Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`Sci Rep. 2016; 6: 31159.
`
`Published online 2016 Aug 8. doi: 10.1038/srep31159
`
`PMCID: PMC4976367
`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via
`inhibition of TLR induced M1 and K63 ubiquitin chain formation
`
`1
`1,2
`3
`3
`3
`Victoria A. McGuire, Tamara Ruiz-Zorrilla Diez, Christoph H. Emmerich, Sam Strickson, Maria Stella Ritorto, Ruhcha V.
`1
`1
`1
`3
`4
`1
`4
`Sutavani, Anne Weiβ, Kirsty F. Houslay, Axel Knebel, Paul J. Meakin,
`Iain R. Phair, Michael L. J. Ashford, Matthias
`3
`a,1
`Trost, and J. Simon C. Arthur
`
`Division of Cell Signaling and Immunology, School of Life Sciences, Wellcome Trust Building, University of Dundee, Dow St, Dundee, DD1 5EH, UK
`
`Department of Chemistry and Biochemistry, Faculty of Pharmacy, CEU San Pablo University, Urbanización Montepríncipe, 28668 Madrid, Spain
`
`MRC Protein Phosphorylation and ubiquitylation Unit, School of Life Sciences, Sir James Black Centre, University of Dundee, Dow St, Dundee, DD1
`
`1 2 3
`
`5EH, UK
`4
`
`Cardiovascular and Diabetes Medicine, Medical Research Institute, School of Medicine, University of Dundee, Ninewells Hospital, Dundee, DD1 9SY,
`
`UK
`a
`
`Email: j.s.c.arthur@dundee.ac.uk
`
`Received 2015 Oct 19; Accepted 2016 Jul 15.
`
`Copyright © 2016, The Author(s)
`
`This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are
`
`included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative
`
`Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit
`
`http://creativecommons.org/licenses/by/4.0/
`
`Abstract
`
`Go to:
`
`Dimethyl fumarate (DMF) possesses anti-inflammatory properties and is approved for the treatment of psoriasis
`
`and multiple sclerosis. While clinically effective, its molecular target has remained elusive - although it is known
`
`to activate anti-oxidant pathways. We find that DMF inhibits pro-inflammatory cytokine production in response
`
`to TLR agonists independently of the Nrf2-Keap1 anti-oxidant pathway. Instead we show that DMF can inhibit
`
`the E2 conjugating enzymes involved in K63 and M1 polyubiquitin chain formation both in vitro and in cells. The
`
`formation of K63 and M1 chains is required to link TLR activation to downstream signaling, and consistent with
`
`the block in K63 and/or M1 chain formation, DMF inhibits NFκB and ERK1/2 activation, resulting in a loss of
`
`pro-inflammatory cytokine production. Together these results reveal a new molecular target for DMF and show
`
`that a clinically approved drug inhibits M1 and K63 chain formation in TLR induced signaling complexes.
`
`Selective targeting of E2s may therefore be a viable strategy for autoimmunity.
`
`Autoimmune disorders represent a diverse range of conditions that remain challenging to treat. The advent of
`
`1
`biological drugs, such as anti-TNF agents, provided a significant advance in the treatment of these conditions ,
`
`however they have the disadvantages of not being orally available and that a proportion of patients do not
`
`respond. The development of new orally available small molecule drugs for autoimmunity is therefore desirable.
`
`Several breakthroughs in this area have recently been made, such as the development of Jak inhibitors and S1P
`2,3,4
`receptor modulating agents, which illustrate the potential of this approach
`.
`
`Dimethyl fumarate (DMF) is a methyl ester known to have immuno-modulatory properties. In combination with
`
`1 of 15
`
`10/18/2016 5:39 PM
`
`Biogen Exhibit 2387
`Coalition v. Biogen
`IPR2015-01993
`
`Page 1 of 15
`
`

`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`5
`other fumaric acid esters, DMF has been in use for many years as a treatment for moderate and severe psoriasis .
`6
`The first report of its use was in 1959 , although it did not gain widespread acceptance until some time later
`7
`following the publication of the first clinical trials demonstrating its efficacy in 1990 . Subsequently, DMF in
`8,9
`.
`
`combination with three salts of ethylhydrogenfumarate was licensed for use in psoriasis in Germany in 1994
`10
`.
