Kaposi’s sarcoma (KS) is a multifocal hyperproliferative neoplasm of the vascular or lymphatic endothelium. KS lesions may also occur on the skin surface of the upper and lower extremities or on the visceral organs like the lung and spleen [1-4]. Four epidemiological subtypes (Classic, Endemic, Epidemic, and Iatrogenic) of KS have been described so far and Kaposi’s sarcoma-associated herpesvirus (KSHV), also known as human herpesvirus-8 was identified as the primary etiologic agent of these subtypes [5-7]. Besides, KSHV causes two lymphoproliferative diseases and an inflammatory syndrome: primary effusion lymphoma, multicentric Castleman disease, and KSHV inflammatory cytokine syndrome [8-10].

KSHV has the highest prevalence in sub-Saharan Africa, the intermediary prevalence in the Mediterranean, and low prevalence in northern Europe, North America, and Asia [4]. Previously, KS was 20,000 times more persistent in acquired immune deficiency syndrome (AIDS) patients than in the general population [11]. With the advent of combination antiretroviral therapy, the incidence of AIDS-related (epidemic) KS substantially decreased in developed countries [12]. But the rapid progression of AIDS and inaccessibility to antiretroviral drugs worsened the condition in several African countries [13]. Countries like Malawi, Uganda, Zimbabwe, and Swaziland have experienced a multitude of KS incidences since the onset of the AIDS epidemic in the early 1980s, making KS the most common cancer in males and the second most common cancer in females [14,15].

The KSHV genome has approximately 140.5 kb long unique region (LUR) flanked by 25‒30 kb of direct terminal repeats [16,17]. The LUR comprises 90 open reading frames (ORFs) that separately express during the latent or lytic phase of viral infection [18]. Though ORF50/Rta (replication and transcription activator) initiates the entire KSHV lytic cycle, its completion depends on ORF57 that ensures the orderly expression of other early and late lytic genes [19]. The ORF57/Mta (mRNA transcript accumulation) is a 455 amino acid long nuclear protein with no substantial sequence homology to any known cellular proteins [20]. This post-transcriptional regulator plays several vital roles in KSHV pathogenesis, for example, binding and stabilization of intron-less transcripts, facilitates intron splicing, nuclear accumulation of coding/non-coding RNAs, cytoplasmic accumulation of mRNAs, promoting protein translation, etc. [21]. Disruption of ORF57 showed to interrupt the KSHV lytic gene expression and eventually halt virion production [22]. Therefore, it could be a potent target for therapeutic intervention to suppress the replication of KSHV in the host cells.

RNA interference is a unique molecular therapeutic technique in which specific non-coding RNAs, such as siRNA or miRNA, silence or downregulate the target mRNA in a sequence-specific manner. Although siRNA and miRNA share a common pathway, they have distinct functions, and in many cases, siRNA outcompeted miRNA in silencing activity. Since siRNA designing allows more sophistication in reducing off-target effects, it can inhibit one particular target mRNA. Besides, modern transfection techniques such as lipofection and nanoparticles have greatly simplified its delivery into cells [23]. siRNA-based treatment has been successfully demonstrated in influenza, hepatitis C, severe acute respiratory syndrome coronavirus 2, cancer, different autoimmune, and neurodegenerative diseases [24-28]. Hence, considering the clinical significance of KSHV, the role of ORF57 in viral pathogenesis, and the therapeutic potentiality of RNA interference, we aimed in the present study to design effective siRNA(s) against ORF57 of KSHV.

The overall pictorial representation of this study is depicted in Fig. 1 in a stepwise manner.

There are several algorithms, standalone tools, and web servers available for designing siRNA against target sequences, each with its own set of benefits and drawbacks [55]. These tools consider a diverse array of mechanisms and features to screen out the best possible siRNAs, and yet none of them can adopt the all-in-one strategy. That’s why a somewhat manual hybrid approach was employed in this study. We utilized a combination of different algorithms (Reynolds, Amarzguioui, Ui-Tei, i-Score, and Sfold) to acquire siRNAs that met most of the prerequisites to be an effective gene silencer. For example, Reynolds, Amarzguioui, Ui-Tei algorithms are some of the most widely used rule-based algorithms, and many tools implement these algorithms for initial siRNA prediction [34,41,60,61]. Second-generation algorithms like i-Score, s-Biopredsi, and DSIR implement different machine-learning methods to predict the most promising siRNAs with almost equivalent accuracy [35,60]. On the other hand, Sfold uses a statistical algorithm-based probability profile approach to predict siRNA accessible regions in the secondary structure of target mRNA [36]. While other algorithms perform calculations based on 19-mer siRNA sequences, s-Biopredsi and DSIR consider 21-mer for their analysis [35]. That’s why from the second-generation category, only the i-Score algorithm was taken into account to ensure uniformity in siRNA length and computation. Only those siRNAs were picked from all of the predicted siRNAs that met every key criterion of these five methods. Furthermore, the full conservation of these siRNA target sites was also verified to ensure that our proposed siRNAs are effective against all KSHV strains.

