Anti-thymoglobulin (ATG) treatment does not reverse type 1 diabetes in the acute virally induced rat insulin promoter-lymphocytic choriomeningitis virus (RIP-LCMV) model

Abstract

Immune modulators such as anti-thymoglobulin (ATG) are under clinical evaluation for the treatment of type 1 diabetes (T1D). Although such agents have cured T1D in the non-obese diabetic (NOD) model, their clinical efficacy has been much lower. In order to improve the odds of successful translation from bench to bedside, we propose to evaluate this agent under more stringent conditions. Here, we evaluated the capacity of ATG to reverse T1D in the acute rat insulin promoter-lymphocytic choriomeningitis virus (RIP-LCMV) model. RIP-LCMV-glycoprotein (GP) mice were treated after new-onset T1D with murine ATG antibodies. Although ATG treatment did not impair viral clearance it failed to reverse new-onset T1D in this model. The CD4 : CD8 ratio was reduced drastically upon LCMV infection due to an expansion of CD8 effectors but ameliorated in ATG-treated mice. Although the percentage of CD4⁺CD25⁺ regulatory T cells (T_regs) within the CD4⁺ population was increased significantly after ATG therapy, their frequency in the periphery was reduced dramatically and never returned to normal baseline. The inability of ATG treatment to cure T1D in a stringent viral model (RIP-LCMV mice) is due at least partially to the inability to maintain or increase a sufficient CD4⁺CD25⁺ T_regs frequency, in striking contrast with what was reported in the NOD model. Our data would argue for the use of multiple animal models to assess efficacy of promising immune-based interventions and select the most potent therapies for future clinical trials.

Introduction

A growing number of immune-based therapies are proving efficient in reversing new-onset type 1 diabetes (T1D) in the non-obese diabetic (NOD) mouse [1]; however, when tested in humans, their efficacy is either reduced or nullified. Several factors influencing translation from bench to bedside have been identified. Among them, the accurate translation of the dosing and regimen from animal models to humans has been challenging so far, raising the question of whether animal models are sufficiently good indicators of therapeutic success in a patient [2]. We have proposed that in order to increase our odds of successfully translating immune-based therapies from bench to bedside, one should evaluate the efficacy of such immune interventions under more stringent conditions and in more than one animal model.

In this study, we decided to evaluate the anti-diabetogenic potential of murine anti-thymoglobulin (mATG) in a stringent model for T1D, the rat insulin promoter-lymphocytic choriomeningitis virus (RIP-LCMV) mice [3], where large numbers of autoaggressive CD8⁺ T cells are triggered by a viral infection. In this model, severe T1D develops rapidly within 10 to 12 days upon LCMV infection. Interestingly, histopathologically, this model bears many resemblances to human T1D, such as the predominance of CD8⁺ T cells, up-regulation of major histocompatibilty complex (MHC) class I in islets and overall smaller infiltration compared to the NOD. It does not exhibit multiple immune defects such as the NOD mouse, many of which are also not found commonly in human T1D, but it lacks the diabetes predisposing MHC.

Methods

Mice

Generation of C57BL/6 RIP-LCMV-glycoprotein (GP) transgenic mice has been described previously [3]. LCMV strain Armstrong (LCMV-arm) was used to trigger autoimmune diabetes in the RIP-LCMV-GP model. The RIP-LCMV constitutes a more acute model than the NOD mice, as diabetes usually occurs within 7–10 days post-LCMV infection in the fast-onset C57BL/6 RIP-LCMV-GP mice. However, RIP-LCMV mice recapitulate many features of human T1D, as described previously. Seven to 10-week-old RIP-LCMV-GP mice were infected with a single intraperitoneal (i.p.) dose of 10³ plaque-forming units (PFU) of LCMV-arm. This study was approved by the La Jolla Institute for Allergy and Immunology Animal Care Committee.

Blood glucose monitoring

Blood glucose values were obtained with the blood glucose monitoring system OneTouch Ultra (LifeScan Inc., Milpitas, CA, USA). Mice were considered diabetic when their blood glucose levels were superior to 300 mg/dl for two consecutive measurements [4].

mATG or control rabbit immunoglobulin (IgG) administration

mATG was prepared as described previously [5]. Newly diabetic RIP-LCMV-GP mice were treated on days 0 and 2 after new-onset type 1 diabetes with mATG or a control rabbit IgG (500 µg/day intravenously; 1 mg total).

