Since the discovery of FAT10 in 1996 as a new member of the ubiquitin‐like family many attempts have been done to unravel the secrets of this molecule regarding its structure, its function and its physiological relevance. One open question is still why does a diubiquitin exist, which modifies substrates and targets them to degradation by the proteasome as ubiquitin does? The creation of a FAT10 knockout mouse could not answer this question as it displays – at least in young age ‐ only minimal phenotypic differences (Canaan, Yu et al. 2006), in contrast to a knockout of the UbC polyubiquitin gene which is lethal (Ryu, Maehr et al. 2007). The fact that FAT10 uses parts of the ubiquitin conjugation machinery, both sharing the E1 enzyme UBA6 and the E2 enzyme USE1 (Chiu, Sun et al. 2007, Aichem, Pelzer et al. 2010), makes it even harder to distinguish pathways and define a specific role for FAT10. However, it’s a particular feature in mammals and is expressed only in cells of the immune tissue and it is inducible in all other body cells upon cytokine secretion during an infection (Raasi, Schmidtke et al. 1999, Lukasiak, Schiller et al. 2008). This restricted expression might be the key to understand the role of FAT10 as a temporally limited mechanism to regulate cellular processes during inflammation or more generally within the immune system.
3.1 FAT10 is conjugated to VCP
It is known for some years now that there must be a lot of FAT10 substrates (Raasi, Schmidtke et al. 2001), but only few could be identified yet (Aichem, Pelzer et al. 2010, Li, Santockyte et al. 2011, Aichem, Kalveram et al. 2012, Bialas, Groettrup et al. 2015). Hence there’s still a need to find new substrates and interaction partners to approach to the physiological function of FAT10. Aichem et al. discovered by a mass spectrometry based screen 571 putative substrates and interaction partners which could be mapped only partially into protein families or distinct processes (Aichem, Kalveram et al. 2012). One interesting hit was chosen as study object for this work. The AAA protein Valosin‐containing protein (VCP) was classified as a putative FAT10‐substrate. The aim of this work was therefore to verify this FAT10ylation and to
substrates, it was interesting to investigate whether there are overlapping or oppositional
functions of FAT10 and ubiquitin in context of VCP associated processes.
As the screen was done in HEK293 cells, because FAT10 can be induced by IFN and TNF very well, we stayed in the same cell line for our investigations and performed first experiments by overexpression of HA‐FAT10 and myc‐VCP. It was observed for other FAT10 substrates such as the FAT10 E2 conjugating enzyme USE1, the autophagosomal receptor p62 or the ubiquitin activating enzyme UBE1 that only a little portion of the existing protein is modified by FAT10 whereas the major portion of the protein remains unmodified (Aichem, Pelzer et al. 2010, Aichem, Kalveram et al. 2012, Bialas, Groettrup et al. 2015). The same was found in case of VCP as only a faint band representing a myc‐VCP‐HA‐FAT10 conjugate occurred after HA‐IP which accumulates slightly under proteasome inhibition (Figure 6 A). More prominent was a non‐covalent interaction shown by a co‐IP of unconjugated myc‐VCP with HA‐FAT10. A non‐
covalent interaction could still be detected with a FAT10 mutant with a mutation of the C‐
terminal GG to AV. The non‐covalent interaction was less pronounced with the mutant than with the wildtype which can be attributed to a decreased expression of the mutant. As expected the mutant couldn’t be conjugated to VCP as also upon MG132 treatment no conjugate could be observed. The conjugates were resistant to treatment with ‐
mercaptoethanol after IP which indicates the formation of an isopeptide linkage between the two proteins. To determine the fate of the conjugate a cycloheximide chase was performed (Figure 6 B). The conjugate was degraded at the same speed as monomeric FAT10 and rescued to the same extend under proteasome inhibition. The non‐covalent interacting VCP decreased, too, but not to the same degree as the VCP‐FAT10 conjugate. This hints to accelerated degradation of the conjugate rather than to limited conjugation due to the reduced FAT10 amount.