`
`More recently, a slow release formulation of DMF has been approved for the treatment of multiple sclerosis
`
`The molecular target of DMF that accounts for its ability to modulate the immune system has been elusive.
`
`Amongst the possible explanations for its action, DMF has been shown to reduce T cell numbers, inhibit NFκB
`11,12
`). In addition, DMF has been found to
`
`mediated transcription and activate the Nrf2 pathway (reviewed in
`
`modulate cytokine production in a number of immune cell types: cytokine production is regulated by several
`
`intracellular signaling systems including NFκB and the ERK1/2 and p38 MAPK pathways, and DMF has been
`
`suggested to modulate these pathways. For example, DMF has been shown to prevent the induction of NFκB
`
`dependent transcription in LPS stimulated dendritic cells as well as TNF stimulated Human Umbilical Vein
`13,14,15
`.
`
`Endothelial Cells (HUVEC) or airway smooth muscle cells (ASMC)
`
`decrease ERK1/2 activation in cells, others have found it to have no effect
`
`The reported effects of DMF on MAPK signaling are less clear. While some studies have shown that DMF could
`14,16,17
`. For p38, DMF has been
`14,18
`. MAPKs can,
`
`reported to either have no effect on activation or to result in an increase in p38 phosphorylation
`
`in part, mediate their cellular effects via the activation of downstream kinases. For example, p38α activates the
`19
`downstream kinases MK2 and MK3 to promote the production of TNF . In addition, p38α can also activate the
`20
`20
`kinases MSK1 and MSK2 . These two kinases, which can also be activated by ERK1/2 , have been found to
`
`have anti-inflammatory functions in macrophages and are required for the maximal induction of IL-10 by
`21,22
`23
`. The ERK1/2 pathway can also activate RSK , however the role that this
`
`macrophages and dendritic cells
`
`activation of both MSKs and RSKs
`
`kinase plays in the regulation of cytokine production is less well established. DMF has been shown to affect the
`14,16,17
`. For instance, in keratinocytes DMF selectively blocked MSK1
`16
`phosphorylation but not ERK1/2 or p38α activation in response to IL-1 stimulation . Similarly DMF also
`
`blocked MSK1 and RSK activation in MIF (Macrophage Inhibitory Factor) stimulated keratinocytes and
`17
`24
`prevented the induction of Cox2 , a known MSK target gene . DMF has also been reported to inhibit MSK1
`
`phosphorylation in LPS stimulated dendritic cells, however in contrast to the data in keratinocytes, in dendritic
`14
`cells DMF was able to reduce LPS induced ERK1/2, although not p38 or JNK, phosphorylation .
`
`In this study we examine the mechanism by which DMF blocks cytokine induction in primary macrophages and
`
`demonstrate that it affects signaling by inhibiting the formation of M1/K63 hybrid polyubiquitin chains.
`
`Results
`
`Go to:
`
`DMF inhibits the transcription of cytokines independently of Nrf2
`
`To test the ability of DMF to block cytokine production in response to TLR agonists, BMDMs were incubated
`
`with various concentrations of DMF for 4 h (Fig. 1A). The cells were then stimulated with the TLR4 agonist LPS
`
`for a further 8 h and cytokine release determined. LPS promoted the secretion of TNF, IL-6, IL-10, IL-13 and
`
`GM-CSF; this was blocked by 50 µM DMF (Fig. 1A). To ensure this was not due to a loss of cell viability, cells
`
`were incubated with 50 µM DMF and viability determined by FACS. DMF did cause some cell death that
`
`increased over time, however the majority of cells were still alive following 12 h of DMF treatment (Fig. 1B).
`
`Cells were then treated with or without DMF in the presence of brefeldin A and monensin to block cytokine
`
`secretion. TNF levels were then measured at a single cell level by flow cytrometry following gating on the live
`
`cell population. This showed that DMF blocked LPS stimulated TNF production in the live cells (Fig. 1C). In line
`
`with the loss of cytokine secretion (Fig. 1A), DMF also repressed the induction of various cytokine mRNAs,
`
`including TNF, IL-6, IL-10, GM-CSF, IL-12p40, IL-23p19 and IFNβ, in response to 1 h of LPS stimulation (
`
`Fig. 2). DMF also suppressed the induction of IκBα mRNA, an established NFκB target gene.