Flow cytometry

Peripheral blood was collected by retro-orbital puncture from each individual mouse and the red cells were lysed using red blood cells lysing buffer (Biolegend, San Diego, CA, USA). Single cell suspensions were stained in fluorescence activated cell sorter (FACS) buffer [1% fetal calf serum (FCS) in 1× phosphate-buffered saline (PBS)] with CD4-fluorescein isothiocyanate (FITC), CD25-phycoerythrin (PE), CD8-peridinin chlorophyll (PerCP) and CD19-allophycocyanin (APC) antibodies. Each population was identified by flow cytometry using a four-colour FACSCalibur flow cytometer (BD Biosciences, San Diego, CA, USA). The number and percentages of the CD4⁺, CD8⁺, CD19⁺ and CD4⁺CD25⁺ populations were determined by analysing the data with FlowJo 7·2·2 software (TreeStar Inc., Ashland, OR, USA).

Quantitative real-time polymerase chain reaction (qPCR) to estimate LCMV copy numbers

The protocol was established by McCausland and Crotty [6]. Briefly, RNA was isolated from 50 µl serum or 20–30 mg tissue samples by using the RNAqueous kit (Ambion, Austin, TX, USA). All samples were frozen at −80°C until RNA extraction. Purified RNA was eluted in a volume of 20 µl and was frozen immediately at −80°C until use to avoid degradation. Ten µl of RNA was reverse-transcribed into cDNA (total volume of 20 µl) using the SuperScript III Reverse Transcriptase kit (Invitrogen, Carlsbad, CA, USA) and a specific primer [nucleoprotein (NP)1-R, AAGCTGAAGGCCAAGATCAT]. Five µl of cDNA was then used as a template for a 25-µl qPCR reaction on a GeneAmp 5700 (ABI, Redwood City, CA, USA), using the primers NP-R and NP-F (GAGGCTTTCTCATCCCAACTAT). A standard curve was generated by using a pCMV-NP plasmid (a gift from Professor Lindsay Whitton, The Scripps Research Institute, La Jolla, CA, USA).

Statistical analysis

Data are expressed as a mean ± standard deviation. Differences between groups were examined for statistical significance using the Mann–Whitney U-test. P-values < 0·05 were deemed significant.

Results

mATG failed to reverse diabetes in the RIP-LCMV model

To evaluate stringently the efficacy of mATG therapy, RIP-LCMV transgenic mice were used. Mouse ATG or rabbit immunoglobulin (rIgG) were administered intravenously after recent-onset T1D. In striking contrast with the NOD model [5], mATG therapy did not cure RIP-LCMV-GP mice (Fig. 1a and b). It is well established that T cells but not B lymphocytes are the main drivers of autoimmune diabetes in the RIP-LCMV model [7]. Therefore, to gain knowledge on how mATG treatment affects the T cell compartment in newly diabetic RIP-LCMV-GP mice, we evaluated the depletion and reconstitution kinetics of both CD4 and CD8 populations. Peripheral blood of RIP-LCMV-GP mice was collected at various time-points post-treatment with mATG or rIgG control. On day 4 after the last injection, a strong lymphodepletion was observed in the peripheral blood of mATG-treated mice (Fig. 1c), underscoring that mATG is functional in the C57BL/6 strain. As described in the NOD mouse model, the number of both CD4⁺ and CD8⁺ T cells decreased significantly after mATG treatment compared to the rIgG-treated control group. The number of CD4⁺ T cells per 10⁶ peripheral lymphocytes dropped from 994 ± 510 post-rIgG to 128 ± 101 post-mATG (*P < 0·05), while the CD8⁺ T cell count decreased from 1331 ± 808 to 101 ± 65, respectively (*P < 0·05). Conversely, no significant difference in the CD19⁺ B cell population was evidenced even though the average was lower in the mATG-treated group compared to control (365 versus 813 CD19⁺ cells per 10⁶ total lymphocytes, P = 0·068).

Fig. 1.

