Protection from articular damage by passive or active anti-tumour necrosis factor (TNF)-α immunotherapy in human TNF-α transgenic mice depends on anti-TNF-α antibody levels


Correspondence: L. Semerano, Department of Rheumatology, Avicenne Hospital (AP-HP) 125 Route de Stalingrad 93000, Bobigny, France.



Active anti-tumour necrosis factor (TNF)-α immunization with the kinoid of TNF-α (TNF-K) induces polyclonal anti-TNF-α antibodies and ameliorates arthritis in human TNF-α (hTNF-α) transgenic mice (TTg). We compared the efficacy of TNF-K to that of infliximab (IFX) and of TNF-K and IFX co-administration, and evaluated whether the titres of anti-hTNF-α antibodies induced by immunization were a determinant of TNF-K efficacy. Forty-eight TTg mice received one of the following treatments: TNF-K immunization (TNF-K group); weekly IFX throughout the study duration (IFXw0–15); TNF-K plus weekly IFX for 4 weeks (TNF-K + IFX); and weekly IFX for 4 weeks (IFXw0–4); PBS. Animals were killed at week 16. Anti-hTNF-α antibody titres and clinical and histological scores were compared. All TNF-K immunized mice (TNF-K and TNF-K + IFX) produced anti-hTNF-α antibodies. Titres were higher in TNF-K versus TNF-K + IFX (P < 0·001) and correlated inversely with histological inflammation (R = −0·78; P = 0·0001) and destruction (R = −0·67; P = 0·001). TNF-K + IFX had higher histological inflammation and destruction versus TNF-K (P < 0·05). A receiver operating characteristic (ROC) analysis of anti-hTNF-α antibody titres identified the criterion cut-off value to discriminate most effectively between the TNF-K and TNF-K + IFX groups. Mice with high versus low titres had less histological inflammation and destruction (P < 0·05). In a model of TNF-α-dependent arthritis, protection from articular damage by TNF-K correlates with the titres of induced anti-hTNF-α antibodies. The co-administration of TNF-K and a short course of infliximab does not result in less articular damage versus solely TNF-K, due probably to lower anti-hTNF-α antibody production. These results are relevant for future development of active anti-TNF-α immunization in human disease.


The advent of tumour necrosis factor (TNF)-α-targeting drugs (anti-TNF-α) have changed the perspectives of rheumatoid arthritis (RA) treatment dramatically over the last decade, giving unprecedented results in terms of disease control and structural damage prevention [1]. Nevertheless, only 25–50% of anti-TNF-α-treated patients achieved remission in controlled clinical trials [2, 3], and even lower remission rates are described in everyday practice [4]. An approximately similar proportion reaches a functional status comparable to that of the general population [5, 6]. Primary or secondary therapeutic failures on anti-TNF-α drugs are not infrequent [7], and there is increasing evidence that the induction of anti-drug antibodies could be a major contributory factor to insufficient response to this class of therapeutics, at least in the case of anti-TNF-α monoclonal antibodies [8-11]. These drawbacks of current anti-TNF-α treatments confirm that there is room for alternative ways to target this key proinflammatory cytokine. Among these, active immunization against TNF-α with TNF-α kinoid (TNF-K) is promising [12, 13]. The chemically inactivated human TNF-α (hTNF-α) is coupled to a carrier protein (the keyhole limpet haemocyanine, KLH). This compound is capable of breaking B cell tolerance to hTNF-α, thereby inducing the production of polyclonal, neutralizing anti-hTNF-α antibodies and avoiding the risk of anti-drug antibody induction [12]. Importantly, TNF-K does not sensitize T cells to native hTNF-α. In the absence of specific T cell help, the rupture of B cells tolerance is transitory, and within 12–20 weeks there is a greater than 50% decline in anti-hTNF-α antibody titres [14-16]. Our group developed the proof of concept of TNF-K applicability in RA using the hTNF-α model transgenic mouse (TTg) [17, 18]. TTg mice develop an hTNF-α-dependent spontaneous arthritis and are therefore the pertinent model to study a TNF-α-targeting strategy. We were able to demonstrate the dramatic efficacy of TNF-K in TTg arthritis, with immunized mice showing mild clinical arthritis scores and prevention of histological joint inflammation and destruction compared to control mice [14-16].