`
`2 of 15
`
`10/18/2016 5:39 PM
`
`Page 2 of 15
`
`

`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`Figure 1
`
`Inhibition of LPS stimulated cytokine induction by DMF.
`
`Figure 2
`
`DMF inhibits LPS induced gene transcription.
`
`DMF has previously been suggested to act via targeting cysteine residues in Keap1 and activating the Nrf2
`25,26,27
`anti-oxidant pathway
`. To test the potential involvement of this pathway in the regulation of cytokine
`
`transcription, Nrf2−/− BMDMs were stimulated with LPS in the presence or absence of DMF and cytokine
`28,29
`. While LPS
`
`mRNA levels measured at 1 h. Nrf2 regulates the transcription of several genes including HO-1
`
`alone did not induce the mRNA for HO-1, this mRNA was induced in DMF treated wild type but not Nrf2
`
`knockout BMDMs, indicating that DMF was able to activate Nrf2 in macrophages (Fig. 3). Nrf2 knockout did not
`
`affect the induction of TNF, IL-6, IL-12p40, IL-23p19 or IκBα mRNAs in response to LPS relative to wild type
`
`cells. There was a small but significant increase (p < 0.05, Students unpaired two tailed t-test) in LPS stimulated
`
`IL-10 and GM-CSF mRNA induction in Nrf2 knockout cells compared to wild type BMDMs (Fig. 3). This
`
`increase in IL-10 mRNA induction is consistent with a recent report showing increased IL-10 production in Nrf2
`30
`knockout dendritic cells
`
`. DMF was able to inhibit cytokine and IκBα mRNA induction in response to LPS in
`
`both wild type and Nrf2−/− cells (Fig. 3). These results suggest that the effects of DMF on LPS induced cytokine
`
`induction are largely independent of any effects on the Nrf2 pathway.
`
`Figure 3
`
`Effect of DMF on LPS stimulated mRNA induction in Nrf2 knockout
`
`BMDMs.
`
`DMF has been proposed to inhibit the action of MSK1 in dendritic cells
`
`14
`. However, the reduced cytokine
`
`production caused by DMF in Fig. 1 is inconsistent with the decreased IL-10 production but increased
`22
`pro-inflammatory cytokine production previously reported in MSK1/2 double knockout macrophages
`
`,
`
`suggesting that DMF must have MSK independent effects. To confirm this, wild type and MSK1/2 knockout
`
`BMDCs were tested. As expected MSK1/2 knockout BMDCs exhibited a lower induction of the MSK1/2 target
`
`gene IL-10 relative to wild type cells (Fig. 4). Induction of IL-12p35 mRNA was increased in the MSK1/2
`
`knockout BMDCs while IL-6, IL-12p40 and IκBα mRNA induction was not greatly changed. As in macrophages,
`
`DMF blocked the LPS induced expression of both IκBα and the cytokine mRNAs tested. This inhibition was
`
`comparable in wild type and MSK1/2 knockout cells (Fig. 4).
`
`Figure 4
`
`Effect of DMF on LPS stimulated mRNA induction in MSK1/2 knockout BMDCs.
`
`DMF inhibits NFκB and ERK1/2 activation in response to TLR agonists
`
`3 of 15
`
`10/18/2016 5:39 PM
`
`Page 3 of 15
`
`

`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`TLR dependent transcription and secretion of cytokines is regulated by the NFκB as well as the ERK1/2 and p38
`31
`
`. The ability of DMF to block the induction of multiple LPS induced genes (Figs 1
`
`MAPK signaling pathways
`
`and 2) suggested that it might have a suppressive effect on both MAPK and NFκB activation in response to the
`
`TLR4 agonist LPS. TLR4 signals via both MyD88 and Trif dependent pathways, which are thought to converge
`31,32,33,34
`. Tak1 in turn activates a complex of IKKβ, IKKα and NEMO. IKKβ then
`
`on the activation of Tak1
`
`activates the classical NFκB pathway via the phosphorylation of IκBα. In addition IKKβ also phosphorylates
`
`p105, and this is required for the activation of Tpl2, which is then able to activate the ERK1/2 pathway. Tak1 also
`31,32,33,34
`.
`
`directly activates the MKKs required for the activation of the p38 and JNK pathways (Fig. 5A)
`
`Figure 5
`
`DMF inhibits multiple signals downstream of TLR4.
