Marcin
Bielinski‡
a,
Lucy R.
Henderson‡
b,
Yuliana
Yosaatmadja§
c,
Lonnie P.
Swift
b,
Hannah T.
Baddock¶
b,
Matthew J.
Bowen
a,
Jürgen
Brem||
a,
Philip S.
Jones**
d,
Stuart P.
McElroy**
d,
Angus
Morrison**
d,
Michael
Speake**
d,
Stan
van Boeckel
e,
Els
van Doornmalen
e,
Jan
van Groningen
e,
Helma
van den Hurk
e,
Opher
Gileadi††
c,
Joseph A.
Newman
*c,
Peter J.
McHugh
*b and
Christopher J.
Schofield
*a
aChemistry Research Laboratory, Department of Chemistry and the Ineos Oxford Institute for Antimicrobial Research, University of Oxford, Mansfield Road, Oxford OX1 3TA, UK. E-mail: christopher.schofield@chem.ox.ac.uk
bDepartment of Oncology, MRC Weatherall Institute of Molecular Medicine, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DS, UK. E-mail: peter.mchugh@imm.ox.ac.uk
cCentre for Medicines Discovery, NDM Research Building, University of Oxford, Old Road Campus Research Building, Roosevelt Drive, Oxford OX3 7DQ, UK. E-mail: joseph.newman@cmd.ox.ac.uk
dUniversity of Dundee, European Screening Centre, Newhouse, ML1 5UH, UK
ePivot Park Screening Centre, 5349 AB Oss, The Netherlands
First published on 30th April 2024
The three human SNM1 metallo-β-lactamase fold nucleases (SNM1A–C) play key roles in DNA damage repair and in maintaining telomere integrity. Genetic studies indicate that they are attractive targets for cancer treatment and to potentiate chemo- and radiation-therapy. A high-throughput screen for SNM1A inhibitors identified diverse pharmacophores, some of which were shown by crystallography to coordinate to the di-metal ion centre at the SNM1A active site. Structure and turnover assay-guided optimization enabled the identification of potent quinazoline–hydroxamic acid containing inhibitors, which bind in a manner where the hydroxamic acid displaces the hydrolytic water and the quinazoline ring occupies a substrate nucleobase binding site. Cellular assays reveal that SNM1A inhibitors cause sensitisation to, and defects in the resolution of, cisplatin-induced DNA damage, validating the tractability of MBL fold nucleases as cancer drug targets.
Potential DNA damage response (DDR) enzyme targets include the glycosylases, helicases, translocases, and nucleases that process damaged DNA.1 Three human ‘MBL-fold’ DNA repair factors, SNM1A, SNM1B (Apollo) and SNM1C (Artemis) are involved in repair of ICLs and DSBs. SNM1A–C have conserved β-CASP and metallo-β-lactamase (MBL) domains (Fig. 1A–C).6–8 The MBL fold is present in many hydrolases and other metallo-enzymes, including Ambler class B Zn(II) dependent β-lactamases, where it has been shown that potent and selective in vivo inhibition can be achieved,9,10 including as a result of optimising hits arising from a high-throughput screen.11 Interestingly, given their central biological importance, there has been relatively little work on targeting MBL-fold nucleases, including SNM1A–C, as anti-cancer or other targets, with one exception being ongoing efforts to develop inhibitors of the mRNA MBL-fold processing Cleavage and Polyadenylation Specificity Factor (CPSF3) as a cancer target12 and as an antiparasitic agent.13,14
SNM1A–C are orthologous with a fungal ancestor, Pso2/Snm1, a key ICL repair factor in Saccharomyces cerevisiae.15 SNM1A/SNM1B are 5′–3′ exonucleases with the unusual capacity to hydrolyse DNA substrates containing lesions, including ICLs (Fig. 1C). SNM1A/SNM1B are both involved in ICL repair. SNM1B is also associated with repair of IR-induced DSB damage. SNM1A has a role in replication-coupled ICL repair,16 functioning with the endonuclease, XPF-ERCC1, to effectively “unhook” the ICL and allow downstream processing. SNM1B has a key role in telomere maintenance, mediated by its interaction with TRF2,17,18 where SNM1B resects the telomeric leading-strand to generate the 3′-overhang necessary for t-loop formation,19 which is required for telomere protection (Fig. 1C). Therapeutic targeting of SNM1B may thus induce senescence and/or loss of tumour cell viability due to telomere attrition. Very recently, a key role for SNM1A as an effector of the break-induced synthesis at telomeres has been unveiled, implying SNM1A is a potential target in tumours that depend upon telomere maintenance governed by the alternative lengthening of telomeres (ALT) pathway.20
SNM1C was first identified in studies of congenital radiosensitive severe acquired immune deficiency (RS-SCID).21 When bound to, and phosphorylated in its C-terminal region (which is distal to the MBL-β-CASP core fold) by DNA-dependent protein kinase (DNA-PK), SNM1C acquires endonuclease activity resulting in hydrolysis of hairpin-end intermediates generated during class-switch recombination and V(D)J recombination (Fig. 1C), a role consistent with the SCID phenotype caused by lack of SNM1C.21 SNM1C also has 5′–3′ exonuclease activity, though this is apparently weaker than that of SNM1A and SNM1B.22 It is likely that the radiosensitivity associated with SNM1C deficiency results from reduced capacity to process chemically damaged DNA termini at DSBs as part of the NHEJ repair pathway. SNM1C is thus a highly attractive target for sensitising cells to IR and radiomimetic drugs.
