Differential impacts of TNFα inhibitors on the transcriptome of Th cells
Arthritis Research & Therapy volume 23, Article number: 199 (2021)
Targeting TNFα is beneficial in many autoimmune and inflammatory diseases, including rheumatoid arthritis. However, the response to each of the existing TNFα inhibitors (TNFis) can be patient- and/or disease-dependent. In addition, TNFis can induce the production of type 1 interferons (IFNs), which contribute to their non-infection side effects, such as pustular psoriasis. Thus far, the molecular mechanisms mediating the drug-specific effects of TNFis and their induction of type 1 IFNs are not fully understood.
Peripheral blood mononuclear cells (PBMCs) were collected from healthy donors and stimulated in vitro with anti-CD3 and anti-CD28 in the absence or presence of adalimumab, etanercept, or certolizumab. Th cells were isolated from the stimulated PBMCs, and their RNA was subjected to RNA-seq and quantitative polymerase chain reaction.
Adalimumab and etanercept, which contain Fc, but not certolizumab, which does not contain Fc, inhibited the expression of several effector cytokines by Th cells within anti-CD3/anti-CD28-stimulated PBMCs. Transcriptomic analyses further showed that adalimumab, but not certolizumab, reciprocally induced type 1 IFN signals and the expression of CD96 and SIRPG in Th cells. The unique effects of adalimumab were not due to preferential neutralization of soluble TNFα but instead were mediated by several distinct mechanisms independent or dependent of Fc-facilitated physical interaction between Th cells and CD14+ monocytes.
TNFis can have drug-specific effects on the transcriptional profile of Th cells.
TNFα inhibitors (TNFis) have emerged as one of the most effective classes of drugs for inflammatory arthritis, including rheumatoid arthritis, psoriatic arthritis, and ankylosing spondylitis. Thus far, there are five commonly used TNFis , including adalimumab (ada), golimumab (gol), etanercept (eta), certolizumab pegol (cert), and infliximab (inf). Ada and gol are human IgG1; eta contains recombinant human trimeric type 2 TNFα receptors fused to human IgG1 Fc; cert is pegylated Fab’ without Fc; inf is human-mouse chimeric IgG1. While there exist differences in affinity and binding valency to TNFα among TNFis, their efficacy and safety profile in rheumatoid arthritis are very comparable. However, some patients may respond to one TNFi but not the others; eta, unlike other TNFis, is not effective for Crohn’s disease or uveitis. The causes of these discrepant clinical observations are still unclear. In addition to higher risk of infection, post-hoc data has uncovered several unexpected side effects of TNFis, including anti-nuclear antibody (ANA), lupus-like diseases, demyelinating diseases, and pustular psoriasis. While there is no head-to-head comparison, cert may be less likely to induce ANA, lupus-like disease, and demyelinating diseases [2,3,4,5]. Furthermore, TNFis can induce type 1 IFN signals, which at least partly explain some of their non-infection side effects [6, 7]. How TNFis induce type 1 IFN signals is still not fully understood. The existing data suggests that neutralization of TNFα prevents the maturation of plasmacytoid dendritic cells (pDCs), which produce a high level of type 1 IFN but lose this ability upon maturation. TNFis therefore sustain the production of type 1 IFN by immature pDCs [6, 8]; however, additional mechanisms very likely exist. Elucidating the molecular mechanisms mediating the differences in bioactivity among TNFis and the induction of type 1 IFN signals will improve the efficacy and safety of treatments and bring us one step closer to personalized medicine in inflammatory arthritis.
TNFα is synthesized as a membrane-bound precursor of 233 amino acid residues (mTNFα), which is then cleaved by metalloproteases to become soluble TNFα (sTNFα) of 157 amino acid residues . sTNFα in trimeric form can bind to type 1 and type 2 TNFα receptors (TNFαRs). TNFαR1 is constitutively expressed in many types of cells, activates NF-kB, AP-1, and caspases in TNFαR1-expressing cells, and is the main mediator of the pro-inflammatory effects of TNFα. The expression of TNFαR2 is limited to immune cells and its contribution to TNFα-induced inflammation is not fully clarified. Emerging data has convincingly demonstrated that mTNFα can also function as a receptor and upon engagement with TNFαR1 or TNFαR2 triggers signaling events in mTNFα-expressing cells, a process known as reverse signaling . Reverse signaling through mTNFα has been shown to induce the expression of E-selectin in human Th cells , activate NF-kB driven-transcription in B lymphoma cells , enhance the production of TNFα in monocytes , and synergize with IL-2 to augment cytotoxicity of NK cells . Reverse signaling through mTNFα can also be triggered by the TFNis , resulting in apoptosis of T cells [15, 16], degranulation of neutrophils [15, 16], and induction of TGF-β in macrophages . Such reverse signaling events may contribute to the discordance in the therapeutic and/or side effects among the TNFis. Despite the observations, the functional consequence of mTNFα reverse signaling is still not fully understood.
Here we report that ada has effects on the transcriptome of primary human Th cells that are not seen with cert, including down-regulation of effector Th cytokines and induction of type 1 IFN signals. The unique effects of ada are not due to preferential neutralization of sTNFα and are mediated by at least three distinct mechanisms. Our data therefore suggest the presence of novel pathways for regulating the expression of Th cytokines and TNFi-induced production of type 1 IFN.
Healthy donor PBMCs were purified from leukoreduction collars obtained from the Crimson Biomaterials Collection Core Facility, which prospectively collects discarded clinical materials matching investigator-defined criteria against available information on clinical samples.