Previously it was reported that siRNA sequences with “GUCCUUCAA” and “UGUGU” motifs can be immune-stimulatory. For example, the “UGUGU” motif induces IFN type 1 and causes downregulation of nonspecific genes [62,63]. Another experiment demonstrated that monocytes are stimulated to generate large amounts of IL-12 when the CpG motif "AUCGAU" is present in RNA oligonucleotides along with a poly-G tail [64]. The existence of the "UGGC" motif in the siRNA guide strand has been found to decrease cell viability [65]. The siRNAs which contained these sorts of motifs were omitted in this phase of screening.

After the conservancy and motif filtration analysis, both the sense and antisense of the selected siRNAs were investigated for the presence of off-target human gene and miRNA seed regions. The siRNA-targeted sequence of viral mRNA should not be identical or nearly identical to any human gene, otherwise, it will create an undesired silencing effect [66]. Off-target effects can be induced by either strand of siRNA (sense or antisense) [45]. Furthermore, siRNA guide strands with seed regions identical to the human miRNA were sieved out in the second phase of off-target filtration. Because miRNA-like off targeting may induce mRNA translation inhibition, which can also result in gene-nonspecific downregulation [67].

For effective siRNA activity, target sites should not be located in SNP sites, and/or in between the first 75 bases of mRNA's start codon or in intron sites [55]. Since all of the target sites for the candidate siRNAs of this study are fully conserved across all mRNA sequences, it can be assured that these target sequences are not in SNP sites. None of the selected siRNAs' target sites are within the first 75 bases of mRNA. Finally, the use of only coding sequences in this analysis eradicated any possibilities of the intron sequences being found in target sites (Table 2). GC content of siRNA has a significant impact on its silencing efficacy, and our suggested siRNAs have GC content in the recommended range [55].

The performance of siRNA is influenced not only by sequence features but also by structural and thermodynamic properties [68]. The incorporation of RNA strands in the RISC complex can be influenced by the thermodynamics of the siRNA duplex. More specifically, RNA strands having lower binding stability at the 5' end of the guide strand are preferred to be integrated into the RISC. Furthermore, the thermodynamic features of nucleotide base-pairing between the siRNA guide strand seed region and off-target mRNAs are primarily responsible for its off-target silencing activity [69]. Therefore, after selecting the desired siRNAs through a stringent filtration procedure, their thermodynamic attributes were also evaluated. The existence of hairpin structures in target mRNA sites can be anticipated by calculating the internal melting temperature (Tm) of the sense strands. Especially, if the Tm value of the siRNA sense strand is greater than 60℃, there is a high probability of hairpin formation, which will eventually reduce the knockdown ability [70]. None of our suggested siRNAs exceed this threshold value of Tm (Table 3).

One of the most essential factors for implementing siRNA target prediction algorithms is the computational estimation of RNA-RNA binding interaction. This RNA-RNA binding interaction is indicated by the free energy of binding (∆G) between a specific siRNA and its target mRNA. In case of weak duplex (mRNA-siRNA) stability, the RISC would not have enough opportunity to cleave the target mRNA. Contrarily, the passenger strand or target mRNA will separate slowly after cleavage by the RISC complex if the siRNA forms a very stable duplex with its antisense strand. All of the finalized siRNAs in this study have free energy values of heterodimer binding (∆G) in between the recommended range (Table 3).

Another important indicator of the strength of base-pairing in a seed duplex complex is the frequency of MFE structure in the ensemble of secondary structures. A high frequency of MFE structure in the ensemble is correlated to a stiff seed duplex binding [71]. The accessibility of siRNA for binding, which can be projected by its folding free energy change, is found to be highly correlated with its efficacy for hybridization with target mRNA [72]. According to prior findings, an RNA molecule with the positive free energy of folding has a greater likelihood of binding to a target site because it will be more accessible [73]. Furthermore, intra-oligomer binding within the antisense strand of siRNA can significantly impede its accessibility for target mRNA [72]. That's why potential folding structures of selected siRNA guide strands were also predicted in this analysis (Fig. 2). Both siRNAs reported in this investigation (siRNA_1 and siRNA_4) appeared to have a high frequency of MFE structure and positive free energy of folding (Table 3).

