Comprehensive Genomic Analysis Reveals the Prognostic Role of LRRK2 Copy-Number Variations in Human Malignancies

Genetic alterations of leucine-rich repeat kinase 2 (LRRK2), one of the most important contributors to familial Parkinson’s disease (PD), have been hypothesized to play a role in cancer development due to demographical and preclinical data. Here, we sought to define the prevalence and prognostic significance of LRRK2 somatic mutations across all types of human malignancies by querying the publicly available online genomic database cBioPortal. Ninety-six different studies with 14,041 cases were included in the analysis, and 761/14,041 (5.4%) showed genetic alterations in LRRK2. Among these, 585 (76.9%) were point mutations, indels or fusions, 168 (22.1%) were copy number variations (CNVs), and 8 (1.0%) showed both types of alterations. One case showed the somatic mutation R1441C. A significant difference in terms of overall survival (OS) was noted between cases harboring somatic LRRK2 whole deletions, amplifications, and CNV-unaltered cases (median OS: 20.09, 57.40, and 106.57 months, respectively; p = 0.0008). These results suggest that both LRRK2 amplifications and whole gene deletions could play a role in cancer development, paving the way for future research in terms of potential treatment with LRRK2 small molecule inhibitors for LRRK2-amplified cases.


Introduction
Identification of the PARK8 locus [1] and mutations in the leucine-rich repeat kinase 2 (LRRK2) gene [2,3], located on chromosome 12 (12q12) in familial cases of Parkinson's disease (PD), more than fifteen years ago is considered a game-changing discovery in our knowledge of this yet incurable

Materials and Methods
A curated set of 176 non-redundant studies publicly available in the online cancer genomic database cBioPortal [48][49][50], comprising 46,595 samples from all types of human malignancies, was initially screened, assessing whether LRRK2 was profiled across those publications; 96 studies endured the screening (full list available in Table S1), encompassing 23,796 different samples from all human cancers excluding those of the eye, the peripheral nervous system, and the thymus. Cases without both MUTs and CNVs data were excluded; a total of 14,286 samples from 14,041 patients remained. The dataset was also queried for MUTs alone, comprising 17,454 patients from 94 studies (2 studies excluded for not profiling LRRK2 MUTs), and for CNVs alone, comprising 15,183 patients from 93 studies (3 studies excluded for not profiling LRRK2 CNVs). For survival analysis of the different types of CNVs, (i.e., amplifications and whole gene deletions), the raw data were downloaded from the cBioPortal database and processed with MedCalc Statistical Software version 19.1.3 (MedCalc Software, Ostend, Belgium). Overall survival was calculated using the Kaplan-Meier Estimate; statistical significance was calculated with a Log-rank Test. Statistical significance of cooccurrence of alterations of LRRK2 and other genes was calculated with Fisher Exact test for p-values and with the Benjamini-Hochberg procedure for q-values.

Results
Among 96 non-redundant studies publicly available in the online cancer genomic database cBioPortal [48][49][50], comprising 46,595 samples from all types of human malignancies, the overall observed prevalence of LRRK2 genetic alterations was 761/14041 (5.4%), with MUTs accounting for 585 cases (76.9%; 426 missense, 155 truncating, 3 inframe indels, 1 fusion) and CNVs accounting for 168 cases (22.1%; 123 amplifications and 45 whole gene deletions); eight cases (1.0%) showed both MUTs and CNVs (5 missense and amplification, 1 fusion and amplification, 1 missense and whole gene deletion, 1 inframe indel and amplification), as shown in Figure 1. All these were somatic alterations, and no germinal LRRK2 mutations were present in the dataset. When considering studies with at least 50 patients profiled, alterations across different cancer types ranged from 18.06% (bladder/urinary tract cancer) to 0.0% (seminoma), as shown in Figure 2. Prevalence and distribution of LRRK2 genetic alterations across the analyzed studies. The 5% of the first and second row is expanded to better visualize the data [48][49][50].
