The thiopurine analogues azathioprine (AZA) and 6-mercaptopurine (6-MP) have been used extensively in the treatment of inflammatory bowel diseases (IBDs).1 AZA is a prodrug that is non-enzymatically converted into 6-MP, which is then metabolised into either 6-methyl-mercaptopurine or 6-thioguanine (6-TG). 6-TG is responsible for the cellular antiproliferative activity, by causing strand breakage through displacement of purine nucleotide incorporation into DNA,2 and its levels have been correlated with a therapeutic response in some but not all studies of patients with Crohn’s disease (CD).3 4 Recently, a RAC1-dependent induction of T cell apoptosis by thiopurines has been proposed as an additional mechanism for their anti-inflammatory activity in CD.5

Interestingly, prolonged administration for several months is required before the anti-inflammatory effect of thiopurines in IBD becomes apparent.6 This delayed onset of action has been traditionally attributed to a slow accumulation of the drug in target tissues.7 8 However, there is also conflicting pharmacokinetic evidence indicating that the drug level in erythrocytes reaches a steady-state level in up to 14 days.9^–11 Moreover, tissue concentrations were shown to rise rapidly, in parallel with the erythrocyte level, in murine models,12 13 and administering an intravenous loading dose did not hasten the onset of effect in patients with CD.14 From an immune perspective, it is hard to reconcile the purported capacity of thiopurines to induce apoptosis of T cells rapidly with their inability to control the inflammatory process until several months of treatment have elapsed. Thus, there is currently no explanation—neither pharmacokinetic nor immunological—for the slow onset of the effect of thiopurines. This knowledge gap impedes efforts to design rational therapeutic approaches that would minimise this major clinical disadvantage.

Therefore, the aim of the present work was to investigate immune mechanisms underlying the delayed onset of the anti-inflammatory effect of thiopurines.

METHODS

Human subjects

The study was approved by the relevant Institutional Ethics Committees, and donors gave written informed consent. Peripheral blood (PB) samples were obtained from healthy blood donors and from patients with CD. Lamina propria lymphocytes (LPLs) were obtained from intestinal biopsies of patients undergoing colonoscopy for CD or for unrelated reasons. LPLs for experiments performed at the Mount Sinai Medical Center were obtained from patients undergoing surgical resection for CD, cancer or diverticulitis.

Reagents

AZA, cycloheximide and prednisolone (Sigma-Aldrich, St Louis, Missouri, USA) were solublised in dimethylsulfoxide (DMSO). 6-MP and 6-TG (Sigma-Aldrich) were dissolved in NaOH. Thiopurines were then diluted in RPMI-1640 (Gibco, Invitrogen, Carlsbad, California, USA) to stock concentrations (1 mM), and cycloheximide and prednisolone were kept in a DMSO stock concentration of 2 or 3 mg/ml, respectively. Drugs were added to cultures at the designated final concentrations.

Monoclonal antibody (mAb) and fluorescence-activated cell sorting (FACS) analysis

Fluorescent-conjugated anti-CD3, anti-CD4, anti-CD8, anti-CD14, anti-interferon γ (IFNγ), antitumour necrosis factor α (TNFα) and anti-interleukin 2 (IL2), and the corresponding mouse immunoglobulin G (IgG) isotype controls were all from PharMingen (San Diego, California, USA). Stained cells were analysed on a Becton Dickinson FACScaliber (Franklin Lakes, New Jersey, USA).

Isolation of mononuclear cells and cell subsets

Peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation on a Ficoll–Hypaque gradient (Sigma). CD4+, CD8+ and CD14+ cell populations were isolated using immunomagnetic selection on separation columns as specified by the manufacturer (BD Biosciences Pharmingen, San Jose, California, USA). Intestinal LPLs were isolated as previously described.15

In vitro cell activation and culture

Isolated CD4+ or CD8+ T cells were resuspended in RPMI with 10% fetal calf serum (FCS), 2 mM l-glutamine and antibiotics. Cells were placed in 96-well flat-bottomed tissue culture plates at 3×10⁵ cells/well, after precoating with anti-CD3 (OKT3) at 2 μg/ml overnight. Upon plating, 1 μg/ml anti-CD28 (BD Pharmingen) was added to the suspension. IL2 (40 IU/ml) was added on day 3. LPLs were similarly cultured after triggering by 1 μg/ml preplated anti-CD2 (BD Pharmingen) with soluble anti-CD28. Isolated CD14+ cells were cultured after stimulation with lipopolysaccharide (LPS; Sigma-Aldrich) at 1 μg/ml.