Based on the proof of concept established in this model and other preclinical and clinical studies, TNF-K entered clinical development for RA and a Phase IIa clinical trial is now terminated (Clinical identifier: NCT01040715). In all the experiments TNF-K showed a slower onset of clinical effect compared with a monoclonal anti-TNF-α antibody [infliximab (IFX)]. This latency results from the time necessary for antibody production by sensitized B cells. Conversely, two or three TNF-K immunizations over 4 weeks resulted in a longer-lasting clinical effect versus IFX given weekly for the same period [15, 16].

The aims of the present study in TTg mice were: (i) to compare the efficacy of TNF-K to that of long-term IFX treatment and of the co-administration of TNF-K and IFX; and (ii) to determine whether the levels of anti-hTNF-α antibodies induced by TNF-K are correlated with articular damage and may therefore represent a prognostic factor for immunized mice.

Materials and methods


Forty-eight male hTNF-α hemizygous TTg, 4–8 weeks old, were purchased from Taconic Farms (Germantown, NY, USA). They were divided into four groups of 10 mice and one group of eight mice, and identified according to the study protocol described below. These mice develop a spontaneous arthritis between 8 and 10 weeks of age [17, 18]. They were weighed and monitored for evidence of arthritis in the four paws throughout the duration of the experiment, and killed at week 16 after arthritis onset. All procedures were approved by the Animal Care and Use Committee of the University of Paris 13 (ethical approval ID: Ce5/2010/036).


We obtained hTNF-α kinoid (TNF-K), a protein complex of hTNF-α and KLH, from Neovacs SA (Paris, France), as described previously [14, 15]. Dulbecco's phosphate-buffered saline (PBS) was purchased from Eurobio (Les Ulis, France), ISA-51 adjuvant from Seppic (Paris, France) and IFX (Remicade®) from Schering-Plough (Levallois-Perret, France).

Study protocol

The study protocol is presented in Fig. 1. Week 0 is defined as the week when treatments were started. The treatment groups were: (i) immunization with TNF-K (TNF-K group), 10 mice; (ii) PBS as negative control (PBS group), 10 mice; (iii) weekly IFX throughout the experiment duration, from weeks 0 to 15 (IFXw0–15 group), 10 mice; (iv) immunization with TNF-K plus weekly IFX from weeks 0 to 4 (TNF-K + IFX group), 10 mice; and (v) weekly IFX from weeks 0 to 4 (IFXw0–4 group), eight mice.

Figure 1.

Study protocol and arthritis scores. All treatments were started at week 0 (black arrow). The tumour necrosis factor-α-kinoid (TNF-K) group (orange diamonds) received three immunizations with TNF-K at weeks 0, 1 and 4 of the experiment. The phosphate-buffered saline (PBS) control group (pink squares) received three injections of PBS at weeks 0, 1 and 4. The infliximab (IFX)w0–15 group (purple triangles) received weekly injections of IFX from weeks 0 to 15. The association group TNF-K + IFX (green circles) received three immunizations with TNF-K at weeks 0, 1 and 4 and weekly injections of IFX from weeks 0 to 4. The IFXw0–4 group (blue squares) received weekly injections of IFX from weeks 0 to 4. The clinical score curves report the mean ± standard error of the mean of the clinical scores for each group at each observation.

TNF-K administration

Animals were injected intramuscularly with 10 μg TNF-K in a 1:1 emulsion with ISA-51 (100 μl) at weeks 0, 1 and 4 of the study.

PBS administration

Animals were injected intramuscularly with 100 μl of PBS in a 1:1 emulsion with ISA-51 (100 μl) at weeks 0, 1 and 4 of the study.

