Introduction

Although Parkinson’s disease (PD) is the second most common neurodegenerative disease [1] the underlying pathogenesis is still widely elusive. Genetic analyses of familial forms of PD lead for the first time to a specific chance to study alterations causative for PD. A protein that plays an important role in this disease is the leucine-rich repeat kinase 2 (LRRK2, PARK8) [2]. Until now, more than 20 mutations were found in LRRK2 accounting for approximately 5–10% of familial PD and up to 3% of sporadic PD cases, both indistinguishable in there pathological hallmarks [3–12]. Therefore, clarifying of the biological function of the 280 kDa multidomain protein LRRK2 will give new insights in the pathogenesis of PD.

The LRRK2 protein consists of a number independent protein domains, and most of the PD associated mutations affect nearly every predicted catalytic and protein–protein interaction domain [13]. One of the known protein–protein interaction domains of LRRK2 is the leucine-rich repeat (LRR) domain [13, 14]. This repeat is flanked by a Rho/Ras-like GTPase domain (termed Roc, for Ras in complex proteins) followed by an associated C-terminal of Roc (COR) domain, which function is still unknown [13]. The Roc-GTPase domain of LRRK2 resembles most closely the Rab (Ras-related in brain) family of GTPases [13], which are central regulators of vesicle formation, actin- and tubulin-dependent vesicle movement, and membrane fusion [15]. C-terminal of the Roc-GTPase and its COR domain LRRK2 contains a protein kinase domain. Additionally, to the previous mentioned LRR domain, there is weak prediction for a single WD40 repeat at the C-terminal end of LRRK2, namely the WD40 domain [16]. WD40 repeats could serve as scaffold for protein interactions and these interactions can occur simultaneously with several different proteins [17]. The N-terminal side of LRRK2 contains 14 evolutionary conserved repeats with unknown function [16]. Although the domain structure of LRRK2 is well known, the elucidation of the biological function of the protein is still in the early stages. By means of RNA interference (RNAi; for a review see [18]), we reduced the mRNA level of LRRK2 in the human dopaminergic neuroblastoma cell line SH-SY5Y. The examination of the alterations in the expression profile (with Affymetrix Human Genome U133 Plus 2.0 Arrays) provoked by the loss of LRRK2 provided us with new insights in the signaling pathways of the protein.

Material and methods

Cell culture and RNA interference

Human SH-SH5Y cells were maintained in D-MEM medium (Invitrogen) supplemented with 15% fetal bovin serum and penicillin–streptomycin.

RNAi of leucine-rich repeat kinase 2 (LRRK2) was performed with two different chemical synthesized siRNAs against LRRK2 (Qiagen) and control siRNA (Qiagen) with random sequence not present in the human genome. The effect of these siRNA (listed in Table 1) was determined with qRT-PCR.

Table 1 Sequences of transfected siRNA (Qiagen)
Full size table

All transfection experiments were performed in independent triplicates with the Cell Line Nucleofection Kit V (Amaxa) according to the manufactures protocol. For the electroporation with program G-04, we applied 2 × 106 cells and 2 μg siRNA.

RNA isolation and cDNA preparation

Total RNA of transfected cells (48h after RNAi treatment) was isolated using the RNeasy Mini Kit (Qiagen) including a DNase digest on the column with RNase-Free DNase Set (Qiagen). The RNA quality was controlled by Lab-on-Chip-System Bioanalyser 2100 (Agilent), and the concentration was determined using a BioPhotometer (Eppendorf).

cDNA synthesis was performed with 1 μg of total RNA using the QuantiTect Reverse Transcription Kit (Qiagen). For examination of genomic DNA contamination, we included a RT-negative control.

Microarray analysis

Microarray analysis was performed using an Affymetrix platform and the according materials and protocols. Fragmented and labeled cRNA of control siRNA treated cells and siLRRK2-1 treated cells (triplicates of each condition) were applied to the hybridization using the Human Genome U133 Plus 2.0 Array (Affymetrix). Data analysis was performed with ArrayAssist 4.0 (Stratagene) with a low stringency filtering scheme of 1.5-fold change in expression level and p value of less than 0.05 after GC-RMA normalization of all data sets.