Analysis of cell apoptosis and proliferation

Apoptotic cells were detected by flow cytometery after staining of cells with anti-Annexin V and propidium iodide (PI) using the Annexin V apoptosis detection kit-I (IQ-Products, Groningen, The Netherlands). Cellular proliferation was determined by carboxyfluorescein succinimidyl ester (CFSE) content, as previously described.16

Cytokine production assays

PB T cells were assessed for cytokine production by intracellular staining, as previously described.16 For assessing LPL cytokine production, supernatants were collected at the designated time points after stimulation, and ELISA assays were employed to determine TNFα (TNFα ELISA kit, Biosource, Invitrogen, Camarillo, California, USA) and IFNγ (eBioscience, San Diego, California, USA) content.

Adhesion assays

Adhesion of CD4+ T cells, activated by two rounds of stimulation with OKT3 and autologous irradiated PBMCs, was performed as previously described17 after preincubation of cells with the designated drugs for 48 h. All experiments were performed in triplicate.

In vivo (mouse) immunisation experiments

Balb/C female mice were obtained from Harlan (Rehovot, Israel). After acclimatisation, mice were co-immunised subcutaneously with 100 μg hen egg lysozyme (HEL) emulsified in incomplete Freund's adjuvant (IFA) (both from Sigma) and with 100 μg ovalbumin (OVA; Sigma) with IFA, twice over 4 weeks. Mice were then rested for an additional 4 weeks to allow for the development of resting memory cells. Thereafter, mice were divided into two groups, one treated with daily intraperitoneal 6-MP at 3 mg/kg, and one treated with vehicle alone. Both groups continued to receive biweekly intraperitoneal immunisations with HEL antigen, while OVA was no longer administered. Mice were sacrificed after 4 and 20 weeks of this protocol. The spleens were aseptically extracted, and splenocytes were labelled with CFSE as above, placed in 96-well flat-bottomed tissue culture plates at 5×10⁵ cells/well and stimulated by HEL (70 μM), OVA (20 μM), medium alone, or phytohaemagglutinin (10 μg/ml) as a positive control. IL2 (40 IU/ml) was added on day 3. Cells were harvested on day 7, and cellular proliferation of gated CD4+ cells was determined by FACS analysis of CFSE content.

Statistical analysis

p Values were calculated by Wilcoxon rank sum test or the Mann–Whitney test as appropriate, using the MedCalc software (Mariakerke, Belgium). Values <0.05 were considered significant.

RESULTS

6-MP and 6-TG abrogate the proliferation of activated lymphocytes

Previous studies addressing the antiproliferative effect of thiopurines on stimulated T cells have either employed DNA incorporation assays18 19 or used suprapharmacological concentrations.20 Therefore, we first gauged the cellular expansion of T cells under a thiopurine concentration (5 μM) shown to be clinically relevant.5 21 To this end, CFSE-labelled PB CD4+ T cells from patients with CD and healthy controls were triggered by plate-bound anti-CD3 and soluble anti-CD28 in the presence or absence of the indicated drugs. Cellular divisions were determined on day 5 by CFSE dilution. As depicted in fig 1A for a specific donor, the thiopurines significantly abrogated the expansion of CD4+ T cells. In multiple experiments, the proliferation of CD4+ cells was reduced by >80% (n = 21, 51% (3.2%), 8% (1.9%), 8.6% (2%), 12% (3.8%) mean proliferation (SEM), for OKT3 alone, 6-TG, 6-MP or AZA, respectively, p<0.01 for all comparisons, fig 1B). There was no difference between the magnitude of proliferation inhibition in healthy controls (n = 13) versus patients with CD (n = 8, p>0.1 for all comparisons).