IFX administration

Animals were injected with IFX intraperitoneally, 10 mg/kg weekly, from weeks 0 to 15 or from weeks 0 to 4, according to the treatment schedule.

Blood samples

Blood samples were collected before the first treatment injection (week 0), then at weeks 5, 14 and at killing (week 16) for anti-hTNF-α and anti-KLH antibody titrations, IFX titrations and hTNF-α neutralizing capacity. Blood collection for each treatment group was performed just before treatment administration.

Clinical arthritis assessment

Weekly clinical assessment of arthritis was started from the receipt of the animals. Evaluations were carried out by an assessor unaware of assignment to treatment groups. Clinical severity of arthritis for each paw (fingers, tarsus and ankle) was quantified on a score ranging from 0 (normal) to 3 (severe inflammation with deformation) [16]. The score of each paw was summed, resulting in an overall arthritis score ranging from 0 to 12. The presented clinical score curves (Fig. 1) are based on mean arthritis score ± standard error of the mean (s.e.m.) at each clinical observation for each treatment group.

Histological arthritis assessment

All animals were killed at week 16. Right forelimbs and left hind limbs were collected, fixed, decalcified, dehydrated and set in paraffin blocks. Slides of 5 μm thickness were obtained. At least four serial sections were realized for each paw in order to obtain a reliable spatial evaluation of articular samples. Slides were stained with either haematoxylin and eosin or safranin-O before microscopic observation (optical microscope). Synovitis (articular inflammation) and bone erosions (articular destruction) were defined on haematoxylin and eosin-stained slides. Lesions were evaluated quantitatively on each slide using a three-point scale ranging from 0 to 3, where 0 = normal articulation; 1 = slight inflammation and thickening of the synovium; 2 = mild thickening of the synovium and mild inflammation with invasion of the subsynovial area by inflammatory cells; and 3 = severe inflammation and massive invasion of adjacent tissues by pannus [19]. Other sections were scored for loss of safranin-O staining as a measure of cartilage proteoglycan depletion, using a scale from 0 to 3 where 0 = no depletion; 1 = depletion of staining and thinning down of the lateral superficial layer; 2 = depletion of staining and thinning down of the central superficial layer; and 3 = severe and mostly complete depletion of staining in the superficial layer [19].

Antibody assays

Sera were obtained and tested for individual anti-TNF-α antibody titres and IFX trough levels from blood samples collected at different time-points during the experiment (weeks 0, 5 and 14) and at killing (week 16). Pooled sera from each group were tested for anti-KLH antibody titres and hTNF-α neutralizing capacity. Specific mouse anti-hTNF-α and anti-KLH antibody titres were determined using a direct enzyme-linked immunosorbent assay (ELISA). Precoated ELISA plates with 100 ng per well of hTNF-α or KLH were incubated with serial dilutions of sera collected from all groups. Specific immunoglobulin (Ig)G was detected by using horseradish peroxidase-conjugated (HRP) rabbit anti-mouse IgG (Zymed Laboratories Inc., now part of Invitrogen Corporation, Carlsbad, CA, USA). To each well, substrate solution was added (Fast OPD; Sigma, St Louis, MO, USA), then the reaction was stopped with H2SO4. The optical density (OD) was read at 490 nm. IFX titration was similar to mouse anti-hTNF-α titration except that we used HRP-conjugated mouse anti-human IgG1 (Zymed) instead of HRP-conjugated rabbit anti-mouse IgG. It is worth noting that HRP rabbit anti-mouse IgG does not cross-react with IFX.