`
`DMF was able to inhibit the activation of the classical NFκB pathway, as 50 µM of DMF was sufficient to block
`
`the degradation of IκBα in response to 30 min stimulation with LPS (Fig. 5B). In addition, 50 µM DMF blocked
`35
`the phosphorylation of IKKβ on Ser177 and 181 (Fig. 5B), sites that correlate with its activation . In agreement
`
`with the loss of IKKβ activity, DMF also blocked the phosphorylation of the IKKβ substrate p105 and inhibited
`
`the activation of ERK1/2, as judged by phosphorylation on its TXY activation motif. DMF treatment also resulted
`
`in a partial inhibition of JNK activation, that was maximal at 75 µM DMF. In contrast, DMF did not inhibit the
`
`activation of p38 at any of the concentrations tested. In the MyD88 pathway, IRAKs are involved in propagating
`
`the signal from MyD88 to Traf6 and then Tak1. During this process IRAK1 becomes modified with K63
`
`polyubiquitin chains, resulting in the disappearance of the IRAK1 band at its predicted molecular weight on
`36,37
`. Interestingly,
`
`immunoblots that can be reversed by treatment of the lysates with deubiquitinating enzymes
`
`the loss of the IRAK1 band was partially reversed by 25 µM DMF and completely reversed by 50 µM DMF
`
`suggesting that DMF may interfere with IRAK1 ubiquitination (Fig. 5B).
`
`To examine this process in more detail, cells were incubated in 50 µM DMF and then a time course of LPS
`
`stimulation carried out (Fig. 5C). As in the previous experiment, DMF inhibited the loss of IRAK1 and IκBα,
`
`blocked the phosphorylation of IKKβ and p105 and greatly reduced the phosphorylation of ERK1/2. The effect on
`
`JNK phosphorylation was more complex. DMF reduced the phosphorylation of JNK at 30 min. In the absence of
`
`DMF, JNK phosphorylation was transient and not observed at 60 or 90 min. In contrast, JNK phosphorylation was
`
`still observed at 60 and 90 min of LPS stimulation in the presence of DMF.
`
`LPS has the ability to signal via both the MyD88 and Trif adaptors. To determine if the effects of DMF on
`
`signaling were common to pathways downstream of both adaptors, cells were stimulated with either Pam CSK
`3
`4
`or poly(I:C). Pam CSK stimulates TLR1/2 and acts via MyD88 and not Trif. The effects of DMF on Pam CSK
`4
`3
`3
`induced signaling mirrored the results seen with LPS (Fig. 5C). Poly(I:C) activates TLR3 and this utilizes Trif
`
`4
`
`and not MyD88 for downstream signaling. Consistent with this poly(I:C) did not strongly promote the
`
`ubiquitination of IRAK1. Poly(I:C) was able to induce both IKKβ and p105 phosphorylation and this was blocked
`
`by DMF. In addition DMF blocked poly(I:C) induced ERK1/2 activation, however, as for LPS and Pam CSK
`3
`4
`there was little effect on p38 activation (Fig. 5C). Poly(I:C) was only a weak activator of JNK, and this activation
`
`was slightly increased in the presence of DMF.
`
`Unexpectedly in these experiments, incubation in DMF alone induced the phosphorylation of p38 and, to a lesser
`
`extent, JNK in the absence of any other stimuli (Fig. 5C, lane 5). To examine this further a time course of DMF
`
`treatment alone was carried out. DMF treatment in the absence of any additional TLR stimulation did not affect
`
`4 of 15
`
`10/18/2016 5:39 PM
`
`Page 4 of 15
`
`

`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`IRAK1 ubiquitination or p105 phosphorylation (Fig. 6A). DMF treatment did however induce phosphorylation of
`
`MAPK signaling in cells
`
`p38, and to a lesser extent JNK, over time. Tyrosine phosphatase inhibitors have previously been found to activate
`38
`
`. Tyrosine phosphatases possess an active site cysteine and may therefore be a target
`39
`for electrophilic drugs such as DMF . Treatment with the tyrosine phosphatase inhibitor pervanadate induced the
`
`phosphorylation of p38 and JNK although, in contrast to DMF, pervanadate also strongly induced ERK1/2
`
`phosphorylation (Fig. 6A). As would be expected pervanadate strongly induced global tyrosine phosphorylation
`
`levels. In contrast, DMF had little effect on global tyrosine phosphorylation levels indicating that it does not act as
`
`a general tyrosine phosphatase inhibitor in cells (Fig. 6B). MAPKs can be dephosphorylated by DUSPs (Dual
`
`Specificity Phosphatases) and inhibition of these enzymes by DMF may account for the p38 and JNK activation.