In an in vitro assay VCP could be FAT10ylated in an ATP dependent manner, but the modification was only minimal (Figure 7). This could be due to the use of just the E1 enzyme UBA6. Addition of the E2 enzyme USE1 did not increase conjugate formation (data not shown), but it can’t be ruled out that presence of a complete enzymatic cascade of E1, E2 and E3 which is expected to be present in cells would increase in vitro FAT10ylation at least slightly. The assumption that minor modification and subsequent degradation is a mechanism for fine‐
tuning the function of the substrate does not hold true for the in vitro situation as no
additional regulatory proteins are present. This would implicate a restriction signal for modification in the structure of the substrate which seems to be quite unlikely. And considering that VCP is very abundant in cells, such a small part of FAT10ylated protein directed to degradation would hardly influence the overall amount of VCP. Thus the conjugation of FAT10 does not appear to be a tool to regulate or even terminate VCP function by its degradation.
In a mass spectrometry based screen for modification of proteins with the ULM ISG15, VCP was found as a substrate (Giannakopoulos, Luo et al. 2005). Similar to FAT10ylation, ISGylation applies only to a minor fraction of VCP. As the authors didn’t follow this discovery in further investigations, the function or consequences of this modification is not known. Thus, the modification of VCP by an ULM is not a unique feature of FAT10, but as no other ULM was published to modify VCP this is a mutuality of the two modifiers comprised of two ubiquitin‐
like domains.
3.2 FAT10 interacts with VCP non‐covalently
The focus of further investigations regarding the interplay between FAT10 and VCP was directed to the non‐covalent interaction observed in overexpression (Figure 6) and in vitro experiments (Figure 7), which could be confirmed under endogenous conditions. After induction of FAT10 expression with IFN and TNF in HEK293 cells VCP could be co‐
immunoprecipitated with FAT10 which was even enhanced after MG132 treatment because of FAT10 accumulation (Figure 8 A). Upon knockdown of FAT10 by specific siRNA the interaction was less due to reduced FAT10 levels (Figure 8 B). Up‐regulation of FAT10 induces apoptosis (Raasi, Schmidtke et al. 2001). Therefore the lower FAT10 expression level might be explained in a way that only those cells survived the pressure of transfection, which expressed lower amounts of FAT10. This was also already discussed in case of USE1 siRNA treatment where the same effect was observed (Aichem, Pelzer et al. 2010). Taken together, with these experiments it could be proved that the non‐covalent interaction of FAT10 and VCP is a physiological phenomenon under inflammatory conditions including cytokine secretion
Several groups could show in yeast and in vitro that VCP can interact non‐covalently with polyubiquitin‐chains attached to substrates or with ubiquitin‐fusion proteins (Dai and Li 2001, Rape, Hoppe et al. 2001, Ye, Meyer et al. 2003). In vitro studies revealed that VCP prefers tetra‐ubiquitin for binding, diubiquitin had only about 5% of the binding efficiency of tetra‐
ubiquitin and for monoubiquitin no binding was observed (Dai and Li 2001). This is in accordance with the observed binding of FAT10 to VCP in cells, as FAT10 resembles the structure of diubiquitin and interacts with a rather small percentage of endogenous VCP.
3.3 FAT10 directly interacts with VCP
Although most publications claim that binding of ubiquitin to VCP is mediated by diverse cofactors (reviewed in (Schuberth and Buchberger 2008, Yeung, Kloppsteck et al. 2008)), it was described that polyubiquitin chains can directly bind to the N‐domain of VCP (Dai and Li 2001, Rape, Hoppe et al. 2001, Ye, Meyer et al. 2003). Therefore it was obvious to test whether FAT10 is able to bind VCP directly. As first in vitro experiments with recombinant proteins showed clearly an interaction (data not shown) we wanted to map the binding site of FAT10 on VCP. Recombinant His‐tagged VCP and truncation mutants lacking at least one domain were used for pull down experiments with GST‐FAT10 (Figure 9). As mentioned above FAT10 interacted with full length VCP and background binding to GST could be ruled out so this can be considered as a true direct interaction. A very strong interaction with the N‐domain was observed and strikingly all other mutants could be pulled down with GST‐FAT10, too. This means that FAT10 has besides the N‐domain at least one additional binding site on VCP.