This study has been approved by Partners Human Research Committee (PHRC), Boston, MA. Informed consent was obtained from participants prior to inclusion to the studies.
Purification and stimulation of PBMCs and Th cells
PBMCs were isolated from leukoreduction collars by Ficoll-Paque PLUS (17-1440-03, GE Healthcare, Pittsburgh, PA) density gradient centrifugation and cryopreserved prior to use. PBMCs were then thawed and plated in 24-well plates (2–2.5 millions/1 ml/well) pre-coated with anti-CD3 (2.5 μg/ml, HIT3a clone, Biolegend, San Diego, CA) in the presence of absence of soluble anti-CD28 (2 μg/ml, Cat. #302914, Biolegend) as well as adalimumab, etanercept, certolizumab pegol, and tocilizumab at indicated concentrations for 24 h before harvesting. In some experiments, Th (CD4+) cells were purified from resting PBMCs by using the human CD4+ T cell isolation kit (Cat. #130-045-101, Miltenyi Biotec, Bergisch, Gladbach, Germany) first before stimulation with anti-CD3 and anti-CD28. In the transwell experiments, Th cells were purified from 5 × 106 of PBMCs and seeded on anti-CD3-coated inserts, whereas the remaining non-Th cells were plated in the bottom chambers of 24-well transwell plates (#140620, ThermoFisher, Waltham, MA). Exclusion of non-Th subsets from PBMCs was carried out with CD8 (#130-045-201, Miltenyi), CD19 (#130-050-301, Miltenyi), or CD14 (#130-050-201, Miltenyi) MicroBeads.
Human fibroblast like synoviocytes were a gift of Dr. Hung Nguyen and were cultivated in DMEM supplemented with 1% penicillin, 1% streptomycin, and 1% glutamine (Gibco) before stimulation with TNFα (10 ng/ml, PHC3015, Life Technology, Carlsbad, CA) for 24 h before harvest.
RNA-seq and data analysis
RNA was prepared from sorted CD4+ Th by using Direct-zol RNA microprep kits (#R2063, Zymo, Irvine, CA), and the concentration and quality were checked by spectrophotometry. RNA samples were delivered to Admera Health (South Plainfield, NJ) or Broad institute (Cambridge, MA) for TruSeq RNA library preparation (Illumina, San Diego, CA) and bulk RNA sequencing. Results were returned in FASTQ format, and transcript-level of RNA-seq analysis was performed using HISAT2-StringTie-Ballgown workflow. Differentially expressed genes were identified with Qlucore Omics Explorer (Lund, Sweden).
Quantitative RNA analysis
RNA isolation, reverse transcription, and quantitative PCR (qPCR) were performed as previously described . The transcript levels thus detected were normalized against that of actin from the same sample. The sequences of the primers used in qPCR are listed in Supplemental Table 3.
Sandwich ELISA was performed using the following kits: human IL-17A ELISA set (#433914, Biolegend), human IL-17F ELISA set (#DY-1335B-05, R&D systems, Minneapolis, MN), and human IL-2 ELISA set (#431804, Biolegend). All ELISA experiments were performed according to manufacturer’s instructions.
PBMCs were stained with anti-CD4 (#317450 and #300508, Biolegend), anti-CD8 (#344750 and #344714, Biolegend), anti-CD14 (#301815, Biolegend), anti-CD19 (#332224, Biolegend), anti-CD69 (#310904, Biolegend), anti-SIRPG (#336606, Biolegend), anti-CD96 (#338405, Biolegend), AF647- (#A20186, Invitrogen, Carlsbad, CA) conjugated adalimumab and control IgG1 (#403502, Biolegend). Stained cells were collected with FACSCanto or LSRFortessa (Becton Dickenson, Franklin Lakes, NJ), and the data was analyzed with FlowJo software (Becton Dickinson)
Statistical analyses were carried out with one-way ANOVA followed by multiple comparisons (Figs. 1, 3, 4H and 6D, and Supplemental Figures 1 and 4), unpaired two-tailed Student’s t test (Figs. 2A and 5A), and paired two-tailed Student’s t test (Figs. 5C–E and 6B, C). The data shown in bar graphs is mean and SEM.
Differential impacts of TNFis on the expression of Th cytokines
We have previously shown that both anti-CD28 and soluble TNFα inhibited the expression of PTPN22 , a gene that is associated with several autoimmune diseases , but only anti-CD28 promoted the production of IL-17A/F in anti-CD3 stimulated PBMCs. Interestingly, the expression of IL-17F in activated PBMCs was inhibited by ada or eta in a dose-dependent manner (Fig. 1A) but not by tocilizumab, a monoclonal human IgG1 against human IL-6R. Ada or eta inhibited the expression of IL-17F at a concentration close to their therapeutic concentration (3.5–50 μg/ml). Surprisingly, cert failed to do so even at a concentration as high as 100 μg/ml (Fig. 1A). Similar trends were observed when we examined the expression of IL-17A and IL-2 (Fig. 1B, C). The discordance between cert and the other two TNFis was also reflected in the protein levels of IL-17A and IL-17F in the supernatant of the stimulated PBMCs (Fig. 1D). By contrast, the expression of IL-10 was not affected by any of the three TNFis (Supplemental Figure 1).