`
`Inhibition of the DUSP would be expected to be specific for the MAPK and not affect the upstream MKK. DMF
`
`was found to increase the phosphorylation of MKK3 and 6 suggesting that its effects on p38 and JNK
`
`phosphorylation were independent of DUSP inhibition (Fig. 6A).
`
`Figure 6
`
`DMF can activate p38 MAPK and can covalently modify its targets in
`
`the cell.
`
`DMF has the ability to covalently modify cysteine residues in its targets, as is proposed for Keap1. We assessed
`
`whether DMF inhibition remained following washout of DMF from the cells. BMDMs were treated for 4 h with
`
`50 µM DMF and then either directly stimulated with LPS or washed extensively and incubated in media without
`
`DMF for various times. This showed that signaling was still inhibited by DMF 4 h after DMF being washed off
`
`the cells, but that inhibition was lost by 16 h (Fig. 6C). This slow loss of inhibition would be consistent with DMF
`
`covalently modifying its target in cells.
`
`In vivo DMF can be converted to MMF. We therefore checked if MMF was able to mimic the effects of DMF on
`
`cell signaling. Unlike DMF, MMF did not inhibit LPS induced ERK1/2 or p105 phosphorylation, or the
`
`ubiquitination of IRAK1 (Supplementary Figure 1).
`
`DMF inhibits E2 conjugating enzymes in vitro
`
`Together these results suggest that DMF may act at an upstream point in the TLR signaling cascade. The
`
`activation of Tak1 and the recruitment of the IKK complex involves the formation of K63/M1 hybrid
`
`polyubiquitin chains which can be added to several proteins in the MyD88 signaling cascade including MyD88
`34,40
`. The formation of polyubiquitin chains requires a cascade of three enzymes; an E1 activating
`
`and IRAKs
`
`enzyme, an E2 conjugating enzyme and an E3 ligase to mediate the transfer of the ubiquitin to the substrate. E2
`
`enzymes use a cysteine in their active site to covalently couple to ubiquitin. DMF is an electrophilic compound
`
`and therefore has the potential to react with the SH group in cysteines (Fig. 7A). We therefore investigated if
`
`M1 chains
`
`DMF could inhibit Ubc13 or UbcH7, the major E2 enzymes proposed to be involved in the formation of K63 and
`41,42,43,44,45
`. In vitro, addition of the E1 UBE1, ubiquitin and Mg-ATP to an E2 is able to promote the
`45
`.
`
`loading of ubiquitin onto the E2, which can be resolved as a mobility shift on SDS polyacrylamide gels
`
`Addition of DMF to this reaction inhibited the loading of ubiquitin onto Ubc13 with an IC50 of between 10 and
`
`20 µM (Fig. 7B). In parallel assays, UbcH7 was more strongly inhibited by DMF with an IC50 of less than 10 µM
`
`and complete inhibition by 50 µM (Fig. 7B). If DMF were to inhibit Ubc13 or UbcH7 via covalent modification
`
`of a cysteine, the molecular mass of the E2 should be increased following DMF treatment. MALDI-TOF mass
`
`spectrometry was therefore used to determine the molecular mass of recombinant Ubc13 and UbcH7 before and
`
`after treatment with DMF. The recombinant Ubc13 has a theoretical mono-isotopic mass of 17295 Da, which
`
`corresponded well with the observed mass of 17301 Da. For Ubc13, an increase in the molecular mass of 144 Da
`
`was seen following incubation in DMF, which would be consistent with the labeling of a cysteine residue in the
`
`5 of 15
`
`10/18/2016 5:39 PM
`
`Page 5 of 15
`
`

`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`protein with DMF via a Michael addition reaction (Fig. 7A). For unmodified UbcH7 a mass of 18157 Da was
`
`observed relative to a theoretical mono-isotopic mass of 18149 Da. Following DMF treatment of UbcH7 3 peaks
`
`were obtained. The major peaks indicated an addition of 144 and 289 Da, which would correspond to addition of
`
`1 or 2 DMF molecules respectively. A minor peak was also obtained with a mass increase of 432 Da, which
`
`would suggest addition of 3 molecules of DMF (Fig. 7C). Of note, Ubc13 only contains the active site cysteine
`
`whereas UbcH7 also contains 2 further cysteines in addition to its reactive site residue. To further test the ability
`
`of 50 or 100 µM DMF to block E2 activity, it was tested against a panel of 24 different E2 enzymes (Fig. 7D).