Because some cofactors bind directly to the C‐terminus of VCP (Rumpf and Jentsch 2006,
Madsen, Andersen et al. 2008, Bohm, Lamberti et al. 2011) it was most likely to be the same for FAT10. But with the C, NC and ND1 mutants still binding could be observed. The reduced signal of the N and NC mutants on the western blot must be due to reduced recognition by the anti‐His antibody as the coomassie stained gel showing protein inputs displayed equal protein amounts.
This direct interaction is an interesting finding as it was believed that connecting VCP to ubiquitinated substrates for extraction is generally facilitated by cofactors (reviewed in (Stolz, Hilt et al. 2011, Meyer, Bug et al. 2012)) assuming that VCP might be involved in the
degradation process of FAT10ylated substrates. Moreover it was shown for other ULMs that they can interact with VCP cofactors, e.g. UBXD7 binds to NEDD8 conjugated to cullin RING ligases (CRLs) (Bandau, Knebel et al. 2012, den Besten, Verma et al. 2012), p47 associates with ATG8 during autophagy (Krick, Bremer et al. 2010) and Ufd1 interacts with SUMO for maintaining genome stability (Nie, Aslanian et al. 2012) but a non‐covalent direct interaction with VCP couldn’t be shown, except for ISG15 (Giannakopoulos, Luo et al. 2005). In this publication the authors speculated, that the above mentioned slight ISGylation (see also 3.1) might occur only at one subunit of the VCP hexamer. Immunoprecipitation of the conjugated ISG15 would then co‐immunoprecipitate five unmodified subunits pretending a non‐covalent interaction after hexamer disassembly by SDS‐PAGE. For the FAT10‐VCP interaction this could be excluded as the in vitro studies for the direct interaction were performed without E1 and E2 enzymes (Figure 9), thus the interaction must be non‐covalent.
Considering that polyubiquitin can directly bind to the N‐domain of VCP, a possible competition of FAT10 and ubiquitin‐chains could be tested in further experiments. In vitro experiments exploring whether FAT10 binding to VCP excludes other cofactors didn’t give coherent results. Initial competition experiments mixing recombinant FAT10, increasing amounts of recombinant His‐p47 and VCP for subsequent VCP‐IP resulted in massive unspecific binding of FAT10. GST‐pulldowns using GST‐FAT10 showed inconsistent binding of recombinant VCP not permitting a conclusion regarding a competition of FAT10 and p47 (data not shown). A clear binding of recombinant VCP to recombinant His‐Ufd1 and GST‐Npl4 couldn’t be observed due to lack of purity of the cofactors (data not shown). Because of time limitations the experiments were not repeated.
Remarkably Liu et al. proposed a VBM motif between the UBA1 and UBA2 domain in NUB1L for a direct binding to VCP (Liu, Fu et al. 2013). The authors suggested that NUB1L promotes the transfer of NEDD8 from Npl4 in the VCP‐Ufd1‐Npl4 complex to the proteasome for degradation (Liu, Yang et al. 2013). And as NUB1L interacts with FAT10 (Hipp, Raasi et al. 2004) we wanted to know which of the ubiquitin‐like domains of FAT10 bind to VCP.
Immunoprecipitation of the single HA‐tagged FAT10 N‐or C‐terminal ubiquitin‐like domain revealed that both domains bind to VCP with a slight gain of the C‐terminal domain. This is quite interesting as FAT10 binds with its N‐terminal domain to the UBA domains of NUB1L
binding. In contrast to Liu et al. (Liu, Yang et al. 2013) our group observed that NUB1L binds to VCP via its UBL domain (G. Schmidtke, unpublished data) which would allow a trimeric complex of VCP, NUB1L and FAT10 provided that FAT10 and NUB1L don’t bind to the same site on VCP. Experiments investigating this issue were not conclusive. GST‐pulldowns with recombinant GST‐FAT10, increasing amounts of recombinant NUB1L and constant amounts of recombinant VCP implicated first a competition of NUB1L and VCP for FAT10 binding.