When considering studies with at least 50 patients profiled, alterations across different cancer types ranged from 18.06% (bladder/urinary tract cancer) to 0.0% (seminoma), as shown in Figure 2. The distribution of missense, truncating, and inframe mutations across the LRRK2 protein is shown in the lollipop plot in Figure 3. Of note, one case of endometrial carcinoma displayed the missense mutation R1441C; one case of stomach adenocarcinoma showed the missense mutation G2019D. Figure 3. LRRK2 missense, truncating, and inframe mutations in the studies analyzed. A scattered mutational distribution is evident. The alterations are color-coded as shown in the legend above [48][49][50].
Overall survival (OS) analysis of all cancer types pooled together demonstrated a worse prognosis for cases with LRRK2 alterations (207/613 deceased, median OS: 94.03 months) in comparison to cases with unaltered LRRK2 (4123/11,356 deceased, median OS: 103.26 months) as shown in Figure 4; this result was statistically significant, with p = 0.0280. The distribution of missense, truncating, and inframe mutations across the LRRK2 protein is shown in the lollipop plot in Figure 3. Of note, one case of endometrial carcinoma displayed the missense mutation R1441C; one case of stomach adenocarcinoma showed the missense mutation G2019D. The distribution of missense, truncating, and inframe mutations across the LRRK2 protein is shown in the lollipop plot in Figure 3. Of note, one case of endometrial carcinoma displayed the missense mutation R1441C; one case of stomach adenocarcinoma showed the missense mutation G2019D.   By stratifying for organ sites, a similar result was observed for prostate adenocarcinoma (LRRK2 altered: 6/16 deceased, median OS months: 120, vs. LRRK non-altered: 127/649 deceased, median OS: 115.13 months; p = 0.037); interestingly, an opposite, statistically significant trend was noted for endometrial carcinoma (LRRK2 altered: 6/85 deceased, median OS: NA, vs. LRRK non-altered: 110/479 deceased, median OS: 110.10 months; p = 0.0001481; Figure 5). No other significant results for other organ sites were noted, with trends pointing toward a better/worse prognosis in different settings ( Figure S1). When analyzing MUTs and CNVs alone, a significant (p = 0.0005102) worse survival was noted within the CNV-altered cases (52/111 deceased, median OS: 55.47 months) in contrast to CNVunaltered ones (4799/13198 deceased, mean OS: 106.57 months); no difference was noted in the cohort analyzed for MUTs only (827/17454; mutated cases: 210/597 deceased, mean OS: 84 months; nonmutated cases: 3711/10945, median OS: 78.8 months; p = 0.870; Figure 6). The majority of CNV-altered By stratifying for organ sites, a similar result was observed for prostate adenocarcinoma (LRRK2 altered: 6/16 deceased, median OS months: 120, vs. LRRK non-altered: 127/649 deceased, median OS: 115.13 months; p = 0.037); interestingly, an opposite, statistically significant trend was noted for endometrial carcinoma (LRRK2 altered: 6/85 deceased, median OS: NA, vs. LRRK non-altered: 110/479 deceased, median OS: 110.10 months; p = 0.0001481; Figure 5). No other significant results for other organ sites were noted, with trends pointing toward a better/worse prognosis in different settings ( Figure S1). By stratifying for organ sites, a similar result was observed for prostate adenocarcinoma (LRRK2 altered: 6/16 deceased, median OS months: 120, vs. LRRK non-altered: 127/649 deceased, median OS: 115.13 months; p = 0.037); interestingly, an opposite, statistically significant trend was noted for endometrial carcinoma (LRRK2 altered: 6/85 deceased, median OS: NA, vs. LRRK non-altered: 110/479 deceased, median OS: 110.10 months; p = 0.0001481; Figure 5). No other significant results for other organ sites were noted, with trends pointing toward a better/worse prognosis in different settings ( Figure S1). When analyzing MUTs and CNVs alone, a significant (p = 0.0005102) worse survival was noted within the CNV-altered cases (52/111 deceased, median OS: 55.47 months) in contrast to CNVunaltered ones (4799/13198 deceased, mean OS: 106.57 months); no difference was noted in the cohort analyzed for MUTs only (827/17454; mutated cases: 210/597 deceased, mean OS: 84 months; nonmutated cases: 3711/10945, median OS: 78.8 months; p = 0.870; Figure 6). The majority of CNV-altered    The co-occurrence of CNV alterations of LRRK2 and other genes located at 12q12 is shown in Table 1. The co-occurrence of CNV-alterations of LRRK2 and other genes involved in cancer pathogenesis located at 12q13-12q15 is shown in Table 2.   The co-occurrence of CNV alterations of LRRK2 and other genes located at 12q12 is shown in Table 1. The co-occurrence of CNV-alterations of LRRK2 and other genes involved in cancer pathogenesis located at 12q13-12q15 is shown in Table 2. The co-occurrence of CNV alterations of LRRK2 and other genes located at 12q12 is shown in Table 1. The co-occurrence of CNV-alterations of LRRK2 and other genes involved in cancer pathogenesis located at 12q13-12q15 is shown in Table 2. No significant clustering of LRRK2 CNV-altered versus LRRK2 CNV-unaltered cases was observed in terms of neoplasm staging ( Figure S2). Sex and age at diagnosis distribution for LRRK2 CNV-altered and LRRK2 CNV-unaltered cases are shown in Figure S3; both results were statistically non-significant (p > 0.05).