Generating of functional categories and pathway analysis

The probeset IDs of the differentially expressed transcripts were integrated in BiblioSphere (Genomatix software GmbH), and Gene Ontology categories were identified for annotated genes. This software facilitates the categorization of the genes to biological processes, while one gene could be associated to more than one functional category. For each GO category, a Z-score is calculated, which indicates how far and in what direction that term deviates from its distribution’s mean, expressed in units of its distribution’s standard deviation. The significance of the GO group increases with higher levels of the Z-score and we choose 3.9 as cut off.

Gene regulation networks were generated using Ingenuity pathways analysis software (ingenuity® systems, www.ingenuity.com). The data set with differentially regulated transcripts and corresponding expression values was uploaded into the application. Each gene identifier was mapped to its corresponding gene object in the ingenuity pathways knowledge base. The genes were overlaid onto a global molecular network developed from information contained in the ingenuity pathways knowledge base. Networks of these focus genes were then algorithmically generated based on their connectivity. All edges are supported by at least one reference from the literature, from a textbook, or from canonical information stored in the ingenuity pathways knowledge base. The functional analysis was also done with this software. Genes from the microarray dataset that met the fold change cutoff of 1.5 and that were associated with a canonical pathway in the ingenuity pathways knowledge base were considered for the analysis. The significance of the association between the data set and the canonical pathway was measured in two ways: (1) A ratio of the number of genes from the data set that map to the pathway divided by the total number of genes that map to the canonical pathway is displayed, and (2) Fischer’s exact test was used to calculate a p value determining the probability that the association between the genes in the dataset and the canonical pathway is explained by chance alone.

Quantitative real-time PCR

The quantitative real-time PCR (qRT-PCR) was performed on the LightCycler 480 System (Roche) in 384-well format. Specific QuantiTect Primer (Qiagen) or Primer designed with the Primer3 Software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) spanning an Exon–Exon boundary were applied for amplification (listed in Table 2). We used 10 μl reaction volume containing 1 μl of 1:5 diluted cDNA and the QuantiTect® SYBR® Green PCR Kit (Qiagen) with PCR settings according to the manufactures protocol. Melting curves were generated at the end of the run. Standard curves of each amplified gene were created, to obtain the PCR efficiency. The cp-values of the genes were detected by the LightCycler 480 Software release 1.2.0.0625 applying the second derivative maximum method. For data analysis, the best out of two reference genes (SDHA and PDHB) was identified by the GeNorm algorithm implemented in the qBASE application [19, 20] and subsequently used for normalization and calculation of the relative expression with REST-384© version 2 [21, 22].

Table 2 Oligonucleotides for the qRT-PCR
Full size table

Results

Efficiency of RNAi

RNAi downregulation of LRRK2 was performed with two distinct siRNA as well as a nonhomologous control siRNA in SH-SY5Y cells. Downregulation of LRRK2 mRNA was confirmed with qRT-PCR. Control siRNA did not result in a change in LRRK2 expression level relative to untransfected cells whereas electroporation with siLRRK2-1 leads to an obvious reduction of LRRK2 mRNA level to a relative expression of 0.25 (Fig. 1). However, the relative expression of LRRK2 in cells treated with siLRRK2-2 achieved only 0.38. Thus, siLRRK2-1 was identified as the most effective mediator of the RNAi of LRRK2 and subsequent experiments were performed using this siRNA.

Fig. 1
figure 1

Reduction of LRRK2 expression by RNAi. SH-SY5Y cells were transfected with two different siRNAs and expression was determined by qRT-PCR. The analysis was done in independent replicates (error bias = SEM). PDHB was used as reference gene. ***p value ≤ 0.001

Full size image

Microarray analysis

RNAi of LRRK2 is likely to influence the expression level of transcripts, associated with the biological function of LRRK2. To identify changes in the expression level of putative interactors, we performed a microarray analysis with Affymetrix Human Genome U133 Plus 2.0 microarrays. The analysis by means of the ArrayAssist 4.0 software identified 187 transcripts differentially regulated between triplicates of control siRNA and siLRRK2-1 transfected SH-SY5Y cells. There were 94 of these transcripts upregulated and 93 transcripts downregulated with a fold change of at least 1.5 and a p value of less than 0.05. A complete list of all differentially regulated genes in this RNAi experiment is provided in the Supplementary material 1.