One possible explanation for the discordance between cert and the other two TNFis is that cert is less potent in neutralizing endogenous sTNFα induced by anti-CD3/anti-CD28. This scenario is unlikely because cert not only restored the expression of PTPN22 in PBMCs (Fig. 1E) but also inhibited TNFα-induced expression of IL-6 by synovial fibroblasts as efficiently as ada or eta (Fig. 1F). In addition, exogenous IL-2, which is critical for the proliferation of T cells, did not rescue the expression of IL17A either (Supplemental Figure 2).
Differential impacts of adalimumab and certolizumab on the transcriptome of Th cells
Th cells are the main expressors of IL-17A/F and IL-2. Indeed, when we isolated Th cells from the stimulated PBMCs, we again found that the transcript levels of IL-17A/F and IL-2 in the purified Th cells were markedly reduced when the PMBCs were stimulated in the presence of ada (Fig. 2A). To examine the impact of ada on the gene expression profile of Th cells, we prepared RNA from Th cells isolated from PBMCs (N = 5), which were pre-stimulated with anti-CD3/anti-CD28 in the presence or absence of ada for 24 h. The RNA was then subjected to RNA-seq.
We subsequently used a threshold of false discovery rate ≤ 0.2 and fold change ≥ 1.8 (+drug/−drug ratio ≤ 0.55 or ≥ 1.8) to identified differentially expressed genes. The volcano plot of p values and fold changes was shown in Fig. 2B. There were only 10 genes, whose expression was downregulated by ada in all five donors (Fig. 2C and Supplemental Table 1). Satisfyingly, IL17A, IL17F, and IL2 were among the most down-regulated genes. There was also reduction in IFNG, IL9, IL21, and IL12RB2. IFNG, IL9, and IL21 are the signature cytokine genes of Th1, Th9, and Tfh cells, suggesting that the effect of ada is not limited to Th17 cells. Expectedly, STRING v11  identified signaling by interleukins as the most significant reactome among the 10 genes (Fig. 2C). PBMCs from two of the five donors (donor #4 and #5) were also stimulated in the presence of cert, enabling us to examine whether the ada-regulated genes were concordantly or discordantly regulated by cert (Fig. 2D). We found that the expression of the cytokine genes was not affected by cert based on the 1.8-fold threshold. By contrast, IL12RB2 and DIXDC1 were comparably downregulated by ada and cert.
Reversely, there were 38 genes, whose expression was concordantly upregulated by ada in all five donors (Fig. 2E, F, and Supplemental Table 2). STRING identified Interferon Signaling as the most significant reactome among the 38 genes (Fig. 2E), which includes MX1, IRF7, ISG15, and OAS3. Indeed, 18 of the 38 genes are type 1 interferon (IFN) inducible genes in human T cells according to Interferome v2.01 . The ada-upregulated genes also include CCL2, SIRPG, and CD96. The latter two genes are preferentially expressed in T cells according to Human Protein Atlas and have been implicated in regulating the function of T cells [23,24,25]. Among the 38 genes, only CCL2 and ISG15 were also induced by cert in donor #4 and #5 based on the 1.8-fold threshold; however, the fold increases were much lower compared to those caused by ada (Fig. 2F).
One possible explanation for the unique effect of ada is that ada, but not cert, inhibits the activation of Th cells. However, ada had no consistent effect on the expression of several transcription factors that are known to be induced in activated T cells and/or critical for the expression of the Th cytokines, such as NR4A1, TBX21, GATA3, and RORC (Fig. 2G). Nor did ada have any impact on the expression of other T cell activation genes, such as PDCD1, ICOS, or FASLG. (Supplemental Figure 3). Ada has been shown to inhibit the activation of T cells through modulating the expression/signaling of NOTCH1 ; however, we did not detect any change in the transcript level of NOTCH1 and NOTCH2 as well as their downstream gene HES1 in our RNA-seq (Fig. 2H). The transcript levels of NOTCH3, NOTCH4, and HES2 are too low (< 1) for meaningful comparison. In agreement with the data shown in Supplemental Figure 1, ada had no impact on the expression of IL10 (Fig. 2H).
We subsequently confirmed the differential expression of the cytokine genes, including IFNG, IL21, and IL9 (Fig. 3A), several type 1 IFN-inducible genes, such as ISG15, MX1, and OAS3 (Fig. 3B), as well as CCL2, SIRPG, and CD96 (Fig. 3C) with qPCR in Th cells obtained from stimulated PBMCs. SIRPG is a receptor-type transmembrane glycoprotein that is expressed almost exclusively in T cells. The unique effect of ada on the transcript level of SIRPG at 24 h was also reflected in the surface level of SIRPG in Th cells 96 h after stimulation (Fig. 3D, E). A similar effect was observed when we examined the surface level of CD96 (Supplemental Figure 4). While the transcript levels of CCL2 and ISG15 were increased by both ada and cert in our RNA-seq analysis, their expression was induced only by ada but not cert when quantified with qPCR.