The neutralizing capacity of anti-hTNF-α antibodies was assessed by L929 cytotoxicity assay, as described elsewhere [14]. Briefly, mouse fibroblast L929 cell line (CCL1) (American Type Culture Collection, Manassas, VA, USA) was cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. The cells were seeded in flat-bottomed 96-well plates at 2·104 cells per well. After 18 h of incubation at 37°C, serial dilution of serum with 2·5 ng/ml hTNF-α dose was added on L929 cells with 1 μg/ml of actinomycin D. After 18 h of incubation at 37°C, [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulphophenyl)-2H-tetrazolium, inner salt]/phenazine methosulphate (MTS/PMS) was added for 4 h at 37°C. The OD at 490 nm was read for each well. The neutralization titre was expressed as the mean of the reciprocal of the serum dilution that neutralizes 50% of hTNF-α activity (NC50).

Antigenicity test

Binding of IFX to TNF-K was measured by direct antigenicity assay. Briefly, eight serial dilutions of TNF-K or hTNF as control were coated onto plates in PBS buffer at a starting dilution of 1000 ng/well. After plate saturation, targeted hTNF-α epitopes were detected using biotinylated IFX and revealed by streptavidin–HRP (Southern Biotech, Birmingham, AL, USA). After 6 min of o-phenylenediamine (OPD) incubation, absorbance was read at 490 nm.

Statistical analysis

Data distribution was preliminary checked by the Kolmogorov–Smirnov test. According to data distribution, values are expressed as mean and standard deviation or median and interquartile range. Serial measurements of clinical scores were analysed considering the area under the curve (AUC) for each mouse as a summary measure; these measures were then analysed as raw data [20]. According to data distribution and number of groups, a parametric [analysis of variance (anova), t-test] or non-parametric (Kruskal–Wallis, Mann–Whitney) test was then performed. Post-hoc comparisons were performed with the appropriate test according to data distribution (Student–Newman–Keuls for parametric data, corrected Mann–Whitney test for non-parametric data). For individual anti-hTNF-α titres and IFX levels, we calculated both the AUC and the geometric mean. Both measures were highly correlated (R = 0·91); however, given the reduced number of blood samples, for subsequent analysis we chose to rely only on the geometric mean, which was therefore used as summary measure, in a t-test for comparison of anti-hTNF-α titres for the TNF-K versus TNF-K + IFX groups and for IFX levels for the IFXw0–15 versus IFXw0–4 and for TNF-K + IFX versus IFXw0–4 groups, respectively. Pooled anti-KLH titres and anti-TNF-α neutralizing activity values were analysed by Kruskal–Wallis test with appropriate post-hoc analysis. Histological scores are given as the arithmetic mean of all articular site scores for each mouse, and were compared with the Kruskal–Wallis test, with post-hoc comparison with the corrected Mann–Whitney test. The criterion value to discriminate between TNF-K-immunized groups on the basis of the geometric mean of anti-hTNF-α antibody titres was calculated with receiver operating characteristic (ROC) curve analysis. Pearson's correlation was used to correlate histological scores and the geometric mean of anti-hTNF-α antibody titres for each TNF-K-treated mouse, and to correlate the histological scores and the geometric mean of serum levels of IFX for each IFX-treated mouse. All statistical analyses were performed with MedCalc software version 10·4 (MedCalc Software bvba, Mariakerke, Belgium).


Effect of TNF-K, IFX and their co-administration on arthritis

When the mice exhibited an average clinical score of 3 (scoring range from 0 to 12; see Materials and methods), treatment was started for all mice (week 0). We waited until each mouse exhibited sign of arthritis in at least one paw and the mean clinical score for each cage was at least 3 (scoring range from 0 to 12; see Materials and methods), then we started treatment for all mice (week 0). As expected, the PBS control group developed severe arthritis rapidly over a 15-week period. Conversely, all treatment groups showed evident amelioration of arthritis versus the PBS group (Fig. 1). We observed a rapid clinical effect mainly in the IFX-treated groups, while in a subsequent phase we also observed a clinical amelioration in the TNF-K group, and an aggravation in the group that received IFX only during the first 4 weeks (IFXw0–4).