Identification of biological context-network analysis

GO categories

The BiblioSphere software (Genomatix software GmbH) was used to assort the 187 differentially expressed transcript to gene ontology (GO) categories of biological processes. A total of 131 of the 187 differentially expressed transcript IDs could be mapped in the analysis, and 65 are alignable with GO categories (Fig. 2). The overview about the differentially expressed transcripts in the influenced 16 GO categories could be found in the Supplementary material 2.

Fig. 2
figure 2

Differentially expressed genes were grouped by GO functional categories of biological processes. The numbers of differentially expressed transcripts of one GO category are shown as bar and the specific Z-score is shown as rhomb

Full size image

We examined four differentially regulated transcripts which are associated with axonal guidance and this category obtained the highest Z-score with 7.1. In the biological category “nervous system development” 14 genes were disregulated with the second highest Z-score of 6.7. Furthermore, four transcripts (Z-score 6.6) within the biological processes “cell cycle arrest” and five transcripts (Z-score 5.9) of “regulation of cell growth” showed a different expression compared to cells transfected with control siRNA. Downregulation of LRRK2 influenced also the expression profile of six transcripts (Z-score 5.9) related to “cell projection morphogenesis”. Additionally, 8 “cell motility genes” (Z-score 5.2) and 14 “anatomical structure morphogenesis genes” (Z-score 5) were disregulated. The biological process “cell communication” highlighted the highest number of differentially regulated transcripts (30, Z-score 4.8) in our microarray experiments. Accessorily, 6 transcripts (Z-score 4.4) of “regulation of multicellular organismal process”, 4 transcripts (Z-score 4.3) of the GO category “MAPKKK cascade” and 26 transcripts (Z-score 4.3) of “developmental process” appeared disregulated in LRRK2 repressed cells. The analysis point out two GO categories with a Z-score of 4.1, these include 17 differentially regulated transcripts of “cell differentiation” and 4 differentially regulated transcripts of “muscle contraction”. Furthermore, we examined the regulation of 8 transcripts (Z-score 4) associated with “regulation of progression through cell cycle” and 25 transcripts (Z-score 3.9) of “signal transduction”. In the GO category “Ras protein signal transduction” four genes (Z-score 3.9) appeared differentially upregulated in LRRK2 repressed cells.

Ingenuity canonical pathway analysis

By means of the ingenuity pathways analysis software (Ingenuity Systems), we analyzed the canonical pathways which are affected by the repression of LRRK2 with RNAi (Fig. 3). A total of 151 of the 187 differentially expressed transcript IDs could be mapped in the analysis and 84 are alignable with functions. We defined a deregulation of transcripts in nine different canonical pathways with a −log(p value) ≥0.98 and a ratio ≥0.017. Affected pathways included axon guidance signaling, actin cytoskeleton signaling, calcium signaling, Wnt/ß-catenin signaling, ephrin receptor signaling, integrin signaling, SAPK/JNK signaling, and Xenobiotic metabolism.

Fig. 3
figure 3

Analysis of canonical pathways affected by LRRK2 RNAi. The analysis was done with Ingenuity Pathways Analysis software. Fischer’s exact test was used to calculate a p value (shown as bars) determining the probability that each biological function assigned to the network is due to chance alone. The ratio (shown as squares) represents the number of differentially expressed genes in a given pathway divided by total number of genes that make up that canonical pathway