A similar inhibition of CD3/CD28-induced proliferation was observed for isolated CD8 cells (fig 1C), and abrogation of CD4+ expansion was also found after stimulation with the superantigen TSST-1 (data not shown).

Thiopurines fail to induce apoptosis of resting T cells, and induce only modest apoptosis of activated T cells at day 5 following stimulation

The proliferation arrest induced by thiopurines prompted us to re-examine their proapoptotic effect on activated T cells. Thus, PB CD4+ cells were stimulated as above with or without the different thiopurines or with 10 μg/ml cycloheximide as a positive control. Apoptosis on day 5 was determined by Annexin V/PI staining. As shown in fig. 2A and B, treatment with thiopurines resulted in a modest but statistically significant increase in the percentage of early apoptotic (Annexin+/PI−) cells by day 5 (4.1% (2.9%) vs 7.1% (3.3%), 6.8% (3.2%) and 8.2% (4.2%) for the percentage of apoptotic cells after stimulation alone vs 6-TG, 6-MP or AZA, respectively, n = 18, p<0.05 for all comparisons). A similar effect was observed with respect to the percentage of the late apoptotic/necrotic (Annexin+/PI+) cells on day 5 (fig 2A,B). No difference was found in the percentage of apoptotic cells in patients with CD (n = 7) compared with healthy donors (n = 11, data not shown).

Although these results were in agreement with previous observations,5 it was of interest to elucidate the seeming contradiction between the ability of thiopurines to inhibit T cell proliferation and to induce apoptosis, and their inability to control the inflammatory process in IBD rapidly. Therefore, we next carefully examined the fate of these cells during the first days after stimulation. For this purpose, CD4+ cells were isolated and stimulated as above, and apoptosis was determined earlier, at day 3. Unlike the findings on day 5, the addition of thiopurines did not induce T cell apoptosis at day 3 poststimulation (fig 2C). In contrast, cycloheximide produced significant apoptosis at both time points (fig 2B,C). In addition, no increase in apoptosis was evident at an earlier time point (6 h poststimulation, data not shown). Similar to the findings in PB T cells, thethiopurines did not enhance the apoptosis of intestinal LPLs obtained from patients with CD and from patients with unrelated disorders on day 3 after anti-CD2/CD28 stimulation (fig 2D). Notably, apoptosis of PB CD+ cells on day 3 poststimulation was not induced even by increasing the concentrations of 6-MP up to 100-fold higher than those achieved in vivo, thereby ruling out any threshold effect (Supplementary fig. 1A). Furthermore, the modest proapoptotic effect of 6-MP and 6-TG, which was evident only on day 5 poststimulation, was restricted to activated cells, since apoptosis of unstimulated PB CD4+ cells was not enhanced by thiopurines even after 5 days of culture (Supplementary fig 1B).

Collectively, these results suggest that thiopurines do not induce apoptosis of resting T cells, and that activated T cells maintain their viability for up to 5 days after stimulation, despite treatment with thiopurines.

Activated T cells differentiate into potent effector cells despite thiopurine treatment

The results indicating that T cells remained viable for up to 5 days after activation prompted us to examine their functionality during this poststimulation period. Thus, PB CD4+ cells from healthy donors (n = 4) and patients with CD (n = 3) were stimulated by anti-CD3/CD28 as above, in the absence or presence of the designated drugs, and cultured for 3 days. Production of proinflammatory cytokines was then determined by intracellular staining after a brief activation by phorbol 12-myristate 13-acetate (PMA)–ionomycin in the presence of a Golgi inhibitor. Because corticosteroids are known to inhibit T cell effector functions, prednisolone at a clinically relevant concentration (1 μg/ml) was used as a positive control. As shown in a representative experiment, unlike the inhibitory effect of prednisolone, the addition of 6-TG or 6-MP did not significantly reduce the number of T-cells producing IFNγ, TNFα or IL2 (fig. 3A).