Comparable binding of TNFis to mTNFα
We then set to investigate the mechanism mediating the unique effect of ada. In addition to neutralizing sTNFα, TNFis can also bind to mTNFα, thereby triggering reverse signaling. It is possible that the discordant effect between ada and cert is due to differential binding to mTNFα. We therefore labeled ada with AF647 and used the labeled ada to stain PBMCs in the presence of FcR block. While very weak staining was detected in Th cells within resting PBMCs, approximately 20–30% of Th (CD4+) cells were stained positive by ada when PBMCs were stimulated with anti-CD3/anti-CD28 for 24 h (Fig. 4A, B and Supplemental Figure 5). Kinetically, ada + Th cells started to appear in Th cells 4 h after stimulation and their percentage gradually increased over the course of 24 h. The staining was almost detected exclusively in CD69+ population. By contrast, much weaker ada staining was detected in CD8+ T cells (Fig. 4B, C) and B cells (Fig. 4B, D). Monocytes (CD4med, CD14+ and/or CD16+) were still identifiable 8 h, but not 24 h, after stimulation and also had weak staining of ada (Fig. 4B, E, and Supplemental Figure 5). The cause for the weak ada staining in non-Th cells is still unknown and could be due to low expression of TNFα and/or rapid cleavage of mTNFα. The ada staining of Th cells was comparably competed away by unlabeled ada, eta, or cert at a molar concentration equivalent to that of AF647-labeled ada (Fig. 4F), suggesting that ada indeed binds to mTNFα and that the discordant effect between ada/eta and cert is not due to differential binding to mTNFα. Both sTNFα and mTNFα can trigger activation-induced cell death of T cells through forward and reverse signaling, respectively [27,28,29]; however, we found that the three TNFis had no differential effect on apoptosis of Th cells in our system (Fig. 4G, H). Thus, the unique effect of ada and eta cannot be explained by a difference in apoptosis.
Non-Th cell-dependent and independent effects of adalimumab
We then examined whether ada acted directly on Th cells to inhibit their expression of cytokines. We purified Th cells first and then stimulated the Th cells with anti-CD3/anti-CD28 in the presence or absence of ada. Surprisingly, ada was ineffective in inhibiting the expression of IL-17A/F and IL-2 in the absence of non-Th cells even though it still enhanced the expression of PTPN22 (Fig. 5A). This observation suggests that the effect of ada on the expression Th cytokines requires non-Th cells. To further determine whether physical contact between Th and non-Th cells was necessary for the action of ada, we set up a transwell co-culture system, in which purified Th cells were stimulated in the top chambers, whereas autologous non-Th cells were cultured in the bottom chambers (Fig. 5B). We found that ada failed to inhibit the expression of IL-17A/F or IL-2 by Th cells when the Th cells were stimulated without physical contact with non-Th cells (Fig. 5C, Th cells transwell) or in the absence of non-Th cells (Fig. 5C, Th cells only). By contrast, the levels of the Th cytokines were reduced in Th cells purified from PBMCs that were stimulated in the presence of ada (Fig. 5C, Th cells PBMC).
Th cells typically do not express type 1 IFN. Expectedly, ada did not induce type 1 IFN signals in Th cells in the absence of non-Th cells (Fig. 5D, Th cells only). However, ada was able to augment the expression of ISG15, MX1, and OAS3 in Th cells from the transwell (Fig. 5D, Th cells transwell). This result indicates that the source of type 1 IFN is non-Th cells and that the ada-induced expression of type 1 IFN by non-Th cells does not require physical interaction between Th and non-Th cells. By contrast, ada effectively enhanced the expression of CD96 and SIRPG in Th cells from the transwell system and also in the absence of non-Th cells (Fig. 5E).
Dependence of CD14+ monocyte and Fc-FcR interaction for the effects of ada on Th cytokines
Type 1 IFN has been shown to constrain Th17 cells by acting directly on Th cells to produce IL- IL-10 [30, 31] or acting on myeloid cells to express IL-27 , which then inhibits the differentiation of Th17 cells. Thus, ada could inhibit the expression of Th cytokines through promoting the expression of type 1 IFN by non-Th cells. This scenario is unlikely because the action of type 1 IFN should not depend on physical contact between Th and non-Th cells. In addition, neither ada nor cert affected the expression of IL-10 by Th cells (Fig. 2H and Supplemental Figure 1). To further elucidate the mechanism of action of ada, we set to identify the non-Th cells that are required for ada to inhibit the expression of Th cytokines. We separately deleted CD8+, CD19+ or CD14+ cells from PBMCs (Fig. 6A). The depleted PBMCs were then stimulated with anti-CD3/anti-CD28 in the absence or presence of ada. Depletion of CD14+ cells alone or together with CD19+ completely abrogated the effect of ada on the expression of IL-17A and IL-2 (Fig. 6B). Ada still subtly inhibited the expression of IL-17F in the absence of CD14+ cells but this residual effect was wiped out by additional depletion of CD19+ cells. By contrast, depletion of CD8+ cells or CD19+ cells alone had little or no effect at all. Supernatant from three of the experiments was also subjected to IL-2 ELISA and the results were consistent with that obtained with qPCR (Fig. 6C). These results indicate that CD14+ cells, but not CD8+ or CD19+ cells, are essential for mediating the inhibitory effect of ada on the expression of Th cytokines. Both ada and eta, but not cert, contain the Fc of human IgG1. It is possible that the unique effect of ada and eta, once bound to mTNFα on Th cells, requires the interaction of their Fc with FcR expressed by CD14+ monocytes and/or other non-Th cells. In agreement with this scenario, we found that pre-treating PBMCs with control human IgG1, which does not compete with ada or eta for binding to mTNFα but is expected to compete with the two Fc-containing TNFis for interaction with FcR, completely blocked the inhibitory effect of ada and eta on the expression of IL-17A and IL-17F (Fig. 6D).