Full size image

Since one gene could be part of various pathways, we identified the differentially upregulated gene CDC42 (cell division cycle 42, fold change 2.47, p value 0.01) as major gene participating in five of the nine identified canonical pathways. Among these are axon guidance signaling, actin cytoskeleton signaling, ephrin receptor signaling, integrin signaling and SAPK/JNK signaling. Additionally, the upregulated guanidine exchange factor ARHGEF7 (Rho guanine nucleotide exchange factor 7, 2.01, 0.02) has a key position in integrin signaling, actin cytoskeleton signaling and axon guidance signaling. The downregulation of CUL3 (cullin 3, −1.77, 0.04) and MGMT (O-6-methylguanine-DNA methyltransferase, −2.70, 0.001) as well as the upregulation of NQO1 [NAD(P)H dehydrogenase, quinone 1, 1.74, 0.03] and PPP2R2A (protein phosphatase 2 regulatory subunit B, 1.59, 0.0003) are features of the impaired xenobiotic signaling. The protein phosphatase subunit is a component of the canonical Wnt/β-catenin signaling pathway. In this pathway, CSNK1G1 (casein kinase 1 gamma 1, 1.86, 0.02), DVL2 (dishevelled dsh homolog 2, 1.53, 0.003), and GNAO1 (guanine nucleotide binding protein alpha activating activity polypeptide O, 1.76, 0.004) are misregulated. Another protein phosphatase subunit PPP1R12B (protein phosphatase 1 regulatory subunit 12B, 2.09, 0.002), which appeared to be upregulated in the microarray experiment, is part of the integrin signaling cascades. The key player of this canonical pathway is the upregulated integrin ITGB5 (integrin beta 5, 1.83, 0.0007). The change in expression profile in ephrin signaling is characterized by the upregulation of GNAO1, CDC42, EFNB2 (ephrin-B2, 1.55, 0.0003) and the downregulation of CXCR4 (chemokine C-X-C motif receptor 4, −2.02, 0.0003). All of these genes participate also in axonal guidance in addition to ARHGEF12 (Rho guanine nucleotide exchange factor 12, 1.56, 0.009), BMP2 (bone morphogenetic protein 2, 1.57, 0.04), SEMA4F (sema domain 4F, -1.61, 0.002), and the previous mentioned gene ARHGEF7. Five genes are upregulated in calcium signaling cascades, namely CACNA1G (calcium channel voltage-dependent, alpha 1G subunit, 1.59, 0.004), CAMKK2 (calcium/calmodulin-dependent protein kinase kinase 2 beta, 1.83, 0.05), TPM1 (tropomyosin 1, 1.79, 0.03), TPM3 (tropomyosin 3, 1.62, 0.013), and TPM4 (tropomyosin 4, 1.62, 0.0009). The disregulation of the actin cytoskeleton signaling is characterized by the upregulation of CHRM3 (cholinergic receptor muscarinic 3, 1.88, 0.0007) and FGF2 (fibroblast growth factor 2, 1.6, 0.03) in addition to ARHGEF7, CDC42, PPP1R12B, and ARHGEF12, which participates in other pathways too. Although only three genes are misregulated in the SAPK/JNK canonical pathway, they reached significance over the −log(p value) threshold (1.3). In this pathway, CDC42 and CHRM3 were upregulated and the kinase ZAK (sterile alpha motif and leucine zipper containing kinase AZK, −1.99, 0.048) was downregulated.

Ingenuity network analysis

Other software tools of the ingenuity pathways analysis software (Ingenuity Systems) were used to examine the connection within the differentially expressed transcripts independently of canonical pathways. 88 of 151 differentially expressed transcript IDs, which could be mapped in the ingenuity analysis are alignable with networks. In Figure 4, a network of differentially regulated genes detected by reduction of LRRK2 with RNAi around the upregulated genes CDC42, ARHGEF7, and FGF2 is shown. In this pathway, the influence of FGF2 on proteins with G-protein coupled receptor activity in the plasma membrane CHRM3 and CXCR4 is highlighted. In extracellular space, SCG5 (secretory granule neuroendocrine protein 1, −1.86, 0.00007), INHBA (inhibin beta A, 1.71, 0.002), GREM1 (gremlin 1, −1.83, 0.02), and BMP2 are misregulated, all of them directly or indirectly connected with FGF2. The regulation cascades around FGF2 are also associated with genes in the cytoplasm like the upregulated gene ARHGEF7. The upregulation of this guanidine exchange factor is in concordance with the upregulation of CDC42, which is in connection with the upregulation of another guanidine exchange factor ARHGEF9 (Cdc42 guanine nucleotide exchange factor 9, 1.58, 0.048). Additionally, ARGHEF12, which is in connection to CDC42 via the unregulated gene Ras, is also upregulated. The network is also characterized by the regulation of two different kinases. The rho-interacting, serine/threonine kinase 21 (CIT, 1.52, 0.002) is upregulated and in relationship to CDC42. The MAPKKK ZAK which is connected over the unregulated JNK with CDC42 is downregulated. The prominent key player of this network CDC42 is in dependency to TP53. We could establish a network around TP53 (Fig. 5) highlighting seven downregulated genes (MGMT, RRM2B, MPHOSPH9, ING3, NDN, ZMAT3, TDG) and seven upregulated genes (NQO1, MAP4, PDE4B, CSNK1G1, CCDC99, CHD3, CDC42). Accessorily to the upregulation of NQO1, its expression regulating gene NFE2L3 (nuclear factor erythroid-derived 2-like 3, 1.64, 0.001) is upregulated too. The same effect could be seen by MAP4 (microtubule-associated protein 4, 4.24, 0.003) and its interaction partner SEPT6 (septin 6, 1.66, 0.002). Additionally, to PDE4B (phosphodiesterase 4B, 1.76, 0.002) there are two more differentially regulated genes with relation to IL4 named PTBP2 (polypyrimidine tract binding protein 2, 2.94, 0.0002) and AUH (AU RNA binding protein/enoyl-Coenzyme A hydratase, −1.58, 0.001).