`
`This showed that while DMF could inhibit all the E2’s tested to some extent, it did exhibit some selectivity with
`
`UbcH7, Ube2W, Ube2T, Ube2S and Ube2R1 being most strongly inhibited.
`
`Figure 7
`
`DMF can inhibit E2 conjugating enzymes in vitro.
`
`DMF blocks the formation of M1 and M1/K63 polyubuititin chains downstream of TLR signaling
`
`Previous studies have indicated that TLR agonists can stimulate the formation of K63/M1 hybrid polyubiquitin
`40,46
`. As, in vitro, DMF inhibited Ubc13 and UbcH7, the E2 conjugating enzymes proposed to be involved
`
`chains
`
`in the formation K63 and M1 chains respectively, its ability to block formation of M1 and K63 chains in cells was
`
`determined. M1-Ub chains were pulled down from RAW264.7 macrophage cell lysates using Halo-tagged NEMO
`40
`. This approach would be expected to pull
`
`beads, which shows >50 fold selectivity for M1 relative to K63 chains
`
`down M1-Ub chains, complexes containing both M1 and K63 chains as well as M1/K63 hybrid chains that are
`40,46
`. As seen in primary macrophages following TLR
`
`proposed to be formed downstream of TLR signaling
`
`stimulation, DMF inhibited the ubiquitination of IRAK1 and the phosphorylation of ERK1/2, p105 and IKK in
`
`RAW264.7 cells (Fig. 8A). Halo-NEMO pulldowns were blotted for M1 or K63 polyubiquitin chains on IRAK1
`
`and IRAK2. LPS increased the amount of M1 and M1/K63 hybrid chains in the pulldowns, and this increase was
`
`inhibited by pretreatment of the cells with DMF (Fig. 8B). Polyubiquitinated IRAK1 and IRAK2 could also be
`
`seen in the pulldowns from LPS stimulated but not untreated cells, and this was blocked by pretreatment with
`
`DMF (Fig. 8B). DMF did not however block all ubiquitination within the cell. Halo-TUBE (Tandem Ubiquitin
`
`Binding Entities) beads were used to pull down all ubiquitin chains within the cells. Blotting of these pulldowns
`
`showed that LPS did not increase the total amount of K11, K48 or K63 linked polyubiquitin chains in these pull
`
`downs and that these levels were also not reduced by DMF (Supplementary Figure 2). Similarly using
`
`Halo-TAB2-NFZ beads to pull down K63 chains showed that neither LPS or DMF affected the amount of K63
`
`chains in these pull downs (Supplementary Figure 2). This probably reflects a high background level of K63 and
`
`K48 chains within the cell so that the changes induced by TLR signaling are too small to be seen. In line with
`
`this, blotting the Halo-TUBE or Halo-TAB2-NFZ pull downs for IRAK2 showed that LPS did induce the
`
`ubiquitination of this protein and that this was blocked by DMF (supplementary Figure 2). In line with the results
`
`in RAW264.7 cells, DMF also greatly reduced the amount of M1 chains that could be pulled down from BMDM
`
`lysates using either Halo-NEMO or Halo-TUBE beads (Fig. 9). Similarly the amount of ubiquitinated IRAK2 that
`
`came down in the Halo-NEMO pulldowns was also reduced by DMF. In contrast, DMF did not reduce the amount
`
`of K48 linked chains in Halo-TUBE pulldowns from BMDMs (Fig. 9).
`
`Figure 8
`
`Effect of DMF on LPS stimulated polyubiquitin chain formation in RAW264.7
`
`macrophages. RAW264.7 cells were incubated in the presence or absence of 50 µM DMF
`
`for 4 h.