Aside from choosing a siRNA guide strand, predicting the secondary structure of the target mRNA is also very imperative for RNAi activity. The previous evidence indicated that the secondary structure of complete mRNA should be postulated, as nucleotides distal from the target site can also modify its (target site's) structure [74]. But the number of potential secondary structures increases exponentially as the length of the sequence increases, making reliable mRNA secondary structure prediction challenging [72]. For that reason, the RNAalifold server, which predicts the consensus secondary structure of mRNA from MSA among several related sequences, was used in this experiment. Target mRNA with unpaired regions at either 5’-end or 3’-end can be silenced more effectively than a fully paired target [75]. As shown in Fig. 3, the target sites of our selected siRNAs contain an unpaired 5' or 3' end.

ORF57 of KSHV plays a vital role in viral lytic replication. Dimerization of ORF57 stabilizes the protein structure and is crucial for its functional activity. Each monomer of ORF57 homodimer is encoded from a single gene and consists of two distinct domains: intrinsically disordered N-terminal domain (NTD) with no defined structural motifs (amino acid residues 1‒166), and helix-rich C-terminal domain (CTD) (amino acid residues 167‒455) [76]. The NTD harbors three nuclear localization signals (NLSs) and interacts with several cellular factors [20,77]. The CTD has an N-terminal ‘arm’ stretching from residues 167 to 222 and a C-terminal globular domain (aa residues 223‒455) having a conserved zinc-binding motif. These three structural elements (arm, globular interface, and zinc-binding motif) are equally significant for ORF57 dimerization as the ‘arm’ region from one monomer docks on the globular surface of a neighboring monomer in an antiparallel fashion, whereas the C-terminus end (aa residues 445‒454) is locked into the globular domain of the same monomer [76].

Interestingly, both the siRNA_1 and siRNA_4 target regions are located in the CDS of N and C-terminal domains of ORF57, respectively. Three NLSs are found in the NTD and the siRNA_1 target sequence (aa residues 98-104) overlaps with the NLS1 (aa residues 101‒107). A previous study demonstrated that disruption of NLSs inhibits the nuclear translocation process of ORF57 [77]. Our designed siRNA_1 targets the NLS1 coding region and could potentially disturb the translation process of the NTD CDS of ORF57. Besides, ORF57-CTD facilitates the dimerization, stability, and function of the protein. Functional studies also revealed the dissociation of ORF57 dimer upon deletion/point mutation of either one of three structural elements as mentioned earlier. siRNA_4 designed in this study falls within the CDS of the globular interface (aa residues 234‒239) of ORF57-CTD. This globular structure maintains the electrostatic interaction with the interface residues of CTD of adjacent monomer to stabilize the dimer [76]. Thus, our proposed siRNA_4 could also inhibit this function by degrading the target mRNA.

It is well established that siRNA-induced post-transcriptional gene silencing starts with the assembly of the RISC [78,79]. The mRNA molecule is cut exactly by cleaving the phosphodiester bond between the target nucleotides which are paired to siRNA residues [80]. The functional ORF57 protein is a dimer of two identical subunits. This indicates the subunits of ORF57 protein are translated from the transcripts of a single copy gene. Hence, the disruption of the ORF57 transcripts by siRNAs could stop the translation process completely. Therefore, we believe that either siRNA_1 or siRNA_4 could be sufficient enough to suppress the ORF57 gene expression completely or at least partially.

There is currently no specific treatment for KSHV-related diseases. The treatment of choice for KSHV patients predominantly depends on various parameters, like the tumor location, a variant of KS, rate of progression, distribution of the lesions, the severity of the symptoms, and immune competence [81]. Although medicines such as rituximab, acyclovir, and others are currently used to treat KSHV-related complications, these are not specific therapies [82,83]. Besides that, severe side effects like kidney toxicity, neutropenia, and neurotoxicity have made the treatments more challenging [84]. That’s why additional studies are required to explore new drugs for KSHV associated diseases. In this circumstance siRNA-based therapy might be a viable alternative as the inhibitory effect of siRNA on different herpes virus replication has already been reported in several studies [85-87]. This research is such an effort to accelerate the discovery of new treatments for KSHV-related diseases. Two potential siRNAs have been screened in this study through a series of comprehensive filtration steps that will hopefully inhibit the translation of the ORF57 gene in KSHV. Since our suggested siRNAs meet all of the requirements for an effective siRNA, it can be expected that they'll be able to inhibit the infection against all KSHV strains. But, as this selection method was entirely based on computational prediction, proper in vitro and in vivo validation is albeit necessary.