Fig. 4
figure 4

Network 1 of differentially regulated transcripts with the key players FGF2 and CDC42. The gene regulation network, established with the ingenuity pathways analysis software contains 20 genes with 11 genes being upregulated (red icons) and 4 downregulated genes (green icons). Five genes are implemented in the network to complete the interactions. The fold change of the regulation in the microarray analysis as well as the p value of the significance analysis is listed below the gene symbols

Full size image
Fig. 5
figure 5

Network 2 of differentially regulated transcripts with the key player TP53. The gene regulation network, established with the ingenuity pathways analysis software contains 25 genes with 11 genes being upregulated (red icons) and 8 downregulated genes (green icons). Six genes are implemented in the network to complete the interactions. The fold change of the regulation in the microarray analysis as well as the p value of the significance analysis is listed below the gene symbols

Full size image

Validation of the microarray experiment

We choose 16 key players of the interaction networks for independent verification by qRT-PCR. These transcripts are in relation to each other as demonstrated in a validation pathway (Fig. 6). We analyzed the transcript level of SDHA and PDHB with qRT-PCR and identified PDHB as the best reference gene with the GeNorm algorithm. Therefore, transcript levels of PDHB were used for calculation of relative expression of the target genes. We could validate 14 of 16 genes, except of CUL3 and MYC. So the results of the qRT-PCR (Fig. 7) showed 87.5% validation efficiency in comparison to the expression data of the microarray experiment. The microarray expression of MYC was not significantly impaired in LRRK2 repressed cells, but with qRT-PCR measurement, we found a minor reduction of this gene. Examining the expression of FGF2, we detected a small upregulation in the microarray analysis and were also able to validate this in qRT-PCR although with low value.

Fig. 6
figure 6

Network of differentially expressed transcripts used for validation in qRT-PCR. The gene regulation network, established with the ingenuity pathways analysis software contains 24 genes with 12 genes being upregulated (red icons) and 4 downregulated genes (green icons). Eight genes are implemented in the network to complete the interactions. The fold change of the regulation in the microarray analysis as well as the p value of the significance analysis is listed below the gene symbols. Because of complexity only some of the known gene to gene interactions are shown. 16 of the shown transcripts were used for qRT-PCR validation

Full size image
Fig. 7
figure 7

qRT-PCR results of array validation. The same type of regulation was observed for all transcripts except CUL3. The signal log ratio (Log2 of relative expression) for the microarray analysis and the qRT-PCR validation is given below the graph for each gene. *p value ≤ 0.05. **p value ≤0.01. ***p value ≤ 0.001

Full size image

Discussion

The biological function of LRRK2, a protein implicated in Parkinson’s disease, is still elusive [23]. RNA interference was performed to repress the LRRK2 mRNA expression in SH-SY5Y cells. We used two different synthetic siRNAs; however, only one siRNA (siLRRK2-1) was able to knock down LRRK2 by 75%. Differences of the LRRK2 mRNA target sequences might explain the discrepancies of siRNA efficiency [24]. The repression of LRRK2 on mRNA did not reflect an unspecific effect of the transfection procedure or siRNA treatment, since control siRNA provoked no effect.

