EBV production in the intracellular compartment and in the supernatant of EBV-transformed B lymphocytes from patients with carbamazepine-induced DRESS (Fig. 5C).

Fig. 5

Induction of EBV by drugs. (A) EBV-transformed B lymphocytes from patients 27, 31, 38, and 40 who have DRESS caused by, respectively, allopurinol (Allo), carbamazepine (CBZ), and sulfamethoxazole (SMX) were incubated with the culprit drug at different concentrations for 72 hours (black columns). EBV production was measured in the intracellular compartment (left) and in the supernatant (right) and compared to EBV-transformed B cells from the same patient incubated without the culprit drug (open bars) or an irrelevant drug such as gentamicin (Genta) (gray bars). (B) EBV-transformed B lymphocytes from four healthy control subjects (group A) were incubated with DRESS-inducer drugs, with an irrelevant drug (gentamicin), or without drug for 72 hours. EBV production was measured in the intracellular compartment (left) and in the supernatant (right). (C) EBV-transformed B lymphocytes from patients 31 (black bars) and 38 (gray bars), who presented with carbamazepine-induced DRESS, were incubated with valproic acid (VPA) at different concentrations. Incubation without drug or with carbamazepine served as negative and positive controls, respectively. EBV production was measured in the intracellular compartment (left) and in the supernatant (right). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.005. Results are representative of two independent experiments (triplicate assays).


Our data show that the adverse drug reaction DRESS seen after administration of some anticonvulsant drugs, antibiotics, or anti-gout medication is a result of cutaneous and systemic manifestations of an immune response, mainly mediated by CD8+ T lymphocytes directed against herpes virus antigens. The clinical and biological features of DRESS are consistent with viral infection: fever, edema of the face, lymphadenopathies, monocytosis, mononucleosis, and hepatitis (25, 26). Indeed, reactivation of viruses of the herpes group (HHV-6, HHV-7, EBV, and CMV) has been repeatedly reported in patients with DRESS (912). In our experiments, we noted reactivation of EBV, HHV-6, and/or HHV-7 in 76% of the DRESS patients. Viral reactivation was seen at the onset of DRESS, consistent with it being an early event in DRESS development. As a consequence of this increased viral presence, EBV-specific CD8+ T lymphocytes that secreted TNF-α and IFN-γ were found both in blood and in involved organs including skin, liver, and lungs. Consistent with this localization, we found that these cells expressed high levels of CLA- and CCR4-homing markers, which allow retention of T lymphocytes in skin and lungs (27).

EBV-specific T lymphocytes were detected in patients not only with EBV reactivation but also without evidence of EBV reactivation. All DRESS patients we tested mounted an abnormal CD8+ T lymphocyte response against EBV, probably related to the presentation of EBV epitopes to the immune system (25). EBV viral DNA was detected in the PBMCs of all DRESS patients, indicating the presence of EBV in all patients. These data support the hypothesis that the culprit drugs may induce reactivation and antigenic presentation of quiescent forms of EBV or other herpes viruses in cells such as B lymphocytes, which could secondarily trigger a multiorgan immune response directed against herpes viruses (25). The viral reactivation could result in a severe immune response only in susceptible people as a result of immune hypersensitivity.

Our in vitro data are consistent with this notion. We observed an increased production of EBV when we incubated EBV-transformed blood B lymphocytes from patients with sulfamethoxazole-, carbamazepine-, and allopurinol-induced DRESS with the corresponding culprit drug and with other unrelated DRESS-inducer drugs. Consistent with these findings and previous reports showing that valproic acid, an anticonvulsant, was able to induce in vitro the production of HHV-6 and EBV (28, 29), we found that valproic acid induced EBV production in EBV-transformed blood B lymphocytes from patients with carbamazepine-induced DRESS. These results are in accord with the fact that anticonvulsant drugs such as valproic acid and carbamazepine inhibit histone deacetylase, which promotes EBV reactivation (30, 31). Whereas the mechanism of EBV reactivation induced by allopurinol and sulfamethoxazole remains unknown and speculative, these drugs may bring about viral proliferation through specific interactions with enzymes that regulate gene transcription of the herpes virus family. We found that drugs that induce DRESS increase viral release by EBV-transformed cells in a dose-dependent manner. These results are consistent with the observation, made in patients taking the anticonvulsant lamotrigine, that the occurrence of skin rashes is related to the drug dose, suggesting a nonallergic or hypersensitivity mechanism (32, 33). Nevertheless, the basis for the susceptibility of affected individuals is still unknown.

