targets
T. brucei protein kinases can be targeted with chemical inhibitors in the low
micromolar and nanomolar range of concentration (Urbaniak et al., 2012). Despite conservation of the protein kinase catalytic domain, protein kinase inhibitors can be designed for selectivity and have been proven good drug candidates (Cohen, 2002; Weinmann and Metternich, 2005; Cohen and
Tcherpakov, 2010). This work identifies 49 protein kinases that are required for
in vivo fitness. For those causing strong phenotypes, leading to parasite
clearance (Table 3-3), it may provide evidence of their potential as drug targets, though a more thorough genetic validation is required, as lethal phenotypes observed upon protein kinase depletion via RNAi are not necessarily due only to protein kinase inhibition. For instance, the absence of the protein in a given
interaction network can cause harm by itself, something that could not be reproduced with a chemical inhibitor.
One method to analyse this would be integration into the RNAi line of a recoded version of the targeted gene. The resulting transcript would be unharmed by the RNAi machinery, while producing the original protein. If this recovers the wild type phenotype, we would confirm that RNAi is on target. If, on the other hand, cells are unable to recover when the recoded gene has key mutations for
activity, e.g. substitution of the catalytic lysine in the second subdomain (Hanks and Hunter, 1995), it would confirm that the same phenotype could be obtained with a targeted kinase inhibitor. One protein kinase in the set, RDK2, has been targeted with this approach and results are described in the next chapter.
Another important consideration is the timing and severity of the outcome. Upon chemical inhibition, the derived phenotype can happen immediately after
addition of the compound. RNAi requires transcription machinery to be active and, as transcript depletion avoids production of fresh protein, time is required to deplete the cell of available stocks. To examine transcript and protein levels with qRT-PCR and western blot in order to understand the timing is critical. In addition to the obvious possibility of designing protein kinase-specific screens targeting them individually, some protein kinases identified here open an
opportunity to develop drug-repurposing strategies. A phylogenetic comparative between the T. brucei and Homo sapiens kinomes showed how 13 out of the 49 protein kinases found to be important in vivo had close human orthologues targeted by at least one of the compounds in the ‘HAT box’ (Peña et al., 2015). Interestingly, among them is Tb927.10.10350 (STE11/Bck1p), from the in vivo only set. These molecules could be chemically modified to achieve T. brucei specificity. This has been also suggested for aurora kinases (Patel et al., 2014) and PI3K/TOR kinase inhibitors (Diaz-Gonzalez et al., 2011). Both pathways have putative components detected also in the in vivo-only set, Table 3-6.In contrast, the other 8 in vivo validated protein kinases are highly divergent from any
human kinase. Three ‘orphan’ kinases in this set are very likely to play parasite- specific functions. Selectively targeting these may be easier than other more conserved protein kinases.
3.4.3.1 Leishmania/Trypanosoma cruzi extrapolation
Most of these 49 protein kinases have close orthologues in Trypanosoma cruzi and Leishmania major (see Figure 3-13/Table 3-5). Extrapolation of results based on gene ontology using T. brucei as a model may be a valuable
perspective to identify potential drug targets, since to date these organisms lack efficient reverse genetics tools for kinome- or genome-wide studies. Examples of protein kinases with a loss of fitness in our study, which also play essential roles for survival of Leishmania include MPK1 (called KFR1 in T. brucei (Wiese, 1998), CK1 (Knockaert et al., 2000; Rachidi et al., 2014), CRK3 (Hassan et al., 2001; Walker et al., 2011) and GSK3-short (Xingi et al., 2009; Ojo et al., 2011). The last three are genetically validated targets and have had HTS screens where inhibitors have been identified. In addition, one of the MAP3Ks observed to have an only in vivo fitness defect, MRK1 (with a role in osmotic resistance), has a L.
major orthologue proven essential for promastigote survival (Agron, Reed and
Engel, 2005). All the main hits causing a loss of fitness in Table 3-3 are
potentially worthy of further consideration in this manner. For example, the L.
donovani orthologue of AUK1 (LdAIRK) has been tested showing implications in
cell cycle but requires further analysis (Chhajer et al., 2016). KKT10, the kinetoplastid-specific kinetochore member, may be targeted with specificity, like CK2A1 and CK2A2,with human orthologues targeted by compounds with trypanocidal activity in the HATbox (Peña et al., 2015), suggesting potential for drug repurposing. A CK2-targetting compound may have enhanced in vivo
efficacy, according to results shown in this chapter. Supporting its potential, a CK2 orthologue in L. braziliensis has been shown to be secreted, helping
macrophage association prior to invasion (Zylbersztejn et al., 2015). Others, like DYRK1, CRK9, AUK3, or RDK2, which are highly divergent from any human
orthologues according to a phylogenetic comparative made by Peña et al (2015), may be interesting kinetoplastid-specific targets too, Table 3-5.