Following experiments displayed simultaneous binding of the three proteins to each other, implicating a trimeric complex. GST‐pulldowns using GST‐NUB1L full length as well as mutants lacking either the UBA domains or the UBL domain resulted in unspecific binding of recombinant 3xFlag‐FAT10 (data not shown).
3.4 VCP function is not necessary for degradation of FAT10 conjugates
The most obvious function to test the impact of the VCP‐FAT10 interaction was the turnover of FAT10 conjugates. Proposing that degradation of most of the FAT10 substrates would rely on VCP function (Liu, Yang et al. 2013), as it is the case for many NEDD8 and ubiquitin substrates, inhibition of VCP should lead to an accumulation of FAT10 conjugates. First steady state levels of FAT10 and ubiquitin conjugates upon VCP inhibition by the specific VCP inhibitors EerI and DBeQ were compared to levels upon proteasome inhibition by MG132 (Figure 11). As expected ubiquitin conjugates accumulated due to EerI treatment to a similar extend as upon MG132 treatment whereas FAT10 conjugates only accumulated upon proteasome inhibition. DBeQ had no influence on ubiquitin conjugate levels, but it slightly decreased FAT10 conjugate as well as monomeric FAT10 amounts. To study this in more detail cycloheximide chases were performed to determine degradation rates of FAT10 conjugates.
Inhibition of VCP by EerI didn’t change speed or efficiency of degradation (Figure 12). The seemingly enhanced degradation caused by DBeQ treatment led to the same overall degradation of 40 % of the conjugates as in all other conditions (Figure 13). The difference between untreated and DBeQ‐treated cells can’t be reasoned by limited efficacy of cycloheximide in the untreated cells because degradation of monomeric FAT10 was observed already in the lysates. The reason why this is not reflected in the IP, neither by the amount of conjugates nor by the amount of monomeric FAT10, can’t be explained. Furthermore, first
investigation of the degradation pathways of FAT10 revealed that FAT10 is not specifically targeted to autophagosomes (Aichem, Kalveram et al. 2012) and second it was published by Chou et al. that DBeQ impairs autophagy (Chou, Brown et al. 2011). Thus no additional degradation pathway for FAT10 conjugates is established upon VCP inhibition which would explain the difference in the degradation rate. Another quite theoretical explanation might be that degradation of many ubiquitinated substrates depends on unfolding by VCP, but upon VCP inhibition the substrates accumulate in a folded state and can’t be degraded by the proteasome. Whereas FAT10 conjugates which are not dependent on VCP might then be preferentially degraded by the proteasome. This scenario becomes pretty unlikely as additional experiments with overexpression of the ATPase‐deficient VCP mutant E578Q (Ye, Meyer et al. 2003) were performed which should have the same effect but didn’t lead neither to accumulation nor to increased degradation of FAT10 conjugates (Figure 14). And in this case any other off‐target effect caused by the inhibitors could be excluded. The reduced amount of FAT10 conjugates as well as monomeric FAT10 might also be due to activation of the UPR by DBeQ treatment which restricts translation by inhibition of eIF2 (Walter and Ron 2011).
This was represented by the reduced FAT10 amount at steady state before cycloheximide treatment observed in the lysates.
Although a dependence of single FAT10 substrates on VCP can’t be ruled out, these results illustrate a difference to ubiquitin as the degradation of the bulk of FAT10 conjugates does not rely on VCP function.
