Effect of sulfasalazine on human neuroblastoma: analysis of sepiapterin reductase (SPR) as a new therapeutic target

We have previously reported on the role of SPR in NB proliferation [32], where we demonstrated a deleterious effect of RNAi-mediated SPR expression knockdown in the MYCN2 NB cell line. We also showed that high SPR mRNA expression was correlated to poor patient prognosis in Kaplan-Meier analysis in the Versteeg-88 NB dataset in the public domain. We now present SPR mRNA expression analysis on all 12 NB cohorts in the public domain (Table 1). We find that high SPR expression is significantly correlated in all four NB cohorts annotated for patient survival and/or prognosis. While in our previous study [32] we could only show a trend for a correlation between SPR expression and tumor MYCN gene amplification in the Versteeg-88 set (P = 0.06), we can now state that SPR expression is significantly higher in patients with tumor MYCN gene amplification in 6 of 8 datasets with MYCN amplification annotation. Considering the different compositions of these datasets with respect to patient age, MYCN amplification, and INSS stage, together with the different array platforms used for the generation of these data, this is a very robust finding. In Fig. 2, we show the results for the largest NB cohort in the public domain, the Kocak-649 dataset. Although this dataset does not contain survival data, the correlations between SPR expression and three important clinical NB parameters are highly significant (Fig. 2, a-c): age at diagnosis (P = 1.9 · 10⁻²³, MYCN tumor amplification (P = 7.9 · 10⁻¹⁵, and INSS stage (various P values < 0.05). In addition, the Kocak-649 dataset shows a significant correlation between SPR and ODC mRNA expression (Fig. 3, R = 0.225, P = 6.5 · 10⁻⁹). This association, although highly significant, has a relatively low R value. However, since we previously found a similar association (R = 0.289, P = 6.2 · 10⁻³) in the Versteeg-88 cohort [32], we felt strengthened in our argument that this correlation is meaningful.

These results show that SPR mRNA expression is highest in all NB clinical groups with poor outcome: high age at diagnosis, tumors with MYCN oncogene amplification, and patients with high INSS tumor stage. Its expression pattern therefore resembles that of ODC, and indeed we found a tentative correlation between SPR and ODC expression. Together, these results prompted us to investigate the specific targeting of SPR alone or together with targeting of ODC as novel NB therapy.

A recent study by Chidley et al. revealed that SSZ blocks BH4 biosynthesis through inhibition of SPR [30]. To examine the inhibitory effects of SSZ in NB cells, we treated SK-N-Be(2)c, SK-N-SH, and LAN-5 cells with increasing concentrations of SSZ (0–400 μM) and measured cell viability 48 h after treatment. As shown in Fig. 4, SSZ decreased the cell viability of all three NB cell lines in a dose-dependent manner. We did not observe overt apoptosis (data not shown), suggesting that SSZ inhibits cell proliferation of NB cells without cytotoxic effects.

To investigate potential signaling molecules and pathways involved in SSZ-mediated cell death, we tested the expression levels of several proteins that regulate cell proliferation, including p27^Kip1, retinoblastoma tumor suppressor protein Rb, Akt/PKB, and p44/42 MAPK (Erk1/2). Western blot analysis did not reveal any significant protein expression differences between SSZ-treated and untreated NB cells (data not shown), suggesting that additional, alternative signaling pathways are activated by SSZ.

To examine if SPR binds SSZ, we performed computational docking simulations. SSZ is an amino-salicylate, specifically 5-((4- (2- Pyridylsulfamoyl) phenyl)azo) salicylic acid (Fig. 1). SSZ has one canonical conformer with an MMFF94-minimized (Merck Molecular Force Field) energy of 83.9 kcal/mol, which was used in the docking simulations [33]. Under physiological conditions the molecule carries a negative charge which may have a role in the interaction with the receptor.

The human SPR crystal structure is available in complex with NADP+ in a hexameric assembly (unpublished data, PDB: 1Z6Z). This biologically active, functional form of SPR exists as a dimer and has 2-fold (180°) rotational symmetry. The SPR monomer is an alpha and beta (a/b) class protein with a 3-layer (aba) sandwich architecture and Rossmann fold topology, and it contains an NADP- binding Rossmann-like domain [34].