Early work to identify SNM1A–C inhibitors revealed moderate inhibition by some cephalosporin antibiotics providing an interesting link to the true bacterial MBLs.23 Squaramides and certain nucleoside analogues also inhibit SNM1A, however, with modest potency.24,25 Crystallographic analyses on SNM1A/SNM1B have revealed a 5′-phosphate binding pocket, which is absent in SNM1C, but which is critical for efficient SNM1A/SNM1B activity.26 Consistent with this observation, the addition of a group mimicking a 5′-phosphate was found to improve the potency of modified nucleoside SNM1A inhibitors.27
Here, we report how a high-throughput screen identified novel SNM1A inhibitor chemotypes. Optimisation enabled the identification of potent hydroxamic acid-based inhibitors, with candidates displaying selectivity for the individual SNM1 nucleases. The selective SNM1A inhibitors are cell-active, sensitise cancer cells to a key crosslinking anticancer drug (cisplatin), and lead to trapping of the SNM1A nuclease at sites of ICL-induced damage. The combined result provides strong evidence for the validity and tractability of SNM MBL-fold nucleases as cancer drug targets.
We worked to obtain information on the binding modes of the hit compounds by soaking SNM1A crystals with them (1 mM final concentration), both in the presence and absence of added ZnCl2. Although the identity of the metal ion(s) used by SNM1s in vivo is uncertain, the available bio-chemical/physical evidence is that this is most likely Zn(II). MBL fold enzymes normally have two metal ion binding sites (M1 and M2, Fig. 1C), although sometimes only one site is occupied or is required for catalysis.28 We, and others, have observed variations in SNM1 active site metal ion occupancy by crystallography, which may relate to differences in purification methods, in particular, when using Ni-affinity chromatographic (IMAC) purification. Indeed, structures of SNM1A–C enzymes are reported in both mono- and di-metal ion active site forms.26,29–31 When only a single metal ion is present it occupies the M1 site and is ligated by 4 protein residues (H732, H734, H737 and D815, in the case of SNM1A). The second metal ion site (M2) only has 3 protein residues as ligands (D736, H737 and D815 in the case of SNM1A); it may be that the M2 metal ion is bound less tightly and as such is not always carried through the purification procedures.
Because the relevant in vivo metalation status of the SNM1A used in the high-throughput screen cannot be unequivocally determined, soaking was performed under two conditions using the orthorhombic SNM1A crystal form. This crystal form grows from conditions containing malonate, imidazole and boric acid (MIB) buffer and manifests an active site where a malonate ion is bound to a single M1 site metal ion (presumed to be a nickel ion due to IMAC purification) (ESI Fig. 2†). To obtain a single metal malonate free form, crystals were resoaked in buffer lacking malonate (0.1 M HEPES pH 7.0, 30% PEG 1000) overnight. To obtain a two-metal ion form the same crystals were re-soaked in buffer lacking malonate, but with ZnCl2 (0.1 M HEPES pH 70, 30% (v/v) PEG 1000, 500 μM ZnCl2), a process enabling occupation of the M2 site with high occupancy (ESI Fig. 2†).