Discussion
To our knowledge, this is the first study to assess the prevalence, characteristics, and prognostic implications of LRRK2 alterations in a large cohort of human malignancies. A significant proportion of cancers (5%, Figure 1) across the majority of tumor subtypes (Figure 2) harbor MUTs and/or CNVs of this gene involved in the pathogenesis of PD; however, the role of LRRK2 in cancer development remains to be elucidated, and the possible driver or passenger role of LRRK2 alterations in cancer needs yet to be established.
The mutational landscape of LRRK2 in this cohort suggests a passenger role of MUTs, given their scattered distribution and absence of clustering in specific domains such as Pkinase and Roc (Figure 3). It is also worth noting that the two most commonly encountered mutations of LRRK2 in PD (and therefore pathogenic, albeit in a different setting of human disease), i.e., G2019S and R1441C, are seldom encountered (R1441C: n = 1, G2019S: n = 0, with a case showing a somatic mutation at the same codon: G2019D). An explanation could be represented by a hypothetical tumor-suppressor role of LRRK2 in normal cellular development, for which a multitude of different inactivating mutations could result in a decrease of LRRK2 levels and activity; however, this hypothesis is not supported by previously published data [33] and our subsequent analysis.
The negative prognostic implications of LRRK2 alterations in the studies analyzed ( Figure 4) seem to suggest, on the other hand, that LRRK2 does not simply represent a bystander during the process of cancer pathogenesis; the conflicting results for prostate adenocarcinoma and endometrial carcinoma also seem to indicate a site-specific biological meaning of such alterations. By querying for MUTs and CNVs alone, the two different Kaplan-Meier plots unravel the negative prognostic role of CNV-altered cases, while the curves for mutated/nonmutated cases almost completely overlap.
By stratifying for the type of copy-number alteration, in comparison to unaltered cases, both whole gene deleted cases (CNV = −2) and amplified cases (CNV = 2) show a worse prognosis, the former being more marked. Notably, LRRK2 amplification has been observed to be of indirect oncogenic potential in papillary renal cell carcinoma and thyroid carcinomas trough MET signaling [39].
Given the nature of the alterations which confer a negative prognostic value to LRRK2, i.e., amplifications and deletions, other genes located within or near the locus 12q12 need to be taken into account when trying to explain a putative role of LRRK2 in cancer development, since duplications and deletions can involve large segments of DNA strands. The list of genes located at 12q12 and their frequency of CNV alterations in LRRK2 CNV-altered and CNV unaltered cases is shown in Table 1; notably, all genes are significantly co-altered along with LRRK2 (percentage of co-alterations range: 5.81-57.56%; percentage of alterations in LRRK2 unaltered cases range: 0.07-0.54%; Log ratio range: 5.21-9.56). By querying the OncoKB database [51,52] for genes located at 12q12, only ARID2 deletion was found to be likely oncogenic, specifically in non-small cell lung cancer [53] and hepatocellular carcinoma [54]; in our analysis, ARID2 whole gene deletion co-occurred in 14.53% of LRRK2 CNV-altered cases.
