Which LRRK2 kinase substrates identified to date are most likely to be true physiological substrates? What limitations/barriers hinder identification of novel substrates?

Despite the clear and strong genetic links between LRRK2 and PD, we still know little about its actual physiological function in the cell. Evidence demonstrating that pathogenic mutations in LRRK2 result in enhanced kinase activity points to a role for possible downstream substrate proteins (appropriate or inappropriate) in the pathway leading to cellular toxicity. With such a clear and critical question, researchers are racing like mad to find these substrates and a number have already been reported to date, such as moesin (Jaleel et al., 2007), 4E-BP (Imai et al., 2008) and more recently MKK3/6 and MKK4/7 (Gloeckner et al., 2009). And of course LRRK2 itself (Greggio et al., 2008).

However, when listening to the whispers in the corners and back rows of meeting and lecture halls, one is left with the impression that many of the identified substrates published to date do not necessarily replicate in all laboratory hands. And progress in identifying these substrates also seems slow given all the potential players involved. What are the issues? Is it lack of the right tools (yes…we know about the antibody problem…), or is it something more fundamental in how we are going about tackling the problem?

Also, although this Research Question focuses on the downstream substrates, it’s worth thinking upstream as well at proteins that may target and modulate LRRK2 itself. A couple of recent publications point to the ubiquitin ligase CHIP as important for targeting LRRK2 to the proteasome (Ding and Goldberg, 2009; Ko et al., 2009). What seems nice here is the replication of findings and this might be a good place for more collaboration among groups. (This is something the MJFF has made a priority in a recent LRRK2 Biology Challenge program currently under review.)

A recent article by Qing et al proposing that synuclein is a direct substrate of LRRK2 illustrates many of the reasons why progress in identifying substrates can be difficult - for every exciting new report there are many ways in which this needs to be validated.

As discussed elsewhere there is very likely to be some link between LRRK2 and synuclein, although the nature of that link is not immediately obvious.  Synuclein is known to be phosphorylated in Lewy bodies and some studies suggest that phosphorylation at Serine 129 may be the key modification in depositied synuclein protein.  Therefore, synuclein phosphorylated at Serine 129 (pS129-Syn) is a really good, logical candidate for a LRRK2 substrate. 

In this context, the experiments reported by Qing et al showing an increase in immunoreactivity for pS129-Syn after incubation of recombinant protein with LRRK2 isolated by pulldown from HEK cells is extremely exciting.  If correct, then there is a simple, direct link between the two genes known to be involved in dominant Parkinson's disease.  However, we should also probably look for some supportive evidence that this occurs in the brain under physiological conditions and this is where interpretation becomes cloudy.

To ilustrate why this is difficult, let's accept that pS129-Syn is a major species found in Lewy bodies and also that Lewy bodies are a major pathology of LRRK2-linked parkinsonism.  Both of these are reasonable propositions, and so it seems reasonable to infer that LRRK2 mutations might promote pS129-Syn formation in the human brain, which is about as definitive as one can get for a physiological context for this disease process.  The problem is that we know that at least three classes of kinases (casein kinase 2 [CK2], used as a positive control here, g-protein coupled kinases [GPRK2/5] and polo-like kinases [PLK2]) are capable of adding phosphate groups to the sidechain of S129 in vitro so the presence of this modified residue in vitro or in vivo is not unambiguous proof that pS129-syn came from LRRK2 action. 

What new data do we need, then, to support the hypothesis that LRRK2 is an in vivo kinase for pS129?  First, some additional in vitro data would be helfpul.  We know that even if two kinases can phosphorylate the same residue of a protein, one may be more efficient than another.  For this reason, it is more likely the PLK2 is a major kinase for pS129 than CK2 as the reaction is quantitatively more efficient with the former than the latter (see eg figure 1A of Inglis et al).  So, seeing if LRRK2 is a quantitatively efficient kinase for pS129 over time for a range of substrate (synuclein) concentrations would be helpful, as would be checking whether other residues are phosphorylated under these conditions.  It would be helpful to see if the reaction occurs at concentrations of LRRK2 and synuclein that are likely to be present in the brain, although this is a difficult question to answer definitively.

Second, if other kinases, many of which are present in cell lysates are capable of phosphorylating this residue then we should really exclude that they are not present in the purified LRRK2 preparation.   The increased reactivity with the G2019S mutation, which appears to be activating for LRRK2 kinase activity in most peoples' hands is very helpful here but it would be better to also include an artificial kinase-dead LRRK2 at the same expression levels to really exclude co-precipitated kinases.  Also, highly purified LRRK2 fragments (eg the material available commercially) would be reassuring and could be run in parallel.