The interaction between Fc of ada and FcR of non-Th cells may cause cross-linking of T cell-bound ada, thereby inhibiting the expression of Th cytokines. However, we found that cross-linking of T cell-bound ada with plate bound antibodies against the Fc of human IgG1 was unable to inhibit the expression of cytokine in purified Th cells (Supplemental Figure 6). Taken together, these results indicate that the unique effect of ada and eta depends on the interaction between their Fc and FcR on non-Th cells and that the contribution of such Fc-FcR interaction is more than causing cross-linking of Th cell-bound ada or eta.
We have uncovered intriguing differences in the biological effect between ada and cert. While both TNFis comparably neutralize sTNFα and induce the expression of PTPN22, ada has additional impacts on the transcriptome of Th cells through at least three distinct mechanisms. It inhibits the expression of Th cytokines, an effect depending on its Fc and physical contact between Th cells and CD14+ monocytes (A in Supplemental Figure 7), induces type 1 IFN signals in Th cells, an effect depending on soluble factors from non-Th cells (B in Supplemental Figure 7), and enhances the expression of CD96 and SIRPG, an effect independent of non-Th cells (C in Supplemental Figure 7).
The observation that ada inhibits the production of Th cytokines only in stimulated PBMCs but not purified Th cells is also consistent with published data [26, 33]; however, several discrepancies exist between our results and the published data. Werner et al. reported that ada and inf at 50 μg/ml, a dose comparable to that used in our study, inhibited the expression of IL-17 and IFNγ in PBMCs stimulated with anti-CD3 for 48 h . Unfortunately, cert was not examined in Werner’s study. Regardless, their data suggest that TNFis attenuates the level of transmembrane Notch-1 and the expression of its downstream gene Hes-1 in T cells. Their data further suggest that the attenuated Notch-1 signaling by TNFis leads to the inhibition of Th cell activation and consequently the expression of Th cytokines. While this putative mechanism can explain the need for physical contact between Th and non-Th cells, it is in conflict with several publications showing a positive role of Notch signals in the activation of T cells [34,35,36,37]. In addition, our RNA-seq data shows that the expression of Notch1/2 and their downstream gene Hes-1 was not affected by ada. Povoleri et al. reported that ada at 1 μg/ml had no effect on the expression of IL-17 and IFNγ by purified CD4+ T cells when analyzed 3 days after stimulation , a finding consistent with our finding that ada did not affect the cytokine expression in Th cells in the absence of non-Th cells. However, they showed that ada inhibited the activation of purified CD4+ T cells and induced the expression of IL-10. These observations are different from our data showing that the expression of IL-10 and several activation-induced genes, such as NR4A1 and PDCD1, in Th cells from stimulated PBMCs was not affected by TNFis. The cause of this discrepancy is unclear and could be due to differences in the duration of stimulation, absence or presence of non-Th cells, and/or dose of ada.
We stimulated PBMCs with plate-bound anti-CD3 and soluble anti-CD28 to optimize the activation of Th cells. This setting may overtake physiological interaction between CD28 on Th cells and B7 on non-Th cells, thereby potentially making it difficult to detect the impact of non-Th cells on the differential effects of TNFis. However, the data shown in Fig. 6 clearly demonstrates a critical role of physical contact between Th cells and non-Th cells, particularly CD14+ cells, in mediating the inhibitory effect of ada on the production of Th cytokines. Thus, non-Th cells are not redundant but instead actively contribute to the differential effect of TNFis in our culture system. Despite our results, it remains unclear why the physical interaction between Th and CD14+ monocytes is required for the inhibitory effect of ada on the expression of Th cytokines. Our data strongly suggest that this interaction is mediated mainly by the Fc of Th cell-bound ada and FcR on monocytes. Such Fc-FcR interaction may induce the production of a soluble factor or the expression of a co-inhibitory surface molecule by monocytes. The transwell experiment shown in Fig. 5 argues against the former scenario. It is intriguing to notice that the expression of CD96 and SIRPG is enhanced by ada. Both CD96 and SIRPG are expressed preferentially, if not exclusively, in T and NK cells [23, 24]. CD96 can function as a co-suppressor and is potentially a target of immune checkpoint therapy . SIRPG has been shown to interact with CD47 , which is ubiquitously expressed, but may have additional ligands. Its function is still unclear but several SNPs at the SIRPG locus are associated with higher risk of type 1 diabetes [39, 40]. While the induction of CD96 and SIRPG by ada is independent of non-Th cells, the expression of their counter-receptors, such as CD111, CD155, and CD47, on monocytes may still depend on the Fc-FcR interaction. It will be of great interest to determine whether the CD96 and/or SIRPG pathway contribute to the inhibitory effect of ada on the expression of Th cytokines.
Alternatively, the Fc-FcR interaction may trigger reverse signaling through ada-bound mTNFα in Th cells. mTNFα reverse signaling can also be triggered by TNFis, resulting in different functional outcomes . For example, inf, gol, and ada induce apoptosis in T cells, and this effect requires three serine residues within the intracellular domain of mTNFα [15, 16], whereas cert induces a mTNFα-dependent non-apoptotic type of cell death . The discrepancy in the functional consequence of mTNFα reverse signaling induced by different TNFis could be due to differences in the affinity and/or binding valency of various TNFis. If the unique effect of ada on Th cytokine expression is due to reverse signaling through mTNFα in Th cells, our data will not only uncover a potentially novel consequence of ada-induced mTNFα reverse signaling, i.e., inhibiting the expression of Th cytokines, but also suggest that the interaction between Fc of TNFis and FcR on non-Th cells can heavily influence the outcome of mTNFα reverse signaling. In agreement with this latter point, inf and ada, but not cert, have been shown to inhibit the proliferation of CD4+ T cells activated by allogeneic PBMCs and induce a distinct CD14+/HLA-DR+ macrophage population in a Fc-dependent manner . It is still unclear whether the inhibitory effect of ada and eta on Th cytokine production can translate into in vivo situation, where endogenous IgG1 may dampen the effect of the TNFis. We do not think the scenario shown in Supplemental Figure 7 occurs in peripheral blood. Instead, we envision that the scenario is playing out in tissue, such as draining lymph nodes or inflamed synovium, where Th cells are activated and/or the concentration of TNFα is high enough to attract TNFis, thereby altering the local ratio between TNFis and endogenous IgG1. Unfortunately, there is still little information on the relative tissue distribution of TNFis and endogenous IgG1 in vivo.