The development of an immune response directed against viruses whose replication and presentation to the immune system is induced or increased by drugs might explain some of the characteristics of DRESS: (i) the long delay before onset after the beginning of drug intake, (ii) the prolonged course after stopping the culprit drug, and (iii) the occurrence of flare-ups of DRESS after switching from a culprit anticonvulsant to another structurally unrelated anticonvulsant.

The TNF-α and IFN-γ secreted by the anti-EBV CD8+ T lymphocytes in cutaneous and visceral lesions likely contribute to disease in DRESS patients. This conclusion is supported by our finding that patients with severe systemic involvement had higher numbers of blood CD8+ T lymphocytes producing IFN-γ, TNF-α, and IL-2 than did patients with less severe systemic disease. In addition, the production of large amounts of TNF-α and IFN-γ by CD8+-activated T lymphocytes is sustained for more than 3 months, possibly accounting for the frequently prolonged course of DRESS after drug withdrawal. The persistent production of high amounts of IFN-γ and TNF-α was associated with down-regulation of the corresponding cytokine receptor gene transcriptions at day 90, possibly a result of a regulatory control loop. We did not observe enhanced transcription of the IL-5 gene or high numbers of IL-5–producing lymphocytes as previously reported to explain blood eosinophilia (13). Rather, we found increased transcription of the IL-17 gene family, which may contribute to blood eosinophilia and pulmonary involvement in patients with DRESS. Indeed, IL-17E transgenic mice have a phenotype similar to that of DRESS patients, including blood eosinophilia, lymphocytosis, lymphadenopathies, and peribronchial inflammation (33, 34).

Overall, this study documents the immune response during DRESS and suggests a mechanism for its pathogenesis. This mechanism is in accord with previous observations that suggest a role for some viruses in DRESS (912) and explains its main clinical and biological features. We thus propose that herpes virus reactivation be tested in patients with possible DRESS and that this criterion be added to the DRESS definition. Corticosteroids, used for their immunosuppressive properties, are the first treatment for severe DRESS patients, but numerous patients relapse when corticosteroids are stopped or tapered, possibly because these drugs may prolong or promote herpes virus reactivation. New studies need to be performed to understand the relationship among drug hypersensitivity, virus reactivation, and immune susceptibility in drug-induced adverse reactions.

Materials and Methods

Patients and sample collection

The study was approved by the Ethics Committee of the Seine Maritime district in France and registered in Clinical Trials (NCT 00213447). Written informed consent was obtained from all patients before inclusion. The prospective study was carried out from November 2002 to December 2005 in 26 dermatological centers in France. Forty consecutive patients with clinical and histological features of DRESS were included when the following criteria of the REGISCAR study were met (36): (i) age older than 18 years; (ii) cutaneous eruption that occurred after starting the intake of a drug known to induce DRESS; (iii) visceral involvement including lymph node enlargement, hepatitis, interstitial pneumonia, and renal or cardiac failure; and (iv) at least one of the following biological abnormalities: eosinophilia (higher than 700 cells/mm3), lymphocytosis (higher than 4000 cells/mm3), or mononucleosis (higher than 1000 cells/mm3). HIV-infected patients and patients who received steroids before the study were excluded. The culprit drug was determined with the REGISCAR criteria. Patients were evaluated during a 12-month follow-up after the occurrence of DRESS. Clinical and biological evaluations were performed at each follow-up visit on days 0, 15, 30, 90, 180, and 360. A skin biopsy was performed at the onset of DRESS in all patients. A liver biopsy specimen was obtained in one patient with hepatic failure when she underwent liver transplantation (patient 37). Bronchial biopsies and a bronchoalveolar lavage were performed in one patient with severe pulmonary involvement (patient 22). Controls included a pool of 50 patients taking drugs without rash and were selected to resemble the distribution of cases with respect to drug intake, ethnicity, age, and gender; 30 patients taking drugs or have taken drugs without side effects and without inflammatory disease at study time (control A); and 20 patients with evolutive inflammatory disease (cancer, infection, or inflammatory disease) (control B) (table S6).

Delineation of TCR gene utilization repertoire by immunoscope analysis

RNA was extracted from 1 × 106 to 10 × 106 sorted CD4+ and CD8+ blood T cells with GenElute Mammalian Total RNA kit (Sigma). Reverse transcription was performed on 10 μl of the RNA solution in a final volume of 40 μl for 60 min at 37°C. The reaction was stopped by heating at 95°C for 5 min. Polymerase chain reaction (PCR) primers specific for the variable and constant domains of the TCR families were used to amplify 24 BV chains. The PCR conditions consisted of denaturation at 95°C for 30 s, annealing at 60°C for 60 s, and extension at 72°C for 60 s for 30 cycles, followed by a 5-min final extension at 72°C. β-Actin was also amplified as a control for complementary DNA integrity. A second five-cycle PCR was performed with the fluorescent primer Cβ. After purification on Sephadex G50 (Sigma), 2 μl of the purified product was analyzed for their length with a sequencer (AB3100 DNA sequencer). The fluorescence intensity of each band was quantified with Genescan software, and the presence of an expanded lymphocyte population was determined (37, 38).

