Female and male TNFα knockout mice and backcrossed control animals on a C57Bl6 background were bred at the Institute of Immunology, Biomedical Sciences Research Center 'Al. Fleming', Greece. The animals were housed in the animal facility of the Department of Rheumatology and Inflammation Research, University of Göteborg, Sweden. The mice were kept under standard conditions of temperature and light, and were fed laboratory chow and water ad libitum. The study was approved by the Ethical Committee of Göteborg University, and the requisitions of the National Board for Laboratory Animals were followed.

Mouse recombinant HMGB1 (rHMGB1) was expressed in Escherichia coli and purified to homogeneity as previously described 8. Preparations were tested for LPS content by the chromogenic Limulus amebocyte lysate assay and contained <2 ng endotoxin/μg rHMGB1. Highly purified recombinant endotoxin-free HMGB1 (pHMGB1) (purchased from HMGBiotech, Milano, Italy) was also used for experiments. LPS from E. coli serotype 055:B5 was purchased from Sigma (Saint Louis, MO, USA).

In the first experiment, TNFα^-/- mice and backcrossed control animals on a C57Bl6 background were injected intraarticularly with 5 μg pHMGB1. In the second experiment, mice received the intraarticular injection of 5 μg rHMGB1 and the contralateral knee was injected with 10 ng LPS, which corresponded to the amount of LPS found in rHMGB1 preparations.

Three days after the intraarticular injections, the optimal time to trigger synovitis 4, the mice were sacrificed. The knee joints were removed, fixed in 4% formaldehyde, decalcified, embedded in paraffin, sectioned and stained with H&E. All of the slides were assessed in a blinded manner by two researchers (RP and AT). The extent of synovitis was judged on an arbitrary scale from grade 0 to grade 3, as described elsewhere 4.

Spleen cells from TNFα^-/- mice and from control mice were prepared as previously described 9 and were stimulated with different doses of rHMGB1 and pHMGB1. The corresponding amount of LPS from E. coli was used for cell stimulation as a control for rHMGB1. Cell culture supernatants were collected after 24 hours and 48 hours for determination of the IL-6 level, as previously described 9.

For the in vitro proliferation assay, splenocyte cultures were prepared; the cells were then incubated for 72 hours in 96-well plates with final concentrations of 0.05 μg/ml, 0.5 μg/ml, and 5 μg/ml pHMGB1. Culture medium was used as a negative control, and concavalin A at a concentration of 2.5 μg/ml as a positive control. The cultures were pulsed with 1 μCi tritiated thymidine 12 hours before harvesting, and the tritiated thymidine uptake was counted in a beta counter. The proliferative response is expressed as the mean ± standard error of the mean (median) counts per minute of triplicate samples from five spleens in each group.

Nonparametric methods were used for statistical comparisons since the data showed a non-normal distribution. Statistical differences between independent groups were calculated using the Mann–Whitney U test. P < 0.05 was considered significant.

To assess the importance of the TNFα signalling pathway in HMGB1-triggered arthritis, TNFα^-/- mice and control TNFα^+/+ mice were given a single intraarticular injection of 5 μg pHMGB1 into a knee joint. This dose has been established in previous experiments to induce arthritis 4. Intraarticular injection of HMGB1 into healthy mouse joints resulted in an overall frequency of 32% to 39% arthritic animals as assessed histologically. No significant differences were found with respect to the severity and incidence of arthritis between mice deficient for TNFα (seven out of 18 mice with arthritis) in comparison with control TNFα^+/+ animals (six out of 19). The inflammation was characterized by mild synovitis (Figure 1).

Six mice in both groups received 5 μg rHMGB1 intraarticularly, which is known to contain a higher concentration of LPS. As a control, the contralateral knee joint of those mice was injected with 10 ng LPS, a dose corresponding to the LPS present in the rHMGB1 preparations used for injection. We could not observe any significant differences regarding arthritis severity in mice receiving rHMGB1 between mice with or mice without the functional TNFα gene (Figure 1). Two mice out of the six injected with 10 ng LPS displayed mild synovitis in both groups irrespective of the presence or absence of the TNFα gene. These results indicate that HMGB1 induces joint inflammation in vivo independently of the TNFα pathway.

We next examined whether TNFα deficiency affects a proinflammatory cytokine response to HMGB1 in vitro. Mouse splenocytes from TNFα^-/- mice and from control TNFα^+/+ mice were stimulated with 0.05 μg, 0.5 μg and 5 μg pHMGB1 for 24 hours and 48 hours, and the IL-6 production in cell culture supernatants was determined. Stimulation with pHMGB1 did not induce significant IL-6 release from TNFα^-/- cells as compared with TNFα^+/+ splenocytes at any time point and irrespective of the pHMGB1 concentration used (Table 1).

In another experiment, spleen cells from TNF^-/- mice and from control mice were prepared and stimulated with 0.5 μg/ml and 5 μg/ml rHMGB1 for 24 hours and 48 hours. As a control, the splenocytes were stimulated with 10 ng LPS, a dose corresponding to that present in the highest HMGB1 preparations. The results show that, upon stimulation with 0.5 μg/ml rHMGB1, the TNFα^+/+ cells released significantly more IL-6 than cells from the knockout mice after 24 hours (P < 0.05) and 48 hours (P < 0.03) (Table 1). No significant differences were detected, however, between spleen cells from TNFα^+/+ mice versus TNFα^-/- mice regarding IL-6 production upon stimulation with 5 μg/ml rHMGB1 after 24 hours and 48 hours. As a control, stimulation with 10 ng/ml LPS (corresponding to the amount found in 5 μg/ml rHMGB1) induced approximately 100 times less IL-6 than stimulation with 5 μg/ml rHMGB1 in TNFα^+/+ cultures as well as TNFα^-/- splenocyte cultures after 24 hours and 48 hours (Table 1).

To assess the impact of HMGB1 on the reactivity of immunocompetent cells in the presence or absence of the TNFα gene, spleen cells from TNFα^+/+ mice and from knockout mice were stimulated with different concentrations of endotoxin-free pHMGB1 and the proliferative response was scored after 72 hours. Upon stimulation with the highest HMGB1 dose (5 μg/ml), TNFα^+/+ mice had a significantly better proliferative response than their knockout littermates (285 ± 51 (median 264) counts per minute versus 197 ± 15 (median 185) counts per minute, respectively; P < 0.05), whereas no differences were seen regarding proliferation at lower HMGB1 concentrations (Figure 2). In addition, knockout mice had a threefold to fourfold lower response to concavalin A, a compound known to act on T lymphocytes, as compared with the TNFα^+/+ mice (3,700 ± 246 (median 3,668) counts per minute versus 7,423 ± 1,043 (median 6,550) counts per minute, respectively; P = 0.009).