Third, and most importantly, LRRK2 needs to be shown directly to produce pS129 in vivo.  In cell models, one should be able to overexpress and knock down LRRK2 and increase/decrease the amount of pS129 recovered.  A negative result here might not be definitive (perhaps the LRRK2 is not activated in a transfectable cell line, for example) so it would be absolutely ideal to confirm this in LRRK2-transgenic and LRRK2-knockout mouse brain.  Of course, if one had a cell permeable potent and specific LRRK2 inhibitor, then one would be able to derive additional supportive data.

There are probably more experiments that could be done, I am using this as an example of why it is difficult to be certain of the relationship between a putative kinase and substrate.  For a first publication that will have huge impact on field, the Qing et al paper has enough data to be able to follow what was done and replicate it.   But from here, a huge amount of work needs to be started, thinking as always of ways to disprove the hypothesis if possible.

In thinking about LRRK2 kinase substrates, one first ponders whether LRRK2 even possesses the ability to phosphorylate other proteins in cells, and that LRRK2 kinase activity is not exclusive to auto-regulatory capacities. The corners and back rows of meeting halls, as alluded to by Dr. Fiske, also harbor comments that LRRK2 is going to resolve as a scaffolding type protein with several domains potently modified by autophosphorylation. This argument is (mistakenly) emboldened by the lack of replication of the published candidate LRRK2 kinase substrates in different laboratories and the lack of success of some biochemical screens that have had some level of past success in linking other protein kinases with substrates.

The current circumstantial evidence that suggests the existence of LRRK2 kinase substrates outweighs the notion that there are none. First, the LRRK2 kinase domain is similar in composition and phylogenetically close to domains associated with proteins with bona fide kinase substrates (e.g., MLKs). Second, full length recombinant LRRK2 catalyzes the transfer of phosphates to myelin-basic protein in vitro at rates that far exceed autophosphorylation (at least in our laboratory). Third, work from the MRC Phosphorylation Unit and Wyeth using a LRRK2 fragment and peptide substrates demonstrate kinetic and binding constants similar to other Ser/Thr protein kinases with identified physiologically relevant substrates.

We probably can say that if the majority of LRRK2 stably interacted with and phosphorylated a (ubiquitously expressed) substrate protein, it would have been identified. Our laboratory, as with many others, continue to plow through LRRK2 interacting proteins and data from arrays with the hopes that we see, even in vitro, phosphorylation kinetics even slightly suggestive of a potential substrate. As yet, we have not identified another protein besides MBP that is not dwarfed by autophosphorylation signal in vitro. While we are disappointed by the reproducibility of newer array technologies (recombinant protein and antibody arrays), there are recent developments in mass spectrometry identification of phosphorylated peptides that may be up to the challenge of identifying LRRK2 kinase substrates in an unbiased manner. The implementation of techniques appropriately powered to identify LRRK2 substrates will be laborious, costly, and completely necessary.

The relationship between LRRK2 and alpha-synuclein, two genes linked to familial PD in an autosomal dominant manner, is still largely unknown. Although studies focussing on the relationship between these two genes are scant, current knowledge points more often than not to a diffuse relationship between LRRK2 and alpha-synuclein with both genes acting at separate levels in a complex pathogenic PD process  The recent report by Qing et al. proposing that LRRK2 directly phosphorylates alpha-synuclein at serine 129 provides the intriguing possibility that the PD pathogenic processes may not be as complicated as it is generally believed to be.

However, the report raises some questions. Many LRRK2 researchers have been confronted with the difficulty of preparing active recombinant enzyme. Qing et al used transfection of eukaryotic expression plasmids in HEK293T cells as expression system, a good choice as this system has previously been shown to yield active LRRK2 enzymes. However, the authors used crude cell extracts as a source of enzyme. The conclusion that LRRK2 directly phosphorylates alpha-synuclein in this system is therefore tenuous as the reaction mix presumably contains multiple alternative kinases endogenously present in these cells. What is the level of purity of the enzyme used in this study? Also, does the method of protein preparation as applied in the study yield active protein? Interestingly, the G2019S mutant which has consistently been shown to possess higher phosphorylation activity than WT yields higher phosphorylation levels of alpha-synuclein while cell extracts of normal cells yields no phosphorylation. Is it then possible that LRRK2 WT or G2019S induces or activates another kinase in HEK293T cells which then effectively phosphorylates alpha-synuclein? Also, the study proposed end point readouts at 6 hours of incubation. Does this mean that the proposed alpha-synuclein phosphorylation is a very slow process? What are the kinetics of the proposed alpha-synuclein phosphorylation? As suggested in previous comments on this issue, the inclusion of a kinase dead negative control would be a very useful addition to the study.