No efficacy of murine anti-thymoglobulin (mATG) polyclonal antibody to reverse new-onset type 1 diabetes (T1D) in rat insulin promoter-lymphocytic choriomeningitis virus glycoprotein (RIP-LCMV-GP) mice. RIP-LCMV-GP mice were infected with LCMV-arm to trigger diabetes. On days 0 and 2 after recent-onset, mice were treated with either mATG (n = 18) or recombinant immunoglobulin (rIgG) (n = 14), both at 500 µg/day; 1 mg total. The blood glucose values of each individual mouse were followed over time post-mATG (a) or rIgG treatment (b). (c) The number of CD4⁺, CD8⁺ and CD19⁺ lymphocytes was determined in the peripheral blood of C57BL/6 mice on day 5 after mATG (n = 10) or control rabbit IgG (n = 10 mice) treatment (both at 500 µg/day; 1 mg total); *P < 0·05. (d) Total CD4⁺ and CD8⁺ lymphocytes were enumerated in the peripheral blood of RIP-LCMV-GP mice infected with LCMV and injected after new onset with either saline solution (saline only; n = 5 mice), mATG (n = 9 mice) or control rabbit IgG (n = 5 mice). The blood was collected at days 7 and 14 post-treatment. The naive group represents the percentage of cells found in the peripheral blood of non-infected and non-diabetic RIP-LCMV-GP mice (n = 12). Data are given as a mean ± standard deviation.

In a kinetic study (Fig. 1d), we observed that the percentage of both CD4⁺ and CD8⁺ populations in the peripheral blood did not return to baseline day 7 after mATG depletion when compared to rIgG control (rIgG versus mATG; 18·80% ± 9·36% versus 1·75% ± 2·42% for CD4⁺ T cells and 20·72% ± 11·41% versus 1·72% ± 2·50% for CD8⁺ T cells; P < 0·0004). More surprisingly, the percentage of total T cells dropped below 10% in both groups on day 14 after treatment. This probably reflects some complications such as hyperglycaemia and weight loss associated with the diabetic status of the mice. Indeed, all mice had to be euthanized on day 14 after the end of the treatment due to these complications, and furthermore, diabetic RIP-LCMV-GP mice injected with only saline solution displayed a similar tendency between days 7 and 14 post-onset of T1D (Fig. 1d). Consequently, mice in the mATG group never reconstituted a sufficient T cell pool after treatment, which could affect their ability to control autoimmunity through induction of CD4⁺ regulatory T cells (T_regs).

Treatment with mATG partially restores the CD4 : CD8 ratio in LCMV-infected RIP-LCMV-GP mice

One of the features of mATG treatment in the NOD mice is the maintenance of a constant CD4 : CD8 ratio despite profound lymphodepletion [5]. We have analysed the variations of this ratio during the course of the disease and after mATG treatment in the RIP-LCMV-GP mice. Infection with LCMV led to a marked increase in the number of CD8⁺ cells, with an average of 142 750 ± 9614 CD8⁺ cells per 10⁶ peripheral cells in naive animals and 251 905 ± 45 109 in infected mice at diabetes onset (**P < 0·0001; Fig. 1d and data not shown). Consequently, the CD4 : CD8 ratio was found to be significantly lower throughout the course of the study in LCMV-infected (from 0·38 ± 0·16 to 0·55 ± 0·14 between days 4 and 7 post-infection) compared to naive (2·37 ± 0·268) RIP-LCMV-GP mice (Fig. 2). We observed that while the CD4 : CD8 ratio was diminished after mATG treatment it was maintained constantly above 1 (from 1·33 ± 0·31 to 1·5 ± 0·137 at days 4 and 14, respectively). Conversely, control rIgG resulted in a CD4 : CD8 ratio constantly lower than 1 (from 0·768 ± 0·2 to 0·831 ± 0·357 at days 4 and 14, respectively). Therefore, although RIP-LCMV-GP-treated mice presented a low CD4 and CD8 T cell count in the peripheral blood at days 7 and 14, the CD4 : CD8 ratio was ameliorated in the mATG when compared to control groups (1·5 ± 0·137 compared to 0·831 ± 0·357 and 0·55 ± 0·14 at day 14, *P < 0·05).

Fig. 2.