Soaking of SNM1A crystals with the ELF hits for which inhibition was validated (see below), that is 1 [IC50 = 2.4 μM] and 2 [IC50 = 2.0 μM], was successful for 1 in the two-metal ion SNM1A form and 2 in the single metal ion SNM1A form (Fig. 2A–H). The electron density maps reveal both 1 and 2 bind directly to the metal ion(s) via oxygen atoms. Interestingly, 1 is an N-hydroxyimide, a pharmacophore that is known to inhibit nucleases and other metalloenzymes including the true bacterial MBLs.32,33
Two molecules of 1 were observed in the SNM1A active site, each positioned to coordinate to M1 or M2 via two oxygen ligands giving an octahedral arrangement (Fig. 2A and D); precedent for an MBL fold inhibitor binding mode involving two molecules comes from the observation of binding of two rhodanine derived inhibitor molecules at the active site of a true MBL.34 The molecule of 1 bound to the M1 site is positioned to hydrogen bond with the main chain amide of D736 and to interact with the side chains of S735, Y879 and S800 (Fig. 2A). The molecule of 1 bound to the M2 site is positioned to form a potential hydrogen bond with H994 and to make a water mediated contact with Y184 (Fig. 2A). The aromatic rings of both molecules of 1 stack with each other in a staggered arrangement, with the ring–ring distances between the two 1 rings being 3.2–3.5 Å (Fig. 2D).
The electron density map for the SNM1A: 1 complex revealed an apparent inconsistency between the ELF reported small molecule structure and the crystallographic data. The ELF report has the methoxy group of 1 at the meta-position, however, the electron density maps clearly imply that the methoxy substituent is in the para-position for both crystallographically observed molecules of 1 (Fig. 2C). Re-synthesis of each of 1, 2, and 3 revealed comparable levels of SNM1A inhibition as determined by the real-time fluorescence-based nuclease assay for each of the regioisomers with IC50 values of 2.4 μM, 2.0 μM, and 2.9 μM, respectively (ESI Fig. 3†).
The structures of 1 and 3 are closely related to N-hydroxyimides that are known to inhibit other nucleases, including XPF-ERCC1 and FEN1 (flap structure specific endonuclease 1).35,36 We also tested the inhibitory potential of a potent and N-hydroxyimide based FEN1 inhibitor against SNM1A, that is AZ1353160 (4). Whilst 4 showed some SNM1A inhibition, this was incomplete and an IC50 value could not be calculated from the fluorescence-based nuclease assay data. A structure of 4 complexed with SNM1A was solved to 1.7 Å resolution (ESI Fig. 4A–D†). As observed with 1, AZ1353160 (4) binds to the di-metal SNM1A form with two molecules of 4 at the active site, both of which are positioned to coordinate to the active site metal ions via oxygen atoms, in a similar manner to that observed for 1 (Fig. 3C). AZ1353160 (4) was obtained as a racemic mixture (ESI Fig. 4B†) and the best fit to the density was obtained with one (R)-enantiomer and one (S)-enantiomer, at the M1 and M2 sites, respectively (ESI Fig. 4C†). As for 1, the two 4 molecules stack in a staggered manner, with differences likely reflecting the benzo-dioxane substituent on the urea nitrogen of 4. The AZ1353160 (4) binding mode with SNM1A contrasts with that observed for it with FEN1,36 where a single inhibitor bridges the two magnesium ions in the FEN1 active site forming two contacts to each metal ion (ESI Fig. 5†).
As for binding of 1 at the M1 site, the one molecule of 2 observed at the active site was also observed to coordinate the SNM1A M1 metal ion form via its two carbonyl oxygens (metal–ligand distances: 2.0 and 2.1 Å) (Fig. 2E–H); its SNM1A binding mode is similar to that of the catechol group of ceftriaxone (5) which inhibits SNM1A and SNM1C with low μM potency23,30 (ESI Fig. 6†). A hydrogen bond is formed between H994 and the 2 diamide nitrogen (Fig. 2E) and the H2 chlorobenzene ring is close to the Y841 and T840 sidechains (Fig. 2E).