`
`6 of 15
`
`10/18/2016 5:39 PM
`
`Page 6 of 15
`
`

`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`Figure 9
`
`Effect of DMF on LPS stimulated polyubiquitin chain formation in
`
`BMDMs.
`
`M1 chains are formed by the LUBAC complex which contains HOIP, HOIL and Sharpin. In this complex HOIP
`47
`acts as the E3 and directly conjugates to ubiquitin via a cysteine during ubiquitin transfer
`. As it is possible that
`
`DMF might inhibit HOIP via reacting with this cysteine in cells, HOIP was immunoprecipitated from cells and its
`
`activity examined in an in vitro ubiquitination reaction. Due to the covalent modification of the cysteine by DMF,
`
`inhibition in the cell should be retained during the immunoprecipitation and in vitro assay. Consistent with
`40
`, HOIP was constitutively active in this assay, and this activity was not affected by pretreatment
`
`previous reports
`
`of the cells with DMF (Supplementary Figure 3).
`
`DMF inhibits K63/M1 chain formation in response to IL-1 and TNF
`
`The formation of K63 and M1 chains is not restricted to TLR signaling and also occurs downstream of other
`34,40
`. To examine chain formation downstream of IL-1, HEK293 cells stably
`
`stimuli including IL-1 and TNF
`
`expressing the IL-1 receptor were stimulated with IL-1 in the presence or absence of DMF. In these cells IL-1
`
`stimulated the phosphorylation of p105 and IKK as well as the degradation of IκBα which was blocked by DMF,
`
`as was the activation of ERK1/2 (Fig. 10A). In Halo-NEMO pulldowns, IL-1 stimulation increased the amount of
`
`both M1 and K63 chains observed, and this was blocked by pretreatment of the cells with DMF (Fig. 10B).
`
`Treatment of HeLa cells with TNF induced ERK1/2, IKK and p105 phosphorylation as well as the degradation of
`
`IκBα. ERK1/2 phosphorylation and IκBα degradation were inhibited by DMF, however in these cells DMF did
`
`not block p105 or IKK phosphorylation in response to TNF (Fig. 10C). NEMO pulldowns showed that TNF
`
`stimulated the formation of M1 polyubiquitin chains, and this was reduced by DMF. TNF only resulted in a minor
`
`increase in the signal for K63 chains, but again this was reduced by DMF (Fig. 10D).
`
`Figure 10
`
`Effect of DMF on IL-1 and TNF stimulated M1 chain formation.
`
`Discussion
`
`Go to:
`
`While the ability of DMF to modulate immune function is well established, its mechanism of action has been
`
`more elusive, and several potential in vivo targets have been suggested. While DMF can activate the Nrf2
`48
`
`anti-oxidant pathway, our results, along with those recently published by Gillard et al.
`
`, indicate that this is not
`
`required for the ability of DMF to block cytokine induction. In our hands DMF has a suppressive effect on several
`
`pathways downstream of TLR signaling, including NFκB and ERK1/2 signaling. We demostrate here that DMF
`
`can inhibit cytokine production in macrophages and that this correlates to the inhibition of NFκB and MAPK
`
`signaling pathways. This would suggest that DMF acts at an upstream point in TLR signaling, and in line with
`
`this we show that DMF can block the formation of signaling complexes containing M1 and K63 polyubiquitin
`
`chains downstream of TLR and IL-1 signaling. Furthermore, in vitro DMF can inhibit UbcH7 and Ubc13, the
`
`major E2 enzymes thought to be required for K63 and M1 chain formation in response to IL-1 or TLR agonists.
`
`DMF did not result in a global loss of polyubiquitin chains. Notably the levels of K63 and K48 chains in pull
`
`downs with Halo-TUBE beads was unaffected by DMF. One explanation is that the majority of these chains were
`
`preformed in the cell prior to the addition of DMF. If the turnover of the chains was slow, no effect of DMF
`
`7 of 15
`
`10/18/2016 5:39 PM
`
`Page 7 of 15
`
`

`
`Dimethyl fumarate blocks pro-inflammatory cytokine production via inhi...