The CD4 : CD8 T cells ratio is improved but not restored after murine anti-thymoglobulin (mATG) treatment in the rat insulin promoter-lymphocytic choriomeningitis virus glycoprotein (RIP-LCMV-GP) mice. Newly diabetic RIP-LCMV-GP mice were treated with either mATG (n = 18 mice) or rabbit immunoglobulin (rIgG) (n = 14 mice) control antibodies, both at 500 µg/day; 1 mg total. On days 4, 7 and 14 after treatment, the peripheral blood was collected to measure the CD4 : CD8 ratio by flow cytometry. A group of naive mice (n = 6) and a group of infected mice (n = 6) were used to evaluate the CD4 : CD8 ratio into untreated animals; *P < 0·05.

mATG therapy after new-onset T1D does not impact viral clearance

Because the lymphocyte counts in the peripheral blood never recovered after mATG therapy, we hypothesized that viral clearance could have been affected. Therefore, we evaluated the number of LCMV copies by qPCR (Fig. 3a). The number of LCMV genome copies was found at a low level in the serum at 20 days post-infection with or without antibody treatment (Fig. 3b). A dramatic increase of LCMV genome copies was observed in the spleens, but not in the kidney, at 35 days post-treatment (Fig. 3c). None the less, similar LCMV genome copies were detected after control rIgG treatment or in LCMV-infected C57BL/6 mice. Thus, we conclude that mATG does not induce any virus reactivation, as the quantities of LCMV genome copies were found to be similar after mATG or rIgG administration.

Fig. 3.

Murine anti-thymoglobulin (mATG) treatment does not affect lymphocytic choriomeningitis virus (LCMV) clearance. Quantitative LCMV real-time polymerase chain reaction (PCR) was used to measure the number of LCMV RNA copies in the serum and various organs of infected mice. (a) Quantitation of the nucleoprotein (NP) standard curve (Ct = cycle threshold). The number of RNA copies of the NP of the LCMV was calculated on day 20 in the serum (b) or day 30 in the spleens and kidneys (c) of mATG or control rabbit immunoglobulin (IgG) (rIgG)-treated rat insulin promoter-glycoprotein (RIP-GP) mice post-infection. As control, LCMV-infected C57BL/6 mice were used and analysed at the same time-points.

ATG treatment augments the percentage but not the frequency of CD4⁺CD25⁺ T_regs in the peripheral blood

As shown Fig. 4a, the percentage of CD4⁺CD25⁺ T_regs within the CD4 population increased significantly post-mATG therapy compared with mice injected with rIgG (at 7 days 23·3 ± 9·0 versus 8·7 ± 1·5, P < 0·05 and at 14 days 20·0 ± 8·1 versus 8·8 ± 2·3, P < 0·05). Previous studies have shown that CD4⁺CD25⁺ T_reg frequency in the peripheral blood was increased on days 7 and 14 post-mATG treatment in the NOD mice [5]. In the RIP-LCMV-GP model, the CD4⁺CD25⁺ T_reg population behaved differently after mATG therapy. Indeed, the frequency of peripheral CD4⁺CD25⁺ T_regs decreased gradually over time (Fig. 4b). This loss of CD4⁺CD25⁺ T_regs was not associated with mATG, as a similar trend is observed after control rIgG treatment. Therefore, in contrast to what was described in the NOD model, mATG was unable to expand T_regs in the RIP-LCMV model, providing at least one explanation for the lack of efficacy.

Fig. 4.

Variations in the CD4⁺CD25⁺ regulatory T cells (T_regs) after murine anti-thymoglobulin (mATG) treatment. Newly diabetic rat insulin promoter-lymphocytic choriomeningitis virus glycoprotein (RIP-LCMV-GP) mice were treated with either mATG or control rabbit IgG (on days 0 and 2 at 500 µg/day; 1 mg total). Days 7 and 14 post-treatment, the percentage of CD25⁺ T_regs within the CD4 population was evaluated by flow cytometry (a) and the frequency of CD4⁺CD25⁺ T_regs was also measured in the peripheral blood (b). Each dot represents a single animal and the average ± standard deviation is shown for each group.

Discussion

A definitive cure for T1D would have to establish immune tolerance by harnessing T_regs to control autoreactive T cells [8]. Several strategies have been tested in mice as well as in humans to block effector T cells (T_eff) and expand islet-specific T_regs. Among them, administration of immune modulators such as anti-CD3 antibodies or anti-lymphocyte serum (ALS) showed efficacy in reversing new-onset T1D in the NOD mouse model [5,9–11].