3.4.4 Potential applications of RITseq used in other pre-existing
libraries
With this study we prove that RITseq can be applied to any pre-existing RNAi library, expanding our capacity to search for genes involved in specific aspects of T. brucei biology for which a screen can be devised. Reduced library size may
be an advantage by increasing resolution and requiring less starting material in order to obtain significant results, as we have shown by in vivo infections. The barcoded primer approach reduced costs and increased our multiplexing
capacity. Possibilities are endless, but we have already used this library to identify protein kinases required for kinetoplast maintenance, cell cycle regulation, withstanding or repairing alkylation-mediated DNA damage, and differentiation of bloodstream to procyclic forms.
Technically it would be possible to use this library to find the target of a kinase inhibitor whose time to kill at a given concentration and EC50 is well profiled. We should be able to find a time frame where a synergistic effect due to overlap of kinase inhibition and kinase depletion can produce a reduction the IC50. The population, after a certain incubation time, would be enriched in the parasites that do not target with RNAi the liable protein kinase.
The library could also be used for the identification of the protein kinase responsible for a particular phosphorylation event, if a specific antibody is available against the phosphorylated peptide. The library, induced and
uninduced, would be fixed before staining with such antibody, which could be directly or indirectly coupled to a fluorophore. Then, cells would be processed through fluorescence-activated cell sorting (FACS) so the non-phosphorylated (and non-stained) population could be isolated. Sequencing the RNAi targets of the unstained population would show which kinase depletions were responsible for the lack of phosphorylation. The main challenge would be choosing the right timing in case the responsible kinase caused a strong loss of fitness. Induction should be sustained long enough for the protein kinase to be depleted so the phosphorylation defect becomes evident. However, it cannot be so long that the cell line of interest is cleared if the protein kinase is essential.
This rationale can be subsequently applied to libraries of genetically modified parasites of many different makes (e.g. overexpression, knock out, fluorescent or luminescent tagging) as soon as we can use universal primers to enrich in a quantifiable molecular label, e.g. a barcode, an RNAi cassette or an
overexpression cassette. For instance, an RNAi library made in a bioluminescent pleomorphic parasite could be used to test which proteins are needed to invade the central nervous system by infecting mice. We could track the progress of the
parasites throughout the host organism to certify invasion of the central nervous system with an in vivo imaging system (IVIS), before purifying DNA from the brain tissue or haematoencephalic fluid. Genomic DNA would be enriched in the cassette and sequencing reads assigned to it filtered in silico. No protein kinase has been reported to date as specifically required to cross the blood brain barrier (BBB) in the parasite, although this signalling is fundamental in the endothelial cells to permit this to happen. Examples of T. brucei-secreted factors proven to have a direct role in BBB crossing are two cathepsin-B-like cysteine-proteases (brucipains). In addition, others have been proposed that may have a role are the closely related oligopeptidase B or acid phosphatases
expressed in the surface (Lonsdale-Eccles and Grab, 2002). Focused RNAi libraries targeting protein kinases or peptidases could be used in this manner. The screen described in this thesis is not foolproof. False negatives in RITseq studies are rare, but can be caused by a failure to produce dsRNA or leaky phenotypes giving a reduced count (due to cell line depletion) both in the induced and uninduced pools (Glover et al., 2014). Starting with a pool of validated cell lines is time-consuming but gives a strategic advantage. False positives can happen due to rare off-target effects. With time, loss in complexity in the induced sample could cause unequal expansion of different clones in the library, due to uneven access to energy resources. Deletion of the RNAi target from the genome after induction has been also detected.
3.4.5 Final conclusion
All together, we showed in this chapter that application of next generation sequencing to pre-existing libraries of RNAi lines enhances greatly our capacity to evaluate the impact of protein depletion on the cell phenotype when
subjected to selective pressures of any kind that mimic relevant cellular
processes. As a proof of concept we used the library to understand what protein kinases in T. brucei are indispensable for mammalian infection, something of interest both for drug discovery and biology. We validated in vivo the effect in loss of fitness that had been observed in vitro for more than 40 protein kinases. Interestingly, we identified at least 9 which showed a degree of essentiality in the context of a mouse infection that could not be reproduced in culture, therefore behaving as potential virulence factors. Most of these in vivo-
indispensable protein kinases showed phylogenetic relationship with orthologues in other organisms that had a role in regulation of stress response pathways, including osmotic shock control, gene expression or autophagy mediated by the PI3K/TOR axis. We demonstrated the contribution of some of these in vivo-only kinases towards serum resistance, and tolerance to osmotic shock.
Severity of the phenotype in more than 20 protein kinases in this screen highlights their potential value as drug targets. We discussed how kinase inhibitor chemotypes are generally enriched in the available collections of
trypanocydal compounds. All of them had orthologue proteins in L. major and T.
brucei, and they may be also essential in those organisms. Until proper high-
content genetic dissection tools, equivalent to RNAi, are developed in these parasites, extrapolation from observations made in T. brucei can be a successful approach.
Given the reduced size of the library, this technique provided the advantage - compared to whole-genome approaches- of increasing greatly our resolution in terms of read depth, permitting more compelling results using a smaller number of parasites, as required for mice inoculation. Possibilities of RITseq and similar genetic approaches are endless.