We therefore conducted the clinical score curve analysis along three time-periods: throughout the whole duration of the experiment, for the initial part of the experiment (from weeks 0 to 7) and for the final part of the experiment (from week 10 to killing). For the overall experiment duration, the analysis was aimed to determine whether one treatment produced lower clinical scores throughout the study duration. There was overall significant variability (P < 0·001), with all active treatments showing lower scores compared to PBS groups (P < 0·05 for all post-hoc comparisons). Despite the qualitative impression of a difference between active treatment groups, with IFXw0–4 appearing less effective (Fig. 1), no significant difference could be detected in terms of AUC for a particular treatment, due probably to lack of statistical power. For the initial part of the experiment, the analysis of the curves from weeks 0 to 7 confirmed the rapid clinical efficacy of IFX versus TNF-K. The clinical scores of the PBS and TNF-K group were higher compared to those of all IFX-treated groups (IFXw0–4, IFXw0–15, TNF-K + IFX) (P < 0·05). This confirmed that TNF-K treatment has a slower onset of action compared to IFX, and that the co-administration of IFX with TNF-K can overcome this latency. For the final part of the experiment, at week 10 of the experiment all IFX- and TNF-K-treated groups had similar clinical scores (P = 0·8). At killing, the group that received IFX during the first 4 weeks (IFXw0–4) displayed significantly higher clinical scores versus the IFXw0–15, TNF-K + IFX and the TNF-K groups (P < 0·05), while no difference was detectable between IFXw0–15 and both the TNF-K-treated groups. This confirms the longer-lasting clinical effect of TNF-K immunization given three times from weeks 0 to 4 compared to IFX given during the first 4 weeks. The best clinical control of arthritis could be assured only by long-term treatment with IFX (group IFXw0–15) or by TNF-K immunization (TNF-K and TNF-K + IFX groups).

Effects of TNF-K, IFX and their co-administration on histological inflammation and destruction

Histological scores of the different treatment groups at week 16 are presented in Fig. 2, ordered from the lowest to the highest. All treatment groups had significantly lower scores of histological inflammation versus PBS (P < 0·05). The TNF-K group was the group with the lowest scores and the only one that displayed significantly lower scores versus the TNF-K + IFX and IFXw0–4 groups (P < 0·05 for both comparisons) (Fig. 2a). Similarly, all treatment groups had significantly lower scores of histological destruction versus PBS (P < 0·05), with TNF-K displaying significantly lower scores versus both the TNF-K + IFX and IFXw0–4 groups (P < 0·05) (Fig. 2b). Histological destruction scores with safranin-O staining showed a similar distribution (not shown). Representative examples of histological sections are shown in Fig. 2c–j.

Figure 2.

(a,b). Box-and-whisker plot with all data (arithmetic mean of all articular site scores for each mouse) plotted for each treatment group of histological inflammation (a) and destruction (b) scores [haematoxylin-eosin (HE) staining], compared for each group. All groups had lower scores than the phosphate-buffered saline (PBS) control group. The tumour necrosis factor-α-kinoid (TNF-K) group had significantly lower scores versus the PBS, infliximab (IFX)w0–4 and TNF-K + IFX groups for both inflammation and destruction scores. *P < 0·05 versus PBS; **P < 0·05 versus PBS and IFXw0–4; ***P < 0·05 versus PBS, IFXw0–4 and TNF-K + IFX. Representative examples (c–j) of histological section for each treatment group stained, respectively, with safranin O (c–f) for cartilage–proteoglycan depletion and with haematoxylin and eosin (g–j) for articular inflammation and destruction. The TNF-K and the IFXw0–15 groups show good protection from cartilage and proteoglycan depletion (c,d) and from histological inflammation (g,h). The PBS group displays marked cartilage depletion (f) and synovial inflammation and destruction (j). In some mice in the TNF-K + IFX group, moderate cartilage depletion (e) and synovial inflammation and destruction (i) were detected.

