Identifying Drug Repurposing Candidates for CLN3 Targeting Proteomics Expression Profile

Abstract

Juvenile neuronal ceroid lipofuscinosis (CLN3) is a rare neurodegenerative disorder lacking effective therapies. This study aimed at developing a drug repurposing approach to identify potential therapeutic candidates for CLN3 using its protein expression profile (CPEP) constructed from proteomics data. Differentially expressed proteins were identified and applied to query the iLINCS database, resulting in 60 FDA-approved drugs with reversal effects on CPEP. These candidates were further prioritized based on regulation strength, coverage, and blood-brain barrier permeability. Top candidates include Vorinostat and Cyclosporine, which have shown promise due to their significant regulation scores and blood-brain barrier permeation probability. These results provide opportunities for further investigation on novel therapies for CLN3.

I. Introduction

CLN3, also known as juvenile neuronal ceroid lipofuscinosis, is a progressive neurodegenerative disorder characterized by the accumulation of lipopigments in tissues, leading to cognitive decline, seizures, and vision loss [1]. The prevalence of CLN3 is estimated to be 2-5 in 100 000, classifying it as a rare disease with limited public awareness and research focus [2]. Currently, treatments are primarily symptomatic and supportive, including seizure management and physical therapy, and no cure or disease-modifying therapy is available [3].

Drug repurposing (DR), or drug repositioning, involves identifying new therapeutic applications for existing drugs and has emerged as a promising strategy for drug development for its time and cost-effectiveness [4]. In this study, we identified and prioritized potential repurposing candidates for CLN3 by targeting its protein expression profile.

II. Methods

In this study, we constructed a CLN3 protein expression profile (CPEP) using proteomics data from CLN3 patients. We then identified drug candidates that could induce a reversal in this expression profile and prioritized them based on their regulation strength against the CPEP.

A. Data preparation

Proteomics data were collected using Mass Spectrometry (MS) from CLN3 participant enrolled in an IRB-approved prospective natural history study (NCT03307304). The proteomic data used in this study were obtained from a previous publication [5], where potential protein biomarkers were identified for CLN3 disease.

B. Differential expressed protein analysis

Differentially expressed (DE) protein were calculated from proteomics data using the Benjamini-Hochberg (B-H) adjusted p-value < 0.05 and a fold change > 1.5 as the cutoff. The CLN3 protein expression profile (CPEP) was then constructed with log2 transformed fold change (LFC) of those DE proteins. The R packages stats [6], EnhancedVolcano [7], and ggplot2 [8] were utilized for DE analysis and visualization.

C. Drug repurposing candidate identification

The CPEP was applied to search against the iLINCS database [9] for experimental perturbagen (investigational chemicals and drugs) that demonstrated reversal effects. Perturbagens with a correlation coefficient < −0.2 with CPEP were selected. Among them, FDA-approved drugs [10] (published on July 30th, 2024) were then selected as repurposing candidates.

D. Drug repurposing candidate prioritization

Gene expression signatures of the candidates were retrieved from the iLINCS via its API [11] using R packages knitr [12], tinytex [13], httr [14], jsonlite [15], htmltools [16], and Biobase [17]. In our previous study [5], we defined two scores, the regulation score (RS) and overall coverage (OC) to measure the regulation strength of each candidate. The RS measures the extent of regulation based on the divergence between the CPEP and drug response signatures, while the OC represents the percentage of CPEP genes whose expression is reversed by the candidate drug. Additionally, each candidate was evaluated for its blood-brain barrier (BBB) permeation probability via literature review. We prioritized the candidates by their regulation score and selected the top five with favorable BBB permeation potential, excluding those with poor BBB permeability.

III. Results

A. DE analysis and CPEP construction

Out of 3,319 proteins in the proteomics dataset, there are 162 DE proteins, including 8 upregulated and 154 downregulated proteins. A volcano plot discloses these 162 proteins exhibiting differential expressions at both transcriptional and translational levels (Figure 1). Subsequently, we constructed the CPEP using the LFCs of these 162 protein expression in CLN3.

Figure 1.

Volcono plot of DE proteins in CLN3. Each dot stands for one protein’s expression, with its log2-transformed fold changes and log10 B-H adjusted p-value in the X and Y axes, respectively. Red dots stand for DE proteins used to construct the CPEP.

B. Drug repurposing candidate identification

As shown in Table 1, 329 drug-induced gene expression signatures were identified from four libraries in iLINCS by applying the predefined Concordance cutoff, calculated between the CPEP and the drug response signatures. These signatures were derived from perturbation experiments of 207 chemicals, including 60 FDA-approved drugs.

Table 1.

Results on Candidates identification from ilincs

		Signatures	Chemicals	Drugs
Signature Libraries	Cancer therapeutics response	106	88	14
	Connectivity Map	6	6
	DrugMatrix	131	98	42
	Pharmacogenomics transcriptional	86	15	7
Total		329	207	63
Unique		-	207	60

The column of Signatures lists the number of signatures identified from each signature library. The column of Chemicals lists the number of chemicals tested in these signatures. The column of Drugs denotes the number of FDA-approved drugs identified accordingly. The row of Total denotes the numeric sum of signatures, chemicals, and drugs identified from all libraries, while the row of Unique lists the unique numbers of chemicals and drugs.

We then plotted a heatmap displaying the reversal expression effect of 60 FDA-approved drugs on CPEP (Figure 2). The candidates were clustered based on the similarity to the expression of CPEP. As illustrated in the heatmap, some candidates, such as Vorinostat and Decitabine (highlighted in Figure 2), demonstrate stronger reversal effects on more CPEP proteins than other candidates.

Figure. 2.

Heatmap of reversal effect on CPEP for 60 drugs. Each row corresponds to a gene in CPEP while each column corresponds to one FDA-approved drug. The bottom row is the LFCs in CPEP in ascending order. Gray area indicates missing values.

Based on the calculated scores of RS, OC, and the BBB permeation probabilities, we obtained the top five repurposing candidates shown in the Table 2. The candidates with high RS, however associated with poor BBB permeation ability, were excluded from our candidate list, such as Decitabine and Dacarbazine.

Table2.

Top Ranked Candidates

Drug	RS	OC	BBB	Indication
Vorinostat	1.02	0.728	0.9861	cutaneous manifestations in patients with cutaneous T-cell lymphoma
Cyclosporine	0.801	0.506	0.9659	prophylaxis of organ rejection
Ifosfamide	0.687	0.574	0.9736	chemotherapy
Vandetanib	0.678	0.456	0.9596	symptomatic or progressive medullary thyroid cancer
Regorafenib	0.637	0.5	0.851	metastatic colorectal cancer

RS: regulation score. OC: overall coverage. BBB: Blood-Brain Barrier permeation probability. Indication: FDA-approved indication. All of these candidates had not been studied as repurposing candidate for CLN3.

IV. Discussion

In this study, CLN3 protein expression profile (CPEP) was constructed by using patient proteomics data and applied to identify 60 FDA-approved drugs with potential reversal protein expression effects. The top candidates were prioritized based on their regulation strength, coverage, and blood-brain barrier permeability, with Vorinostat emerging as a promising repurposing option. Vorinostat is currently FDA-approved for treating cutaneous manifestations in patients with cutaneous T-cell lymphoma and has a high blood-brain barrier permeability. Though it has not been directly studied for CLN3 disease, its effects on neurodegenerative disorders [18], suggest it may have potential in addressing CLN3 pathology. Future studies should focus on validating these top candidates’ efficacy in in vitro and in vivo assays, with the ultimate goal of accelerating the drug development of CLN3.