`
`https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976367/
`
`would be expected on the levels of these chains. In line with this, LPS did not increase the levels of either K48 or
`
`K63 chains in the Halo-TUBE pulldowns indicating the presence of a significant number of K48 and K63 cells in
`
`resting cells. These observations does not mean TLR signaling cannot promote the K48 or K63 linked
`
`ubiquitination of specific proteins only that any such modifications do not measurably alter the total amount of
`
`K48 and K63 chains in the cell. TLR signaling has been shown to promote K63 and M1 polyubiquitin chain
`
`formation on a number of signaling proteins including IRAK1. Previous reports have indicated that many of these
`40,46
`, and in line with this we see increased amounts of K63
`
`chains actually from as hybrids of K63 and M1 chains
`
`and M1 chains in pull downs with Halo-NEMO beads, which selectively pulls down complexes containing M1
`
`chains. The TLR induced formation of these K63 and M1 chains was blocked by pretreatment with DMF (Figs 8,
`
`9, 10).
`
`DMF is converted to MMF in vivo. The effects we observed on signaling were specific for DMF, as we did not
`
`observe any effect of MMF on LPS induced signaling. This is in agreement with a previous study showing that
`48
`. In macrophages we
`
`DMF, but not MMF, repressed TLR induced cytokine induction in primary human PBMCs
`
`show that DMF blocked the activation of the IKK complex and thus prevented both p105 phosphorylation and
`
`IκBα degradation. Previous reports have also indicated that DMF may inhibit NFκB transcriptional activity. For
`
`example, DMF reduced NFκB activity in LPS stimulated dendritic cells and TNF stimulated HUVEC or
`13,14,15
`
`. A common feature of these earlier studies was that DMF inhibited NFκB DNA binding or
`
`ASMC
`
`relocalization to the nucleus, but did not have a major effect on the degradation of IκBα. The reason for the
`
`difference between these studies and our findings is not clear. One potential explanation is the different cell types
`
`and stimuli used in the various studies may influence the results seen for IκBα degradation. In line with this, the
`
`ability of DMF to inhibit IκBα degradation in TNF stimulated HeLa cells was much less than in TLR stimulated
`
`macrophages. Another issue could relate to how DMF was used. The concentrations of DMF were similar in the
`
`different studies, however the pre-incubation times in these previous papers were not clear and were likely
`
`different to the 4 h preincubation used in our studies. As DMF can act as a covalent inhibitor it is able to display
`
`time dependent kinetics of inhibition (data not shown) and thus longer pre-incubation times might give a more
`
`pronounced inhibition of IκBα degradation.
`
`The effects we observed on NFκB and MAPK activation could be explained by inhibition of TLR signaling at the
`
`level of K63/M1 polyubiquitin chain formation and thus blocking the activation of Tak1. If DMF did act at the
`
`level of Tak1 or ubiquitination, it would also be expected to inhibit p38 activation. This is complicated by the
`
`observation that p38 was activated by DMF alone (Fig. 6), which would obscure any effect of DMF on TLR
`
`induced p38 activation pathways. One potential mechanism would be the inhibition of tyrosine phosphatases, as
`
`this is known to activate MAPK signaling. This is however unlikely as DMF did not result in increased global
`
`levels of phosphotyrosine. Another explanation could be activation of a MAP3K for p38 and JNK. One possible
`
`candidate might be Ask1, which is also known to be expressed in macrophages and can contribute to p38
`49
`50
`activation by LPS . Ask1 is activated by oxidative stress via a complex mechanism involving Trx1 . It is
`
`possible that DMF may activate Ask1 via modification of cysteine residues on Trx1 or via targeting proteins
`
`involved in the response to oxidative stress. Further work would be required to resolve this point.
`
`The effects of DMF on signaling have recently been suggested to have similarities to those produced by BAY
`48
`11-7082 . Interestingly while BAY 11-7082 was originally described as an IKK inhibitor a recent study has
`45
`shown that BAY 11-7082 can act in cells via inhibiting E2 enzymes, including Ubc13 and UbcH7 . As for DMF,
`
`the mechanism found for this involved the covalent modification of the active site cysteine in the E2.
`
`Taken together our results, along with those of others, would indicate that DMF has complex effects in cells and
`
`is unlikely to act via a single target. Instead it is likely that the overall therapeutic effects of DMF in MS or
`
`psoriasis are due to the ability of DMF to directly modify the function of a number of cellular proteins via the
`
`modification of reactive cysteines in th