Long-term treatment with TNFis is known to induce type 1 IFN signals, which contribute to the development of some of the non-infection side effects of TNFis, such as pustular psoriasis [6, 7]. Existing data suggest that TNFis lead to an arrest of pDCs at an immature TNFα-producing stage [6, 8]. Our data, however, suggest that short-term exposure of stimulated PBMCs to ada but not cert also induces the expression of type 1 IFN, probably by non-Th cells. We still do not have physical evidence for the presence of type 1 IFN. Nor do we know how ada uniquely induces the type 1 IFN signals. As the type 1 IFN signals were also detected in Th cells from the transwell, its induction is independent of Th/non-Th interaction. One plausible scenario is that ada and cert, while binding comparably to mTNFα, trigger a distinct reverse signaling cascade inside non-Th cells, resulting in the expression of type 1 IFN. This scenario is consistent with published data showing that various TNFis can act directly on myeloid cells with discordant outcomes [42,43,44]. For example, eta is less effective in inhibiting LPS-induced expression of IL-1β by human monocytes, whereas cert does not elicit the release of myeloperoxidase from human polymorphonucleocytes . If ada indeed induces the expression of type 1 IFN, better understanding of the molecular mechanism mediating this effect of ada may lead to novel approaches of inhibiting TNFis-induced production of type 1 IFN and preventing some of the non-infection side effects of TNFis.
Adalimumab compared to certolizumab has an unique impact on the gene expression of Th cells through at least three distinct mechanisms that are independent of neutralizing soluble TNFα.
Availability of data and materials
The raw RNA-seq data will be available at NCBI database or upon request to the corresponding author.
Peripheral blood mononuclear cells
Plasmacytoid dendritic cells
Quantitative polymerase chain reaction
Mitoma H, Horiuchi T, Tsukamoto H, Ueda N. Molecular mechanisms of action of anti-TNF-alpha agents - comparison among therapeutic TNF-alpha antagonists. Cytokine. 2018;101:56–63. https://doi.org/10.1016/j.cyto.2016.08.014.
Verma HD, Scherl EJ, Jacob VE, Bosworth BP. Anti-nuclear antibody positivity and the use of certolizumab in inflammatory bowel disease patients who have had arthralgias or lupus-like reactions from infliximab or adalimumab. J Dig Dis. 2011;12(5):379–83. https://doi.org/10.1111/j.1751-2980.2011.00522.x.
Beigel F, Schnitzler F, Paul Laubender R, Pfennig S, Weidinger M, Goke B, et al. Formation of antinuclear and double-strand DNA antibodies and frequency of lupus-like syndrome in anti-TNF-alpha antibody-treated patients with inflammatory bowel disease. Inflamm Bowel Dis. 2011;17(1):91–8. https://doi.org/10.1002/ibd.21362.
Schreiber S, Khaliq-Kareemi M, Lawrance IC, Thomsen OO, Hanauer SB, McColm J, et al. Maintenance therapy with certolizumab pegol for Crohn’s disease. N Engl J Med. 2007;357(3):239–50. https://doi.org/10.1056/NEJMoa062897.
Kemanetzoglou E, Andreadou E. CNS Demyelination with TNF-alpha blockers. Curr Neurol Neurosci Rep. 2017;17(4):36. https://doi.org/10.1007/s11910-017-0742-1.
Conrad C, Di Domizio J, Mylonas A, Belkhodja C, Demaria O, Navarini AA, et al. TNF blockade induces a dysregulated type I interferon response without autoimmunity in paradoxical psoriasis. Nat Commun. 2018;9(1):25. https://doi.org/10.1038/s41467-017-02466-4.
De Bandt M. Anti-TNF-alpha-induced lupus. Arthritis research & therapy. 2019;21(1):235. https://doi.org/10.1186/s13075-019-2028-2.
Palucka AK, Blanck JP, Bennett L, Pascual V, Banchereau J. Cross-regulation of TNF and IFN-alpha in autoimmune diseases. Proc Natl Acad Sci U S A. 2005;102(9):3372–7. https://doi.org/10.1073/pnas.0408506102.
Horiuchi T, Mitoma H, Harashima S, Tsukamoto H, Shimoda T. Transmembrane TNF-alpha: structure, function and interaction with anti-TNF agents. Rheumatology (Oxford, England). 2010;49(7):1215–28.
Szondy Z, Pallai A. Transmembrane TNF-alpha reverse signaling leading to TGF-beta production is selectively activated by TNF targeting molecules: therapeutic implications. Pharmacol Res. 2017;115:124–32. https://doi.org/10.1016/j.phrs.2016.11.025.