Anti-hTNF-α antibody titres induced by TNF-K immunization

Because the co-administration of IFX and TNF-K (TNF-K + IFX) did not result in lower clinical and histological scores compared to either treatment alone, we investigated whether this might be due to lower efficiency of TNF-K immunization at inducing anti-hTNF-α antibodies in this group. Despite the fact that anti-hTNF-α antibodies were detected in all mice, mice in the TNF-K + IFX group had lower anti-hTNF-α antibody titres versus those in the TNF-K group (P < 0·05) (Fig. 3a,b). Conversely, the two groups did not differ in terms of anti-KLH antibody titres (P = 0·7) (analysis conducted on pooled sera for each group, data not shown). To confirm a link between anti-hTNF-α antibodies and histological damage, we subsequently investigated whether the anti-hTNF-α antibody titres were correlated with histological scores. For this purpose, the geometric mean of anti-hTNF-α antibody titres at three different blood samplings was calculated for each mouse. We then tested the correlation between the geometric mean of anti-hTNF-α antibody titres and the histological scores for each mouse, finding an inverse correlation with all histological scores (Fig. 4). This correlation was good and significant for both inflammation [correlation coefficient (R) = −0·78; P = 0·0001] (Fig. 4a) and destruction scores (R = −0·67; P = 0·001) with haematoxylin and eosin (Fig. 4b) and moderate and significant for safranin O destruction scores (R = −0·51; P = 0·01, not shown).

Figure 3.

Anti-human tumour necrosis factor (hTNF-α) antibody titres in tumour necrosis factor-α-kinoid (TNF-K) (a) and TNF-K + infliximab (IFX) (b) groups as detected by enzyme-linked immunosorbent assay (ELISA). The graph shows antibody titres in the three different blood samples for each mouse. The TNF-K group had higher titres than TNF-K + IFX (P < 0·01). A receiver operating characteristic (ROC) curve analysis on the geometric mean of anti-hTNF-α antibody titres for each mouse identified the value of 4211 as the cut-off value to discriminate between the two groups (dotted line). Nine mice in the TNF-K group and four mice in the TNF-K + IFX groups had higher anti-hTNF-α titres than the cut-off.

Figure 4.

(a,b) Tumour necrosis factor-α-kinoid (TNF-K)-treated mice [TNF-K and TNF-K + infliximab (IFX) groups]: correlation (with scatter diagram and regression line) between the geometric mean of TNF-K-induced anti-hTNF-α antibody titres and histological inflammation scores (a), and histological destruction scores (b) for each individual mouse. (c,d) IFX-only treated mice (IFXw0–15 and IFXw0–4 groups): correlation between the geometric mean of IFX trough serum levels and histological inflammation scores (c) and histological destruction scores (d) for each individual mouse.

Because anti-hTNF-α antibody titres were correlated inversely with histological scores of inflammation and destruction, and as the TNF-K + IFX group had lower anti-hTNF-α antibody titres versus TNF-K, we aimed to determine whether this factor might have accounted for the higher histological scores of the TNF-K + IFX group. We performed a ROC curve analysis on the geometric mean of anti-hTNF-α antibody titres in order to identify the criterion cut-off value to discriminate most clearly between the TNF-K and TNF-K + IFX groups in terms of antibody production. Nine mice in the TNF-K group and four of the mice in the TNF-K + IFX group had higher antibody titres than the cut-off (4211 dil−1) (Fig. 3). We then compared the histological scores for mice having higher or lower antibody titres than the cut-off, and we showed that mice with higher antibody titres had significantly lower inflammation and destruction scores (with both haematoxylin and eosin and safranin O staining) (P < 0·05 for all differences; Fig. 5).

Figure 5.

Box-and-whisker plot with all data (arithmetic mean of all articular site scores for each mouse) plotted of histological inflammation (a) and destruction (b) scores [haematoxylin-eosin (HE) staining] categorized on the basis of anti-human tumour necrosis factor (hTNF-α) antibody production: higher (1) or lower (0) than the cut-off value of anti-hTNF-α antibody geometric mean that best allows to discriminate between TNF-K and TNF-K + infliximab (IFX) groups (4211) (see Fig. 4). Mice with higher anti-hTNF-α antibody titres had both lower inflammation and destruction scores. *P < 0·05.