3.5 FAT10 does not disturb the hexamerization of VCP
As VCP didn’t seem to influence FAT10 function – at least not on its function as a modifier leading to degradation of its substrates‐ one might speculate that it could be the other way round. Considering the direct interaction it might be possible that FAT10 acts like a cofactor or other functional regulator of VCP. It was first described in Arabidopsis that the plant UBX domain‐containing protein (PUX) 1 regulates Cdc48 by disassembly of the hexamer (Rancour, Park et al. 2004), which could be shown for mammalian UBXD9 (TUG, ASPL), too (Orme and
subsequently an analytical gel filtration was performed. If FAT10 would destroy the VCP hexamer peaks at higher elution volumes corresponding to the molecule size of single VCP monomers, dimers or trimers should be visible in the chromatogram. However, no change in the elution volume of the peak corresponding to a hexamer was observed, meaning that FAT10 didn’t interfere with the oligomerization state of VCP under these experimental conditions. This matches with the finding of Bogan et al., who showed that although the N‐
terminal UBL domain of UBXD9 becomes cleaved which exposes a C‐terminal diglycine motif resembling a ULM, that for the disassembly of the hexamer the C‐terminal part of UBXD9 containing the UBX domain was responsible (Bogan, Rubin et al. 2012). Therefore it appears to be plausible that FAT10 does not influence the hexameric structure of VCP.
3.6 FAT10 does not influence the ATPase activity of VCP
Another way of regulating VCP is the modulation of its ATPase activity. It was shown before that the core cofactor p47 decreases the ATPase activity by 80% (Meyer, Kondo et al. 1998).
Recently it was published that the cognate cofactor p37 increases the VCP activity and that deletion of a region in p47 makes it an activating factor (Zhang, Gui et al. 2015). Assuming that FAT10 may act as a regulating factor of VCP a colorimetric ATPase assay was used to investigate the influence of FAT10 on VCP activity. The first experiments displayed a high increase in ATPase activity upon 3xFlag‐FAT10 addition (Figure 17). Due to high variances in duplicate samples, an ATPase assay using 32P‐labelled ATP was performed instead which gave more homogenous results and validated the previous results. As none of the ubiquitin‐like modifiers was reported to interact directly with VCP it seemed to be suitable to use ISG15 and SUMO besides linear diubiquitin, which resembles the structure of FAT10, as negative controls. Additionally a tagless FAT10 was used to rule out any influence of the 3xFlag‐tag.
However, as all control proteins (Figure 19) and finally newly purified 3xFlag‐FAT10 (Figure 20) didn’t cause an increase in VCP activity, one has to conclude that the first 3xFlag‐FAT10 preparation was contaminated with a bacterial ATPase giving a false positive result.
Since the rise in activity for increasing amounts of FAT10 alone from the first 3xFlag‐FAT10 preparation was not the same as for 3xFlag‐FAT10 with VCP it could be speculated that
theoretically there could have been another factor in the preparation – maybe just co‐purified ATP ‐ which activated VCP.
Accordingly, contrary to first promising results one has to conclude that pure recombinant FAT10 is not able to modulate the ATPase activity of VCP in vitro.
3.7 FAT10 might influence the degradation of 1AT under ER stress
One of the best studied functions of VCP is its role in Endoplasmic Reticulum associated
degradation (ERAD). Thus it was investigated whether FAT10 influences the degradation of ERAD model substrates such as a mutant of 1AT. This Z mutant does not fold correctly, stucks in the ER as it can’t be secreted as the wildtype protein and therefore induces ER stress (Lomas, Evans et al. 1992). Additionally secretion of 1AT WT was blocked by treatment with Tunicamycin which inhibits N‐glycosylation and hence the protein is restrained in the ER.
Surprisingly FAT10 did not seem to have an effect on the degradation of the 1AT Z mutant, whereas degradation of 1AT WT was enhanced (Figure 21). As ERAD results in degradation of the substrates by the proteasome one would expect a noticeable accumulation of proteins upon proteasome inhibition. However, the accumulation was rather minor which raises the question of validity regarding the observed degradation while a failure of MG132 can’t be
Surprisingly FAT10 did not seem to have an effect on the degradation of the 1AT Z mutant, whereas degradation of 1AT WT was enhanced (Figure 21). As ERAD results in degradation of the substrates by the proteasome one would expect a noticeable accumulation of proteins upon proteasome inhibition. However, the accumulation was rather minor which raises the question of validity regarding the observed degradation while a failure of MG132 can’t be