Phenotype and cytokine analysis

Phenotype and cytokine secretion profile by CD4+ and CD8+ peripheral blood T lymphocytes were assessed on the 25 first included consecutive DRESS patients. Freshly isolated patients and control PBMCs were studied by FACS (FC500, Beckman Coulter) with monoclonal antibodies directed against activation markers (such as CD27, CD25, CD26, CD38, CD69, CD70, CD71, CD134, CD45RA, CD45RO, and HLA-DR), cutaneous homing receptors (such as CLA and CCR4), and cytotoxic markers (such as FasL). Stimulated PBMCs with phorbol 12-myristate 13-acetate–ionomycin for 6 hours in the presence of brefeldin A were fixed and permeabilized with IntraPrep (Beckman Coulter) and incubated with the following antibodies for analysis of intracytoplasmic cytokines: IL-2, IL-5, IFN-γ, TNF-α, IL-4, IL-8, IL-13, and IL-10 (Beckman Coulter). To characterize the phenotype and cytokine secretion of expanded blood T lymphocytes, we first determined by immunoscope the repertoire of T lymphocytes on day 0 and subsequently characterized these expanded T cell populations by FACS using the corresponding monoclonal antibodies to BV. Results were analyzed with FlowJo 7.2.4 (Tree Star).

Tetramer analysis

Soluble major histocompatibility complex (MHC) peptides were produced in Escherichia coli. Briefly, the recombinant soluble part of HLA-A*0201, HLA-B*0702, and β2-microglobulin was folded with the following synthetic peptides derived from EBV antigens: GLCTLVAML from BMLF1, CLGGLLTMV from LMP2, and RPPIFIRRL from EBNA3A. Soluble MHC was purified by gel filtration on Superdex 75 (Pharmacia) and biotinylated overnight with BirA enzyme (Avidity). Biotinylated complexes were further purified with gel filtration and anion exchange on a Resource Q column (Pharmacia). Tetramerization was performed by gradually adding tetramer-grade streptavidin-phycoerythrin (Molecular Probes) to MHC at a 1:4 molecular ratio. EBV-specific T cells (CD8+tetramer+CD3+) were identified by FACS. Controls included 50 healthy subjects with HLA-A*0201 and/or HLA-B*0702.

Microarray analysis

Patients with carbamazepine- or minocycline-induced DRESS were pooled separately for analysis. RNA from total PBMCs and CD4+- and CD8+-sorted blood T cells was amplified. Expression of 894 genes of immunological interest was analyzed with PIQOR TM arrays (Miltenyi). Gene expression was individually compared between day 0 and day 90 in the minocycline-induced DRESS patients. Control groups included patients taking carbamazepine and healthy patients without medication. The data discussed here have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE23015 (14).

CDR3 sequence

Amino acid sequences of TCRβ V(D)J junctional regions (CDR3) of expanded CD8+ T cells identified by the immunoscope technique in the blood, skin, and liver of patients with DRESS were determined. TCRβ rearranged sequences were amplified with one of the 24 TCRBV family–specific oligonucleotides and a 3′ TCRBC constant as described (37, 38). The PCR products were cloned with a TOPO TA Cloning kit (Invitrogen) and sequenced with a BigDye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems). The amino acid–converted CDR3 sequences obtained were compared to those in databases and to published CDR3 sequences (1522).

Viral reactivation analysis

Viral DNA from EBV, CMV, HHV-6, and HHV-7 was quantified by real-time PCR in the serum and PBMCs of the patients at days 0, 15, 30, 90, 180, and 360 and in PBMCs from 50 healthy controls from a blood donor center. For each sample positive for HHV-6, a determination of the variant of HHV-6 was performed. Sera from each patient were examined for the presence of IgG and IgM against CMV, of IgG against EBNA peptide from EBV by ELISA, and of IgG and IgM antibodies against HHV-6 by immunofluorescence assay. For each virus, the presence of viral DNA as detected by PCR in serum and/or a high viral load found in PBMCs (≥1500 copies per microgram of DNA) was scored as viral reactivation. The presence of IgM against HHV-6 alone without HHV-6 detection by PCR in PBMC or serum was not scored as HHV-6 reactivation. The detection of IgG against EBNA was considered to be evidence against an EBV primary infection (3941).