The hTNF-α neutralizing capacity was evaluated by L929 cytotoxic assay at weeks 5, 14 and 16 on pooled sera from each group. The pooled sera from the TNF-K group displayed higher hTNF-α neutralizing capacity versus those from the TNF-K + IFX group (P < 0·05) (not shown). Conversely, the TNF-K and IFXw0–15 groups displayed similar neutralizing capacity.

IFX serum levels and histological scores

As expected, the geometric means of IFX serum levels were higher for all mice in the IFXw0–15 group versus all those in the IFXw0–4 group (P < 0·001). The geometric mean of the serum levels of IFX showed inverse correlation with histological inflammation (R = −0·56; P < 0·05) and destruction scores with haematoxylin and eosin (R = −0·57; P < 0·05) (Fig. 4c,d), and with safranin O staining (R = −0·56; P < 0·05) (not shown).

Lower efficiency of TNF-K immunization in the TNF-K + IFX group

We hypothesized that the lower production of anti-hTNF-α antibody in the group that received TNF-K and IFX co-administration might be due to binding of IFX to the molecules of TNF-α exposed by the TNF-K, with consequent hindered interaction between TNF-K and B cell receptors and/or formation of immune complexes between TNF-K and IFX, with subsequent higher clearance of both. To evaluate this hypothesis, we assessed whether IFX would bind to TNF-K in a direct antigenicity test. As shown in Fig. 6, IFX binds to TNF-K, which confirms that the molecules of hTNF-α exposed by the TNF-K keep expressing the conformational epitope recognized by IFX, even after chemical inactivation and coupling to KLH. We also observed that the serum levels of IFX in the TNF-K + IFX group at the three blood samplings were always below the detection threshold, while the IFXw0–4 group, which had received IFX on the same schedule, had detectable levels of IFX at day 35 (P < 0·001 for geometric mean comparison between the two groups).

Figure 6.

Binding of infliximab to tumour necrosis factor-α-kinoid (TNF-K) (direct antigenicity).


In this study we show that the histological efficacy of TNF-K correlated with the titres of anti-hTNF-α antibodies induced by active anti-hTNF-α immunization. The co-administration of a short course of IFX with TNF-K overcame the latency in TNF-K clinical effect, but did not result in lower histological damage versus either strategy alone. Both TNF-K and long-term IFX treatments resulted in comparable clinical control of arthritis and prevention of histological inflammation and damage. IFX had a more rapid clinical effect than TNF-K, but this did not result in less histological damage at the end of the study (week 16). In addition, the TNF-K group was the only one to show significantly lower histological scores compared to the group that received IFX over 4 weeks and versus the group that received both TNF-K and IFX. In all TNF-K immunized mice (both TNF-K and TNF-K + IFX groups), histological destruction and inflammation scores showed a significantly good inverse correlation with the titres of anti-hTNF antibodies, and mice with higher serum levels of anti-hTNF-α induced by TNF-K immunization had significantly lower histological scores. This provides evidence for a dose–response effect of anti-hTNF antibodies induced by TNF-K, which explains the efficacy of active immunization. A similar correlation between IFX serum levels and histological scores was found in IFX -treated mice. It is noticeable that, in TTg mice, we have demonstrated previously the importance of anti-hTNF-α antibody generated after TNF-K immunization, as the transfer of sera from TNF-K immunized mice to naive mice induced strong protection from TNF-α-galactosamine-induced shock [14]. The anti-TNF-α inhibition capacity of sera (with Kd ranging from 5 × 10−8 M to 10−10 M [14]) was not higher than those of IFX (Kd from 10−9 to 10−12 M [21]) or adalimumab (Kd between 5·8 and 8·7 10−11 M [22]), which is reassuring concerning potential excessive TNF-α inhibition on TNF-K treatment.