We explored feasible binding modes both for the SPR monomer and the dimer. The docking computations were carried out on each binding mode by geometric complementarity and semi-flexible docking to allow for inherent receptor flexibility. From each computation, the 50 lowest energy-docking positions were saved for further analysis. The presumed SSZ-binding sites were ranked by conservation score, specifically by the frequency of occurrence of a residue in a contact surface. The contact surface was delimited as an area consisting of the residues inside a 3.6 Å radius of the ligand.

Based on the conservation scores of all the residues, we identified the main binding location within the NADP-binding Rossmann-like domain. A consensus of five binding regions constituted the receptor pocket comprising residues Gly11, Ser13, Arg14, Phe16 (Region 1), Ala38, Arg39 (Region 2), Asn97, Ala98, Gly99, Ser100 (Region 3), Tyr167 (Region 4), and Leu198, Thr200, Met202 (Region 5). Thus, the binding pocket appeared to contain 2 basic polar residues, 5 neutral polar residues, and 7 neutral non-polar residues. Due to the presence of 2 arginine residues, the site has a basic, positively charged character which may be essential for SSZ binding. Most or all of SSZ exists in a non-protonated, negatively charged state at neutral pH, as the acidic pK_a of carboxylic acid is 2.3 and the pK_a of the sulfonamide nitrogen is 6.5, i.e. less than half-protonated at pH 7.0 [35].

The same residues listed above are involved in NADP+ binding, but the complete NADP+ binding site extends beyond these residues (Table 2). The monomeric or dimeric state of SPR did not affect the location of the SSZ binding site in the simulations, indicating that dimerization does not directly block the access of ligand to the receptor. Table 2 also lists the dimer interface residues. Indeed, the interface residues do not share common elements with the SSZ/NADPH+ binding pocket. Only Tyr167, which is part of both ligand sites, is found in the vicinity of an interface residue, i.e. Cys168.

Figure 5 shows the binding of SSZ to SPR monomer and dimer, respectively. Both chains were found to simultaneously bind ligands in the dimer. While the SSZ site is close to the N-terminus in the primary structure, it appears near the middle of the protein in the 3D fold. The binding pocket is not in very close contact with the dimerization interface and only a few side chains project into the joint neighborhood. The figure also shows the NADP+ binding site of SPR in side-by-side comparison and overlay mode with SSZ. The superimposition of the ligands clearly illustrates that the two binding sites are essentially the same. The geometric center of SSZ and NADP+ is separated only by about 0.5 Å from each other in the superimposed binding pockets. Thus, from Fig. 5 and Table 2 it appears that the binding site for SSZ coincides with the region previously identified in NADP+ binding in the X-ray structure. As a consequence, this could help elucidate the interaction between SSZ and SPR in in vitro and in vivo studies.

To test whether the combined treatment with SSZ and DFMO induces synergistic cell death in NB, we treated SK-N-Be(2)c and LAN-5 cells with different concentrations of SSZ and DFMO. We used two common methods to analyze drug-drug interactions, the isobologram and the combination index (CI) analysis. For both combination analyses, we measured the SSZ and DFMO interaction at 50 % effect level. We first determined the single-agent IC₅₀ concentration for SSZ and DFMO in NB cell lines SK-N-Be(2)c and LAN-5 (Fig. 6, a and b) using an MTS cell viability assay after 48 h of treatment. SSZ exhibited an IC₅₀ value of 133.1 μM for SK-N-Be(2)c and 337.2 μM for LAN-5 cells. DFMO showed an IC₅₀ value of 4.07 mM for SK-N-Be(2)c and 5.79 mM for LAN-5 cells. Subsequently, we combined SSZ and DFMO at different concentrations based on each IC₅₀ value to treat the two NB cell lines, generated isobolograms, and calculated the CI values illustrating the observed synergy. As shown in Fig. 6c and Table 3, SSZ and DFMO combinations revealed slight synergism in SK-N-Be(2)c cells when drug concentrations were below 29.64 μM and 1.80 mM, respectively. Strikingly, SSZ and DFMO showed strong synergism in LAN-5 cells when drug concentrations were below 1.20 μM and 1.21 mM, respectively.