Harashima S, Horiuchi T, Hatta N, Morita C, Higuchi M, Sawabe T, et al. Outside-to-inside signal through the membrane TNF-alpha induces E-selectin (CD62E) expression on activated human CD4+ T cells. J Immunol. 2001;166(1):130–6. https://doi.org/10.4049/jimmunol.166.1.130.
Zhang H, Yan D, Shi X, Liang H, Pang Y, Qin N, et al. Transmembrane TNF-alpha mediates “forward” and “reverse” signaling, inducing cell death or survival via the NF-kappaB pathway in Raji Burkitt lymphoma cells. J Leukoc Biol. 2008;84(3):789–97. https://doi.org/10.1189/jlb.0208078.
Rossol M, Meusch U, Pierer M, Kaltenhauser S, Hantzschel H, Hauschildt S, et al. Interaction between transmembrane TNF and TNFR1/2 mediates the activation of monocytes by contact with T cells. J Immunol. 2007;179(6):4239–48. https://doi.org/10.4049/jimmunol.179.6.4239.
Yu M, Shi W, Zhang J, Niu L, Chen Q, Yan D, et al. Influence of reverse signaling via membrane TNF-alpha on cytotoxicity of NK92 cells. Eur J Cell Biol. 2009;88(3):181–91. https://doi.org/10.1016/j.ejcb.2008.09.001.
Ueda N, Tsukamoto H, Mitoma H, Ayano M, Tanaka A, Ohta S, et al. The cytotoxic effects of certolizumab pegol and golimumab mediated by transmembrane tumor necrosis factor alpha. Inflamm Bowel Dis. 2013;19(6):1224–31. https://doi.org/10.1097/MIB.0b013e318280b169.
Mitoma H, Horiuchi T, Hatta N, Tsukamoto H, Harashima S, Kikuchi Y, et al. Infliximab induces potent anti-inflammatory responses by outside-to-inside signals through transmembrane TNF-alpha. Gastroenterology. 2005;128(2):376–92. https://doi.org/10.1053/j.gastro.2004.11.060.
Pallai A, Kiss B, Vereb G, Armaka M, Kollias G, Szekanecz Z, et al. Transmembrane TNF-alpha reverse signaling inhibits lipopolysaccharide-induced proinflammatory cytokine formation in macrophages by inducing TGF-beta: therapeutic implications. J Immunol. 2016;196(3):1146–57. https://doi.org/10.4049/jimmunol.1501573.
Chang HH, Tai TS, Lu B, Iannaccone C, Cernadas M, Weinblatt M, et al. PTPN22.6, a dominant negative isoform of PTPN22 and potential biomarker of rheumatoid arthritis. PLoS One. 2012;7(3):e33067.
Chang HH, Ho CH, Tomita B, Silva AA, Sparks JA, Karlson EW, et al. Utilizing a PTPN22 gene signature to predict response to targeted therapies in rheumatoid arthritis. J Autoimmun. 2019;101:121–30. https://doi.org/10.1016/j.jaut.2019.04.012.
Rawlings DJ, Dai X, Buckner JH. The role of PTPN22 risk variant in the development of autoimmunity: finding common ground between mouse and human. J Immunol. 2015;194(7):2977–84. https://doi.org/10.4049/jimmunol.1403034.
Szklarczyk D, Gable AL, Lyon D, Junge A, Wyder S, Huerta-Cepas J, et al. STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2019;47(D1):D607–D13. https://doi.org/10.1093/nar/gky1131.
Rusinova I, Forster S, Yu S, Kannan A, Masse M, Cumming H, et al. Interferome v2.0: an updated database of annotated interferon-regulated genes. Nucleic Acids Res. 2013;41(Database issue):D1040–6. https://doi.org/10.1093/nar/gks1215.
Georgiev H, Ravens I, Papadogianni G, Bernhardt G. Coming of age: CD96 emerges as modulator of immune responses. Front Immunol. 2018;9:1072. https://doi.org/10.3389/fimmu.2018.01072.
Barclay AN, Brown MH. The SIRP family of receptors and immune regulation. Nature reviews. 2006;6(6):457–64. https://doi.org/10.1038/nri1859.
Stanko K, Iwert C, Appelt C, Vogt K, Schumann J, Strunk FJ, et al. CD96 expression determines the inflammatory potential of IL-9-producing Th9 cells. Proc Natl Acad Sci U S A. 2018;115(13):E2940–E9. https://doi.org/10.1073/pnas.1708329115.
Werner L, Berndt U, Paclik D, Danese S, Schirbel A, Sturm A. TNFalpha inhibitors restrict T cell activation and cycling via Notch-1 signalling in inflammatory bowel disease. Gut. 2012;61(7):1016–27. https://doi.org/10.1136/gutjnl-2011-301267.
Zheng L, Fisher G, Miller RE, Peschon J, Lynch DH, Lenardo MJ. Induction of apoptosis in mature T cells by tumour necrosis factor. Nature. 1995;377(6547):348–51. https://doi.org/10.1038/377348a0.
Sytwu HK, Liblau RS, McDevitt HO. The roles of Fas/APO-1 (CD95) and TNF in antigen-induced programmed cell death in T cell receptor transgenic mice. Immunity. 1996;5(1):17–30. https://doi.org/10.1016/S1074-7613(00)80306-4.
Zhang M, Wang J, Jia L, Huang J, He C, Hu F, et al. Transmembrane TNF-alpha promotes activation-induced cell death by forward and reverse signaling. Oncotarget. 2017;8(38):63799–812. https://doi.org/10.18632/oncotarget.19124.