Co-administration of active (TNF-K) and passive (IFX) anti-hTNF-α immunization overcomes the delay in therapeutic activity of TNF-K, but the group that received the co-administration did not fare better than the groups receiving either treatment alone with regard to global clinical scores of arthritis. Moreover, this group had higher histological scores versus either treatment alone (with differences that reached significance versus TNF-K group). This worse histological effect may be due to the fact that in the TNF-K + IFX group, and specifically in six of 10 mice, lower levels of anti-hTNF-α antibody were detected compared to the group that received only TNF-K. When histological scores were compared after categorization based not on the group of treatment, but on the titres of anti-hTNF-α antibody, the mice with low titres had significantly higher histological inflammation and destruction versus all other TNF-K-treated mice (from both TNF-K and TNF-K + IFX groups).

Thus, the TNF-K + IFX-treated mice were protected at the beginning of the experiment by IFX administration over the first 4 weeks while, later on, when the effect of IFX subsided, they had lower protection from anti-hTNF-α antibody induced by TNF-K immunization compared to the TNF-K group. Conversely, anti-KLH antibody titres did not differ between the two groups, suggesting that the lower efficacy of immunization is not dependent on the lack of T cell help. As TNF-K and IFX were administered together, we can speculate that IFX might have bound some molecules of hTNF-α onto TNF-K. This binding could prevent anti-TNF-α-specific B cell activation or accelerate clearance of immune complexes between IFX and TNF-K, or both. To partially support this hypothesis, we demonstrated that IFX binds efficiently to TNF-K in vitro, and that the process of chemical inactivation and coupling of hTNF-α to KLH does not alter the epitope recognized by IFX on the hTNF-α molecule. Moreover, the mice in the TNF-K + IFX group had undetectable serum levels of IFX, while in the IFXw0–4 group (which had received IFX on the same schedule) the drug was detectable. This supports a possible higher elimination of IFX in the former group, due possibly to the formation of immune complexes between IFX and TNF-K. The main limit of the study is that the co-administration strategy was tested with only one anti-TNF-α agent (IFX). It would be of interest to test the clinical and histological effect of the co-administration of TNF-K with other anti-TNF-α treatments, such as adalimumab, etanercept or certulizumab pegol. Clinically, it would be important to determine whether or not the lack of efficacy of the co-administration is IFX-restricted.

The results of this study bring two important elements for the future development of active anti-TNF-α inhibition in RA. First, the histological efficacy of active immunization, comparable to that of a long-term treatment with a standard anti-TNF-α drug, depends upon the amount of induced anti-hTNF-α antibodies. Secondly, the association of a passive (IFX) and active anti-TNF-α treatment allows a more rapid disease control compared to TNF-K, but does not result in less articular damage. The identification of prognostic factors of therapeutic success and the effect of the co-administration of passive and active anti-TNF-α immunization are major issues of interest for the further potential clinical development of TNF-K in human disease.

These results were obtained in the TTG mouse, which develops a spontaneous erosive polyarthritis dependent mainly on deregulated constitutive production of hTNF-α. As anti-TNF antibodies generated by TNF-K target hTNF-α and not murine TNF-α, this model is the only relevant one to test the efficacy of an anti-hTNF-α treatment.


We thank Monique Etienne, Simone Béranger (University of Paris 13), Stéphane Chambris (animal facilities, University of Paris 13), Delphine Lemeiter and Moufida Mahmoud Bacha (University of Paris 13, EA4222, Li2P) and Beatrice Drouet (Neovacs) for their outstanding technical assistance.


TNF-K is patented and the patent is held by Neovacs SA (Paris, France). G.G.V., E.B. and O.D. are scientist employees with Neovacs SA. M.C.B. has been a consultant for Neovacs SA and his laboratory has received unrestricted research grants from Neovacs, Pfizer, UCB Pharma and Roche. The other authors declare that they have no competing interests.