The validation of this finding will require first a thorough confirmation at the in vitro level to discriminate between LRRK2 phosphorylation and phosphorylation by other cellular kinases. The use of proteins with a high level of purity and confirmed activity should help to address this issue. If the phosphorylation of alpha-synuclein by LRRK2 is true and physiological, then we have an extremely exciting development on our hands.

Two main observations have driven researchers to focusing on the identification of LRRK2 substrates: first, the most common LRRK2 mutation G2019S enhances kinase activity and, second, this activity is required for mutant proteins to be toxic in neuronal models.  However, despite the number of laboratories involved in this “treasure hunt”, no convincing physiological substrates have been confirmed to date.  All reported substrates either require denaturation to be phosphorylated or are used in large molar excess compared to LRRK2.
 
In contrast, LRRK2 autophosphorylation seems to be relatively efficient, at least in vitro, and has been confirmed by many different groups.  This last observation suggested us that understanding autophosphorylation in more detail might provide some clues about LRRK2 function.  Combining data from three different collaborating laboratories, we were able to map several LRRK2 autophosphorylation sites (Greggio et al., BBRC, 2009).  We found, quite surprisingly, that LRRK2 phosphorylates itself predominantly within its GTPase/ROC domain.  Phosphorylation of a threonine1343 was consistently observed with several different methods and we could confirm that this residue authentically resulted from autophosphorylation by using a kinase dead LRRK2 molecule. T1343 residue is predicted to be involved in GTP binding in either of the two models of the ROC domain that are available.  Therefore, we hypothesize that T1343 will impact GTPase function, which suggests an intriguing inter-molecular regulation of the GTPase domain by the kinase domain.

 
There are several important next steps that we have started in the lab.  First, confirmation of the results in vivo is important and probably the key reagents are to develop phosphorylation specific antibodies against the various sites that we have found to date.  Second, we have strong evidence that there are additional sites outside of the two that we have validated to date.  Additional mass spectrometry approaches are probably needed to identify these sites.  Third, we need to show that each of the putative phosphorylation sites have functional effects, which we are approaching using both biochemical and cellular assays.

It would be interesting to discuss these data with other researchers working on the identification of LRRK2 autophosphorylation sites and see, for instance, if similar results were found.

For some time, researchers used autophosphorylation of LRRK2 or transphosphorylation of the generic MBP substrate as a readout of LRRK2 activity.  We undertook a KESTREL (kinase substrate tracking and elucidation) screen and identified the ERM proteins as efficient substrates of LRRK2 in vitro (1).  The development of LRRKtide as a peptide substrate has enabled the necessary high throughput screening process for small molecule inhibitors to proceed. However, it proved to be difficult for us (and no doubt others) to ascertain whether modification of the ERM substrates in vivo was a result of LRRK2 activity, as a multitude of kinases have been reported to modify these substrates.  It became clear an alternate approach was necessary.

In collaboration with Lew Cantley and Jessica Hutti the substrate specificity of LRRK2 was assessed by screening a positional scanning peptide library (2).  The substrate preferences revealed by this screen were substituted into a longer version of the widely used LRRKtide, resulting in a high affinity substrate for LRRK2, termed Nictide-RLGWWRFYTLRRARQGNTKQR.  The utility of this novel substrate is several fold, including the potential to predict LRRK2 substrates and the ability to assay endogenous LRRK2.  Additionally, we discovered that LRRK2 exhibits a preference for Thr residues over Ser residues. 

Combining Nictide with a new, specific antibody against LRRK2 has enabled the first reported assessment of endogenous LRRK2 kinase activity in the context of an immunoprecipitation-kinase assay.  This, at the very least, answers the question of whether endogenous LRRK2 possesses kinase activity.  It is our hope that the Nictide sequence could be used to probe sequence databases for the identification of bona fide LRRK2 substrates.  If successful, this would prove a lucrative discovery for the researcher.  We felt it best to open this expedition to the PD research community in order to make the discovery process more rapid, furthering our field as well as drug discovery endeavors. 

We also describe the characterization of a novel pharmacological scheme, using tool compounds, for investigating whether potential substrates are indeed in vivo substrates for LRRK2.  We found that widely utilized inhibitors of the rho-associated kinase (ROCK), such as Y-27632 and H-1152, also exhibited inhibitory properties against LRRK2 at concentrations similar to that of ROCK.  Additionally, we described inhibitors that are selective for ROCK over LRRK2 (GSK429286A and GSK269962A) as well as sunitinib, which is more selective for LRRK2 over ROCK. 