Zhang L, Yuan S, Cheng G, Guo B. Type I IFN promotes IL-10 production from T cells to suppress Th17 cells and Th17-associated autoimmune inflammation. PLoS One. 2011;6(12):e28432. https://doi.org/10.1371/journal.pone.0028432.
Ramgolam VS, Sha Y, Jin J, Zhang X, Markovic-Plese S. IFN-beta inhibits human Th17 cell differentiation. J Immunol. 2009;183(8):5418–27. https://doi.org/10.4049/jimmunol.0803227.
Guo B, Chang EY, Cheng G. The type I IFN induction pathway constrains Th17-mediated autoimmune inflammation in mice. J Clin Invest. 2008;118(5):1680–90. https://doi.org/10.1172/JCI33342.
Povoleri GAM, Lalnunhlimi S, Steel KJA, Agrawal S, O'Byrne AM, Ridley M, et al. Anti-TNF treatment negatively regulates human CD4(+) T-cell activation and maturation in vitro, but does not confer an anergic or suppressive phenotype. Eur J Immunol. 2020;50(3):445–58. https://doi.org/10.1002/eji.201948190.
Dua B, Upadhyay R, Natrajan M, Arora M, Kithiganahalli Narayanaswamy B, Joshi B. Notch signaling induces lymphoproliferation, T helper cell activation and Th1/Th2 differentiation in leprosy. Immunol Lett. 2019;207:6–16. https://doi.org/10.1016/j.imlet.2019.01.003.
Palaga T, Miele L, Golde TE, Osborne BA. TCR-mediated Notch signaling regulates proliferation and IFN-gamma production in peripheral T cells. J Immunol. 2003;171(6):3019–24. https://doi.org/10.4049/jimmunol.171.6.3019.
Dar AA, Bhat SA, Gogoi D, Gokhale A, Chiplunkar SV. Inhibition of Notch signalling has ability to alter the proximal and distal TCR signalling events in human CD3(+) alphabeta T-cells. Mol Immunol. 2017;92:116–24. https://doi.org/10.1016/j.molimm.2017.10.013.
Adler SH, Chiffoleau E, Xu L, Dalton NM, Burg JM, Wells AD, et al. Notch signaling augments T cell responsiveness by enhancing CD25 expression. J Immunol. 2003;171(6):2896–903. https://doi.org/10.4049/jimmunol.171.6.2896.
Brooke G, Holbrook JD, Brown MH, Barclay AN. Human lymphocytes interact directly with CD47 through a novel member of the signal regulatory protein (SIRP) family. J Immunol. 2004;173(4):2562–70. https://doi.org/10.4049/jimmunol.173.4.2562.
Westra HJ, Martinez-Bonet M, Onengut-Gumuscu S, Lee A, Luo Y, Teslovich N, et al. Fine-mapping and functional studies highlight potential causal variants for rheumatoid arthritis and type 1 diabetes. Nat Genet. 2018;50(10):1366–74. https://doi.org/10.1038/s41588-018-0216-7.
Reddy MV, Wang H, Liu S, Bode B, Reed JC, Steed RD, et al. Association between type 1 diabetes and GWAS SNPs in the southeast US Caucasian population. Genes Immun. 2011;12(3):208–12. https://doi.org/10.1038/gene.2010.70.
Vos AC, Wildenberg ME, Duijvestein M, Verhaar AP, van den Brink GR, Hommes DW. Anti-tumor necrosis factor-alpha antibodies induce regulatory macrophages in an Fc region-dependent manner. Gastroenterology. 2011;140(1):221–30. https://doi.org/10.1053/j.gastro.2010.10.008.
Nesbitt A, Fossati G, Bergin M, Stephens P, Stephens S, Foulkes R, et al. Mechanism of action of certolizumab pegol (CDP870): in vitro comparison with other anti-tumor necrosis factor alpha agents. Inflamm Bowel Dis. 2007;13(11):1323–32. https://doi.org/10.1002/ibd.20225.
Waetzig GH, Seegert D, Rosenstiel P, Nikolaus S, Schreiber S. p38 mitogen-activated protein kinase is activated and linked to TNF-alpha signaling in inflammatory bowel disease. J Immunol. 2002;168(10):5342–51. https://doi.org/10.4049/jimmunol.168.10.5342.
Shen C, Assche GV, Colpaert S, Maerten P, Geboes K, Rutgeerts P, et al. Adalimumab induces apoptosis of human monocytes: a comparative study with infliximab and etanercept. Aliment Pharmacol Ther. 2005;21(3):251–8. https://doi.org/10.1111/j.1365-2036.2005.02309.x.
We thank the anonymous blood donors of Crimson Core.
This work was supported by Flagship Program of Precision Medicine for Asia-Pacific Biomedical Valley, National Health Research Institute, Taiwan; National Institutes of Health (grant numbers AR070171, AR074788, and AR070253). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Ethics approval and consent to participate
This study has been approved by Partners Human Research Committee (PHRC), Boston, MA. Informed consent was obtained from participants prior to inclusion to the studies.
Consent for publication
Written informed consent and agreement for publication were obtained from all participants.
The authors declare that they have no competing interest.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Ho, CH., Silva, A.A., Tomita, B. et al. Differential impacts of TNFα inhibitors on the transcriptome of Th cells. Arthritis Res Ther 23, 199 (2021). https://doi.org/10.1186/s13075-021-02558-z
- TNFα inhibitors
- Th cells
- Type 1 interferon