Early studies showed that administration of ALS to newly diabetic NOD mice restored tolerance and prolonged survival of islet allografts [10], later suggested to act through the induction of CD4⁺CD25⁺ T_regs[11]. Mouse ATG is a clinical equivalent of ALS and is known to deplete lymphocytes in vivo and can be used effectively in a variety of therapeutic settings. Similarly to ALS therapy, administration of mATG after new-onset T1D in NOD mice stopped disease progression and induced long-term tolerance [5]. Short-course treatment of NOD mice with mATG induced a strong T cell depletion, with a lymphocyte count returning to normal after 30 days. Protection from T1D after mATG administration was associated with (1) an enhancement of CD4⁺CD25⁺ T_regs suppressive function and (2) an alteration of dendritic cell (DC) profile and function, skewing the T helper type 1 (Th1)/Th2 balance in vivo. Another important observation is the time-dependent induction of tolerance in NOD mice [5]. ATG treatment was efficacious only when administered late in the prediabetic phase (at 12 weeks of age) or after recent-onset T1D. One can argue that such a time-dependent activity might also be seen in the RIP-LCMV model for T1D and that efficacy might benefit from an earlier intervention before diabetes onset. However, in this model the therapeutic window is rather small when compared to NOD mice, as the prediabetic phase lasts for only 9–12 days post-LCMV infection. In this model, treating just before viral infection or days 0–7 post-infection with the same mATG dosage (500 µg/injection; 1 mg total) might be detrimental for optimal viral clearance, as lymphodepletion will occur before complete clearance of the virus. Therefore, only lower doses resulting in incomplete depletion could improve the safety/efficacy ratio when administered earlier in the disease course. The current mATG dose (500 µg/day on days 0 and 3; 1 mg total) was used based on the initial study by Simon et al. [5], demonstrating that a 1·0 mg/animal dose (approximately 50 mg/kg of body weight for a 20 g mouse) was optimal because it represented the minimal dose of mATG, providing an equivalent degree of peripheral blood lymphocyte depletion targeted in human therapeutic settings utilizing anti-thymoglobulin. In human clinical trials, the total ATG dosage administered for the treatment of new-onset T1D (i.e. [12] and the START clinical trial (Stimulating Targeted Antigenic Responses to Non-Small Cell Lung Cancer) at http://www.type1diabetestrial.org) or after islet transplantation [13,14] usually varies from 6 to 27 mg/kg of body weight. Therefore, based on this knowledge, it will be valid to test lower mATG doses in animal models to maintain efficacy while reducing adverse events. In RIP-LCMV mice a reduction in the mATG dosage might block effector T cells without affecting viral clearance.

In the RIP-LCMV-GP mice the LCMV GP33-specific CD8⁺ T cells but not the CD4⁺ T cells are essential for T1D development [3]. This constitutes another major difference with the NOD model, where both CD4⁺ and CD8⁺ T cells are thought to participate in T1D pathogenesis and islet destruction [15]. Furthermore, the CD4 : CD8 ratio at T1D onset changes considerably between NOD and RIP-LCMV-GP mice (i.e. ∼2 and ∼0·5, respectively). Future studies will need to examine whether such variations from animal model to animal model have important therapeutic implications, in particular when using low-dose lymphodepleting antibodies such as ATG that might not target CD4⁺ and CD8⁺ populations equally.

Here, we demonstrated that mATG therapy did not significantly affect the disease incidence in the RIP-LCMV-GP model for T1D. This failure in controlling autoimmunity was not due to an inability to clear the LCM virus but rather the incapacity of mATG to expand CD4⁺CD25⁺ T_regs which might, in return, affect the suppression of effector T cells in vivo. These data are in striking contrast with the results obtained with FcR-non-binding anti-CD3 antibody which promotes T_regs and reverses new-onset T1D in both NOD and RIP-LCMV models [4,9]. We would argue that evaluating the efficacy of immune interventions under stringent conditions and in more than one animal model could help in selecting the most potent immune modulators for future clinical trials, especially as human trial populations can be expected to be genetically heterogeneous and with a wide range of T1D severity (kinetic of the disease, frequency of effector T cells, etc.).

Disclosure

All authors have declared that they have no conflicts of interest.