There has been much effort put into understanding the role of LRRK2 since it’s implication in the pathogenesis of Parkinson’s disease.  However, there is much more to uncover, such as which proteins interact with and/or are modified by LRRK2, where does LRRK2 reside in vivo to exert these functions and also are there any inputs from upstream modifiers that could regulate LRRK2 function.  Basic science inquiry into the intrinsic functions of LRRK2, (as well as other proteins encoded by the PARK loci) and how these processes may be perturbed in disease state or in the cause thereof, will result in the development of cell based assays to asses the utility of small molecule inhibitors for the disruption of LRRK2 function.  This will facilitate potential development of a therapeutic regime for treating PD, familial or idiopathic.  As the field moves forward, it is crucial to begin to rigorously distinguish between properties of LRRK2 that are artifacts of experimental protocols from real facets of LRRK2 biology that require incorporation into models of LRRK2 function. 

A full text PDF of Nichols et. al, 2009 is available at http://www.biochemj.org/bj/imps/abs/BJ20091035.htm

In 2007, our group and others began developing ways to visualize LRRK2 conformations using native PAGE and gel filtrations, and found that a proportion of LRRK2 protein is present as dimer-sized structures in lysates. The Cookson group published an initial insightful study in 2008 that combined autophosphorylation studies together with the first approaches that allowed measurements of LRRK2 dimers via native-gels and SEC separations. Prior to that, LRRK2 self-interaction was demonstrated clearly through the use of co-immunoprecipitations of differentially tagged LRRK2, but these approaches fail to distinguish oligomeric LRRK2 from dimeric LRRK2, and thus are only successful at excluding monomeric LRRK2 from everything else.

As the Cookson group demonstrated in their recent publication, we knew that LRRK2-kinase inactivating mutations strongly reduce dimer-formation and instead shifts LRRK2 to higher-molecular weight oligomers, at least in full-length LRRK2 protein in lysates. We sought to develop ways to stably separate the LRRK2 conformations present in soluble lysates and measure activity, and found appreciable kinase activity only with dimeric LRRK2. The implication is that the dimer conformation itself may be the active form, and therefore may lend additional insight into how pathogenic LRRK2 mutations change LRRK2 activity. We became convinced after many experiments that some pathogenic mutations tend to increase LRRK2 dimer concentrations relative to the total pool of LRRK2 protein. The effects are subtle but this is consistent with what we see with these mutations through in vitro kinase assays. Going the other way with pharmacological inhibition of LRRK2 kinase activity produces more dramatic results that don’t need statistics, and leads to the first possible utility of a LRRK2 dimer:

1- Potential LRRK2 kinase inhibitors may be immediately tested on cells expressing LRRK2, and coupled with native-gels, one can get an idea whether the compound might be effective in living cells for ablating LRRK2 kinase activity via the visualization of LRRK2 dimers. This approach may be more comprehensive than looking at, for example, a LRRK2 autophosphorylation site, but clearly less effective than looking a genuine LRRK2 substrate.

Another possible utility:

2- LRRK2 expression is strong in human peripheral cells. Is the ‘active’ LRRK2 dimer species higher in blood cells from PD cases, and could this have diagnostic value?

Finally,

3- Protein-protein interaction technologies may be able to exploit LRRK2 dimerization for the visualization of LRRK2 activity in living cells, and help sort out what activates LRRK2. One challenge may be the identification of fusion tags that don’t change the properties of LRRK2. For example, in our hands, eGFP fused to either the N or C terminus changes the localization of LRRK2 and causes perinuclear aggregates we don’t see with untagged LRRK2.

A recent publication from Mark Cookson's group suggests that 4E-BP is likely not a direct substrate for LRRK2 kinase activity, following up on initial studies by Imai et al. (2008). They confirmed that LRRK2 can phosphorylate 4E-BP in vitro, but it's activity was relatively weak compared to LRRK2 autophosphorylation. Furthermore, an alternative kinase (MAPK14/p38alpha) was shown to phosphorylate 4E-BP at the same site and could phosphorylate 4E-BP in vivo, whereas LRRK2 could not.

All in all, the report seems to put the lid on 4E-BP as a LRRK2 substrate, instead pointing to 4E-BP phosphorylation due to p38-mediated stress responses. However, though not likely a direct target of LRRK2, 4E-BP still may be a target for neuroprotection, and has been recently genetically linked to PD though EIF4G1, as presented by Matt Farrer at WFN 2009 and reported on PD Online by Mark Cookson.