As a small percentage of T cells divide despite thiopurine treatment, we next examined whether cytokine production was restricted only to this small dividing population. CFSE-labelled CD4+ T cells were stimulated as above, and production of the designated cytokines was determined on day 5 by intracellular staining. We found that the cycle-arrested T cells in cultures with thiopurines were equally capable of cytokine secretion as were their actively dividing counterparts stimulated in the absence of the drugs (fig 3B). To increase the relevance of these findings to IBD, intestinal LPLs (n = 6) were stimulated with plate-bound anti-CD2/CD28, and the production of cytokines in the supernatant was assessed by ELISA 18 h after stimulation. These experiments showed that the thiopurines do not affect the secretion of IFNγ and TNFα by activated LPLs (fig 3C), whereas prednisolone significantly inhibits the production of these cytokines.

Given the results in T cells, and because cells of the monocytic lineage have been implicated in the pathogenesis of IBD,22 23 we have also studied the effect of thiopurines on LPS-stimulated PB CD14+ monocytes. Similar to the findings in T cells, none of the thiopurines enhanced the apoptosis of LPS-stimulated CD14+ cells on day 3 after activation, nor inhibited their cytokine secretion (Supplementary fig 2A, B). Taken together, these findings show that antipurine metabolites do not effectively inhibit cytokine secretion, by either dividing or non-dividing activated T cells in the PB or the intestinal mucosa, or by stimulated monocytes.

To extend these observations further to additional stimulation-induced lymphocyte effector functions, we also performed adhesion and cytotoxicity assays in the presence or the absence of 6-MP, 6-TG or prednisolone. These experiments showed that the thiopurines do not reduce cell-mediated killing of target K562 cells by PB T-NK (natural killer) effector cells (fig 4A). Additional experiments showed that adhesion of T cells to collagen IV and fibronectin is only marginally inhibited by these drugs (fig 4B). In contrast, prednisolone significantly abrogated both of these effector functions.

Prolonged administration of 6-MP, in vivo, causes the depletion of the specific CD4+ memory pool to repeatedly encountered antigens, but not to antigens to which it was previously exposed

The above findings suggest that repeated bouts of activation can still result in tissue damage mediated by cycle-arrested yet functional T cells. While this may account for the lack of an early effect of thiopurines, it also raises a question as to the mechanism responsible for their eventual late anti-inflammatory activity. Thus, it was of interest to study this seeming paradox in mice, in vivo, in the context of prolonged 6-MP treatment and repeated antigen encounter. For this purpose, we employed a novel experimental design in order to distinguish between the specific effects of 6-MP on memory T cell clones re-encountering antigen versus its effects on resting memory T cells. Thus, Balb/C mice were initially co-immunised with HEL and OVA, and then rested for 4 weeks to allow for the development of resting memory cells. Subsequently, the mice were repeatedly immunised biweekly with HEL alone (without OVA), under treatment with daily 6-MP or vehicle alone (fig 5A). Mice were sacrificed after either 4 or 20 weeks of this protocol. The magnitude of the splenic CD4+ memory response was determined ex vivo by CFSE-based measurement of the proliferation of CD4+ T splenocytes in response to either antigen. The results showed that short-term administration of 6-MP for 4 weeks had no effect on the memory response to either antigen (fig 5B). In contrast, the prolonged administration of 6-MP for 20 weeks markedly abrogated the memory T cell response to the re-encountered HEL antigen, compared with mice treated with vehicle alone, while re-exposed to this antigen (7.6% (6.1%) responding cells vs 23.8% (3.2%), p = 0.01, fig. 5C). Importantly, in these same mice, the prolonged administration of 6-MP did not affect the T cell memory response to the previously primed OVA antigen that was not reintroduced during 6-MP treatment (22% (3.2%) vs 29.9% (6.8%), p>0.1, fig 5C). Furthermore, the mean total number of splenocytes in the mice treated with 6-MP for 20 weeks was significantly lower compared with mice treated with vehicle alone (fig 5D).

Collectively, these results suggest that repeated antigen encounters over a prolonged course of time are a prerequisite for specific memory depletion induced by long-term thiopurine treatment.