Induction of EBV production by the culprit drug

PBMCs from four healthy controls from group A and from four patients with carbamazepine-, sulfamethoxazole-, or allopurinol-induced DRESS were collected at days 180 or 360 and isolated by Ficoll-Paque density gradients (Amersham Pharmacia). A B cell–enriched fraction was prepared by rosetting with sheep red blood cells (Eurobio). EBV-transformed B lymphoblastoid cell lines (B-LCLs) were generated with supernatant from the marmoset B cell line B95-8 (from D. Gilbert, University of Rouen). B-LCLs (5 × 106) were incubated for 72 hours with carbamazepine (10 and 50 μg/ml), sulfamethoxazole (0.5 and 1 mg/ml), or allopurinol (10 and 50 μg/ml), and the virus load was quantified both intracellularly and in the supernatant with the LightCycler EBV Quantification kit (Roche). Controls included incubation of cells without drug and with an irrelevant drug such as gentamicin (10 and 50 μg/ml), which does not induce DRESS. In addition, B-LCLs derived from patients presenting with carbamazepine-induced DRESS were incubated with valproic acid (0.3, 1, and 2 mM).

Statistical analysis

The sample size of 40 patients was calculated a priori to be sufficient to represent the main drugs involved in DRESS. With 40 patients, there was an 87.2% chance that a drug accounting for 5% of all DRESS cases could be represented in at least one patient and a 98.5% chance that a drug accounting for 10% of all DRESS cases would be represented in at least one patient. This sample size also ensured a maximal width of ±0.15 for 95% confidence intervals in the proportions of DRESS patients with any given dichotomous characteristic. Fisher’s exact test and Mann-Whitney nonparametric test were used for two-group comparisons of dichotomous and continuous characteristics, respectively. A two-sided P value of <0.05 was considered significant. In view of the increased levels observed in DRESS patients relative to controls, a one-sided 0.05 type I error was used to detect increased concentrations of cytokines in DRESS patients with severe visceral involvement. SAS software version 9.1 (SAS Institute) was used.

Supplementary Material

Table S1. Culprit drugs, visceral involvement, blood abnormalities, viral reactivation, HLA I typing, and CD8+ T cell repertoire analysis of the 40 DRESS patients.

Table S2. Microarray analysis of DRESS patients.

Table S3. Viral reactivation analysis of the DRESS patients and healthy controls.

Table S4. CDR3 sequences obtained from expanded populations of DRESS patients that present homology with CDR3-specific EBV sequences.

Table S5. Recognition of EBV-derived peptides by CD8+ T lymphocytes from DRESS patients and healthy controls.

Table S6. Clinical characteristics of controls.

Fig. S1. Follow-up of the CD4+ and CD8+ T cell counts, phenotype, and cytokine secretion.

Fig. S2. Induction of EBV by DRESS inducer drugs.


  • * These authors contributed equally to this work.

  • Citation: D. Picard, B. Janela, V. Descamps, M. D'Incan, P. Courville, S. Jacquot, S. Rogez, L. Mardivirin, H. Moins-Teisserenc, A. Toubert, J. Benichou, P. Joly, P. Musette, Drug reaction with eosinophilia and systemic symptoms (DRESS): A multiorgan antiviral T cell response. Sci. Transl. Med. 2, 46ra62 (2010).

References and Notes

  1. Acknowledgments: We thank J. C. Roujeau for critical review of the manuscript and inclusion of patients; V. Bajzik for technical help; P. Loiseau for HLA typing; and J. M. Bonnetblanc, B. Milpied, S. Barbarot, A. Barbaud, F. A. Legal, J. Chevrant Breton, and P. Bernard for including patients in the study. We also thank R. Medeiros for editing the manuscript and V. Ferranti for the monitoring of the study. Funding: The project was supported by the French Society of Dermatology, a national program of clinical research (Programme Hospitalier de Recherche Clinique), and the René Touraine Foundation. Author contributions: D.P. and B.J. performed immunological experiments (repertoire, ELISA, flow cytometry, and experiments with EBV-transformed cells) and data analysis and co-wrote the manuscript. S.J. supervised the flow cytometry analysis. P.C. performed immunohistochemical experiments. S.R., L.M., and V.D. performed viral PCRs and data analysis. H.M.-T. and A.T. performed tetramer experiments. J.B. performed the statistical analysis. M.D. helped to design the study and helped with the clinical analysis. P.J. and P.M. designed the study, provided overall expertise into interpretation of the experiments and analysis, and co-wrote the paper. Competing interests: The authors declare that they have no competing interests. Accession numbers: This study is registered as clinical trial NCT 00213447.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article