CNVs of other genes located at 12q12-12q15 are known to be oncogenic, such as KMT2D deletions [55], ERRB3 amplifications [56], CDK4 amplifications [57], and MDM2 amplifications [58]. Correlations between CNV of those genes and LRRK2 in our analysis are presented in Table 2. Notably, amplifications of MDM2 and CDK4 are both associated with liposarcoma [57,58]. Given the oncogenic role of amplification of ERRB3, CDK4, and MDM2, it could be speculated that LRRK2 represents only a passenger of 12q amplification; it must be noted, though, that the major impact on prognosis is observed in LRRK2 whole gene deleted cases, and the concordance between LRRK2 and KMT2D whole gene deletions is 4.07% in our analysis, and a significant proportion of LRRK2-amplified cases lack amplification of MDM2, CDK4, and ERBB3. Besides, MDM2 acts as a down-regulator of p53 by initiating its ubiquitination which leads to proteasomal degradation [59], and LRRK2 has been described to phosphorylate p53 [36], therefore protecting it from MDM2-induced degradation; in this view, the 20.93% of cases harboring both LRRK2 and MDM2 amplifications could not rely on the latter as a driver of tumor progression. The same applies to the 15.12% of CKD4-amplified cases, as LRRK2 induces p21, a known CKD4 inhibitor [60], via p53 phosphorylation [36].
The hypothesized role of an increase in kinase activity of LRRK2 in cancer development, if confirmed via in vitro studies, potentially paves the way for the use of small molecule LRRK2 inhibitors beyond PD: LRRK2-amplificated cases are putative candidates for such treatment. Our analysis indeed shows that LRRK2 amplification, given its negative prognostic significance, could play a role in cancer development and may be clinically actionable. And yet, we also note a dismal prognosis for LRRK2 whole gene deleted cases. In this view, an inhibition of LRRK2 activity could prove to be more harmful than beneficial. However, when considering that both amplifications (gain of function) and whole gene deletions (loss of function) of LRRK2 are associated with a worse prognosis in cancer, it must be acknowledged that the modulation of the kinase activity by LRRK2 small molecule inhibitors in G2019S-mutated PD patients might lead to a normal kinase function rather than a loss of kinase activity. However, given the complex interplay between the different domains of LRRK2, an inhibition of its kinase domain could have pleiotropic effects on the function of the whole protein [61,62]. Further research is needed, both in vitro and in vivo, in order to better characterize the potential anticancer effects of LRRK2 small molecule inhibitors.
Our analysis provides insights on the potential biological balancing role of LRRK2, a fact hinted at by the different biological processes in which this gene is involved. A number of other genes have been described to possess both oncogenic and tumor-suppressor functions [63]. A subset of these genes possesses kinase activity (e.g., MAP2K4, MAP3K4, PRKAR1A, PRKCB), similarly to LRRK2. In this view, any CNV alteration could play a role in the disruption of normal cellular homeostasis which ultimately leads to cancer. Notably, a small molecule LRRK2 inhibitor, LRRK2-IN-1, demonstrated activity in colorectal and pancreatic cancer via direct inhibition of DCLK1 [64]; the clinical utility of such therapeutic agents could also spawn further from LRRK2-overexpressed cases.

Conclusions
Herein, we demonstrated that a significant subset of human malignancies shows LRRK2 genetic alterations. The most frequently encountered mutations in PD, G2019S, and R1441C are almost nonexistent in sporadic human malignancies. LRRK2 CNV, both amplifications and whole gene deletions, confer a poorer prognosis in terms of OS in comparison to unaltered cases; the latter cases show the worse prognosis. In contrast, LRRK2 somatic MUTs show no prognostic significance. These data support the fact that both LRRK2 overexpression and complete loss-of-function could play a role in cancer development, and potentially pave the way for future research to investigate the potential treatment of amplified cases with LRRK2 small molecule inhibitors; the dismal prognosis of LRRK2 whole gene deleted cases also needs further research. In contrast, LRRK2 non-activating MUTs are of probable passenger significance.