DISCUSSION

This study addressed the mechanism for the slow onset of the anti-inflammatory effect of thiopurines. The results indicate that thiopurines have a potent antiproliferative effect at clinically relevant drug concentrations. Interestingly, despite this proliferation arrest, no significant apoptosis induction of stimulated T cells was evident until day 5 after stimulation, even at suprapharmacological 6-MP concentrations. Importantly, despite thiopurine treatment, these activated T cells could successfully acquire effector functions such as cytokine secretion, adhesion capacity and cytotoxicity. In contrast, prednisolone markedly inhibited this differentiation of intestinal and PB T cells into effector cells, which are instrumental in mediating intestinal inflammation and tissue damage.24 25 These findings support a concept whereby despite administration of thiopurines, T cells can still differentiate into potent effector cells after each bout of activation during the course of disease, and further propagate tissue damage. This may result in the inability of thiopurines to control inflammation more rapidly and induce a fast therapeutic response in patients, in contrast to agents such as corticosteroids.26

Following T cell priming, the expanded antigen-specific T cell population later contracts into a long-term memory population, whose size is determined by the magnitude of the initial T cell burst.27 Re-exposure to the cognate antigen results in the expansion of antigen-specific memory populations, recruitment of cells into tissues and differentiation into effector memory cells.28 29 Our mouse experiments showed that prolonged (20 weeks) treatment with thiopurines leads to contraction, rather than expansion, of the CD4+ memory pool for a repeatedly encountered antigen (HEL). In contrast, the mice maintained the memory response to a different antigen (OVA), to which they were not re-exposed during 6-MP treatment.

Antigen stimulation is a stochastic event, affecting at random a fraction of the specific memory cells, and only those present locally for antigen presentation. Thus, these data suggest a model whereby upon antigen exposure in vivo, the relevant fractions of antigen-specific memory cells can be activated and recruited to the site of inflammation, regardless of treatment with thiopurines. Successive episodes of antigen exposure could thereby cause recurrent T cell activation and resultant bouts of local inflammation. However, consecutive repeated encounters with the particular antigens eventually culminates in the selective depletion of the memory pool specific for these antigens, mediated by thiopurine-induced apoptosis and activation-induced apoptosis in the lack of compensatory clonal expansion. This, in turn, would predictably lead to the late downregulation of the inflammatory process driven by pathogenic antigens. In contrast, memory clones, that do not encounter their priming antigens, remain in their resting memory phenotype and retain their population size despite 6-MP treatment. Notably, rather than any single antigen, a multitude of antigens are probably involved in triggering the immune response in IBD. Some of these putative antigens, such as the intestinal microbiota, may alter over the course of time.30 When interpreted within the framework of our model, such antigenic shifts could be hypothesised to lead to escape from the effect of thiopurine and to intermittent relapse of disease, but other mechanisms may be responsible as well.

The in vitro findings for intestinal LPLs were similar to those found for PB T cells. Nevertheless, our murine experiments investigated the memory response in splenic CD4+ cells. Therefore, they cannot ascertain where the elimination of pathogenic T cell clones in IBD would take place. Lamina propria T cells are known to be proliferatively hyporesponsive to stimulation via the T cell receptor (TCR)/CD3 pathway compared with PB T cells.31 Whether these lamina propria T cells actively divide, in vivo, remains controversial.32 33 However, antigen presentation by professional antigen-presenting cells in secondary lymphoid organs is known to cause the expansion and differentiation of intestinal T cells, which subsequently migrate to the intestinal wall.34 Thus, it is possible that the incremental memory T cell depletion by thiopurine primarily affects these expanding populations of T cells in the secondary lymphoid organs (or the systemic circulation), prior to their recruitment into the inflamed intestinal wall.

In conclusion, the inability of thiopurines to terminate immune-driven inflammatory processes rapidly may be related to “lingering” activated immune cells with preserved effector function, that continue to exert tissue damage after stimulation. Subsequently, recurrent antigen encounters in the context of prolonged administration of the drug may lead to incremental depletion of antigen-specific T cell memory clones and eventual control of disease.