Overall, the microarray analysis resulted in 187 differentially expressed transcripts, characterized by at least 1.5-fold change and a p value <0.05 with 94 transcripts being up-regulated and 93 transcripts being downregulated. Fourteen of sixteen differentially expressed transcripts (87.5%) were confirmed by qRT-PCR, which is in accordance with quotas mentioned in literature [25].

By means of the analysis of the GO categories, the canonical pathways and the networks impaired by the reduction of LRRK2, we could clearly demonstrate the influence in GTPase mediated signal transduction. The upregulated Ras related GTP-binding protein CDC42 is implicated in the regulation of cell morphology, motility and cell cycle progression [26, 27]. It is a key player in five identified canonical pathways and in our established network 1. Through the Rho GTPases like RHO, RAC, and CDC42, the regulation of actin-based motility occurs, which is critical for nervous system development [28]. Previous studies highlighted the role of LRRK2 in regulating neurite process morphology and pointed out that the downregulation of LRRK2 with short hairpin RNA vectors results in a prominent increase in neurite length [29]. GTPases can switch between “inactive” GDP bound and “active” GTP bound conformations. The nucleotide exchange is catalyzed by GDP/GTP exchange factors (GEF), converting the GTPase in a conformation that can interact with downstream effector proteins [15]. The downregulation of LRRK2 results, additionally to the upregulation of CDC42, in the upregulation of three different guanidine exchange factors (ARHGEF 7, 9, 12). Additionally, the guanine nucleotide binding protein GNAO1 as well as RGS20, which has GTPase activator activity, are upregulated. Since LRRK2 has a Roc-GTPase domain [13], its function could be closely related with guanidine exchange factors and GTP binding proteins.

Furthermore, West et al. [30] showed that the LRRK2 GTPase activity functions independently of the kinase activity, but the kinase activity requires the GTPase activity, and both are required for neurotoxicity of LRRK2. They also suggest that LRRK2 requires co-factors for the binding and hydrolysis of nucleotides in the GTPase domain, reinforcing the analysis of ARHGEF 7, 9 and 12 as possible mediators.

Recent studies highlighted that overexpression of LRRK2 with autosomal-dominant mutations leads to disinhibited kinase activity, a reduction in neuron process length and complexity and the accumulation of tau-positive inclusion with lysosomal characteristics [29]. More detailed analysis in the kinase signaling cascades pointed out that mutations linked to familial PD in LRRK2 did not obviously perturb the MAPK pathway [30]. However, we could show upregulation of CHRM3 and downregulation of ZAK (a MAPKKK), both implemented in SAPK/JNK canonical pathways [31]. In addition, CAMKK2 is upregulated participating in MAPKKK cascades [32, 33]. Especially worth mentioning is the upregulation of CIT in our RNAi experiment. This protein has accessorily to its kinase activity a GTPase regulating activity [34] suggesting it could have relevance in the pathways influenced by LRRK2 reduction.

Recent studies discussed the influence of p53 protein level in the degeneration of the nigrostriatal dopaminergic neurons in parkinsonian brain [35] and the impact of the phosphorylation of p53 on cell death in dopaminergic neurons [36]. The protein p53 is known to induce cell cycle arrest or apoptosis by differential transcription of target genes and through transcription independent apoptotic function [37]. By means of our network 2, we demonstrate the effect of the loss of LRRK2 on p53 signaling cascades. We found 12 misregulated transcripts with direct and 7 misregulated genes with indirect contact to TP53. In response to LRRK2 reduction, the cell cycle influencing transcripts CDC42, MAP4, SEPT6, MPHOSPH9, and the apoptosis related transcripts NQO1, CSNK1G1, RAG2, ZMAT3, MGMT, NDN as well as ING3 are effected in their expression level. By means of the supplied analysis, we highlighted the impact of the loss of LRRK2 on p53 signaling, which is in relation to the previously mentioned publications [35–37].

Considering the interaction partners of LRRK2 which were identified by Dächsel et al. [38], four groups of interacting proteins could be found. They are chaperone proteins, proteins linked to cytoskeleton and trafficking, proteins associated with phosphorylation, and others like eukaryotic translation initiation factors. Among them are the chaperone proteins HSPA9 and HSPA8 known as binding partners of p53 [39, 40].

The Protein p53 decreases the expression of HSP90AB1 and HSP90AA1 [41, 42] both likewise interaction partners of LRRK2 [38]. Additionally, p53 can regulate the expression of FGF2 and vise versa [43, 44]. This upregulated growth factor is associated with actin cytoskeleton signaling [45]. It can regulate the expression of VIM [46], which is a potential interaction partner of LRRK2 [38] and also participating in cytoskeleton signaling [47].

Additionally to FGF2, there is an upregulation of MAP4, TPM1, TPM3, TPM4, KIF21B and a downregulation of ATG4A, all of these affecting cytoskeleton signaling. In detail, MAP4 is a positive regulator of microtubule stabilization [48], ATG4A has microtubule binding opportunities [49] and KIF21B can influence microtubule based movement [50]. Interestingly, LRRK2 was recently shown to phosphorylate the actin-binding proteins moesin, esrin, and radixin [51]. The LRRK2 RNAi-regulated TPM proteins found here also have actin-binding activity [52], making them obvious candidate substrates of LRRK2. These results enforce the hypothesis that the biological function of LRRK2 is closely related to actin based motility, which is critical for nervous system development [28]. This theory is supported by the regulation of genes associated with neurogenesis and neurotransmitter catabolism and differentiated genes implemented in axonal guidance, triggered by the downregulation of LRRK2. The downregulated genes of these biological functions are PRIMA1, RPS6KA3, SEMA4F, SLC18A1, NDN, GREM1, and SCG5. There are also six upregulated genes, namely CHRM3, GNAO1, INHBA, FGF2, EFNB2, and CIT.

The loss of LRRK2 also impairs cell cycle signaling cascades shown in the differentially upregulation of CDC42, APBB2, INHBA, SEPT6, FGF2, CIT, and the downregulation of MPHOSPH9, PPP6C, CUL3, NDN, and ZAK. Especially worth mentioning is the upregulation of INHBA, which is in close connection to the upregulated transcript BMP2. Both proteins can increase the expression of BCL2L1 [53, 54], which is an interaction partner of GOSR1 [55]. GOSR1 can interact with the vesicle mediated transport regulating protein SCFD1 [56], which is a potential interaction partner of LRRK2 [38].

The studies of Dächsel et al. [38] additionally highlighted 2 eukaryotic initiation factors as interaction partners of LRRK2, namely EIF2C2 and EIF2C1, both interacting with each other [57]. In the same study Chun and Rana [57] pointed out that EIF2C2 can bind EIF4E and it may increase the expression of FGF2 [58], which is upregulated in our microarray experiment. Furthermore, we found changes in expression level of three eukaryotic initiation factors (EIF4E3, EIF2S3, EIF3S9).

Another known interaction partner of LRRK2 is the E3 ubiquitin ligase Parkin [59]. Although we did not observe parkin misregulated in our siRNA experiment, we got hints that signaling cascades of ubiquitinylation are influenced with the downregulation of BRAP and ATG4A and the upregulation of FBOX3b and RNF165. Additionally, binding of parkin with actin proteins occurs [60], and four differentially regulated genes of our network (TPM1, TPM3, CIT, FGF2) are in relation with actin proteins [34, 61, 62]. This emphasizes the previous discussed potential role of LRRK2 in cytoskeleton signaling.

Clearly, further research has to separate the primary effects of the depletion of LRRK2 from the secondary ones. Nevertheless, our microarray experiment highlights that the loss of LRRK2 influences numerous signaling cascades, most prominently those implicated in neurite growth and maintenance. Consistent with the identification of actin-binding proteins as direct substrates of LRRK2 [51] as well as knock-down studies in rodents [29] and C. elegans [63], our findings suggest that LRRK2 is involved in actin cytoskeleton-related processes, possibly including neurite outgrowth and collapse, synaptic plasticity, and axonal/dendritic transport. Impairments in these central neuron functions may underlie neuronal dysfunction and neurodegeneration in PD.