The obtained structures reveal both 1 and 2 (and, likely 3) bind directly to the active site metal ion(s) of SNM1A via two oxygen atoms and likely inhibit in a competitive manner with the DNA substrate, displacing the Zn(II) ion complexed hydrolytic water/hydroxide that, in models of the reaction mechanism, is positioned for hydrolysis of the phosphodiester bond in the DNA backbone. Ceftriaxone (5) is an inhibitor of SNM1A/SNM1B, complexing the active site metal ion via its catechol group.23 We compared the inhibition by ceftriaxone (5) with that of 1, 2 and 3 and found each 1, 2 and 3 to be more potent and consistent inhibitors of SNM1A than ceftriaxone (5) (ESI Fig. 3†). We therefore employed 1 as a positive control in subsequent fluorescence-based and gel-based nuclease SNM1 inhibition assays. Although, we have not pursued this in our current work, the different observed binding modes for 1 and 2 suggest that it may be possible to identify inhibitors selective for binding to forms of SNM1A with one or two metal ions.
In the ELF screen a structurally similar quinazoline-based inhibitor, K1 (6) was identified as an SNM1A inhibitor, with an IC50 value of 17.4 μM. Resynthesised K1 (6), however, had a higher IC50 value for SNM1A (129 μM); despite being less potent than 1 and 2; K1 (6), however, possesses a heteroaromatic structure we considered more amenable to modification. We therefore developed a synthetic scheme (ESI Fig. 8†) suitable for the versatile synthesis of K1 (6) analogues.
We hypothesised that the carboxylic acid of K1 (6) may bind to one or both of the active-site metal ions. Indeed, when the carboxylic acid of 6 was replaced with normally less effective chelating groups, such as ester (7), amide (8) or hydroxyl (9) groups, there was a marked loss of SNM1A inhibition (Fig. 3B and C). Given the potency of both 1 and 3 as SNM1A inhibitors and the information about binding provided by the SNM1A: 1 crystal structures (Fig. 2), we substituted the carboxylic acid of K1 (6) with a hydroxamic acid (10), a change which resulted in a 3-fold increase in potency (Fig. 3B and C). Hydroxamic acid-based inhibitors containing different amino acid derivatives were then synthesised and tested against SNM1A (11–14); all showed significant improvement in potency compared to K1 (6). Compound 13 containing an L-alanine based sidechain was identified as the most potent inhibitor against SNM1A with an IC50 of 0.8 μM (Fig. 3B and C).
There is structural similarity between SNM1A, SNM1B, and SNM1C and major mechanistic features are likely conserved between the three nucleases. We therefore tested (13) and other hydroxamic acid analogues for inhibition of SNM1B and SNM1C. Due to differences in catalytic turnover between the three enzymes, different concentrations of SNM1B (1 nM) and SNM1C (2.5 nM) were used compared to that for SNM1A (0.5 nM) in the assays. Note that the substrate utilised for the high-throughput fluorescence based and subsequent SNM1A and SNM1B assays differed slightly for SNM1C reflecting its (primarily) endonucleolytic, rather than exonucleolytic, activity (ESI Fig. 9†).
Many of the hydroxamic acid compounds inhibit all three of SNM1A, SNM1B and SNM1C, though there were notable differences in potency (ESI Fig. 9†). The most potent inhibitor against SNM1A was 13 with an IC50 of 0.8 μM. In general inhibitors with a small sidechain at position 2′, such as with allylamine (12, 13 and 14) or dimethylamine (22 and 23) sidechains showed the best inhibition against SNM1A, while inhibitors with bulkier sidechains manifested a decrease in SNM1A potency. Interestingly introduction of an extra carbonyl group at position 2′ (as in amide 20), resulted in a 100-fold decrease in potency when compared to 13, while against SNM1B and SNM1C inhibitors 13 and 20 show inhibition with a similar range. Similarly, introduction of an ethoxy group at position 2′ caused a much steeper drop in potency against SNM1A, compared to SNM1B and SNM1C.
SNM1B and SNM1C can tolerate more bulky aromatic sidechains at position 2′, such as benzylamine (16) or furfurylamine (19) derivatives. 16 was the most potent identified SNM1C inhibitor with an IC50 of 1.1 μM; 19 was the most potent SNM1B inhibitor with an IC50 of 2.5 μM. In terms of selectivity, it is notable that 18, which has a diethylamine sidechain at the 2′ position, was only potent against SNM1C. Furthermore, based on results for compounds 22, 23 and 24, SNM1A and SNM1B likely do not tolerate groups larger than a halogen atom at the 6′ position, because the introduction of a methoxy group resulted in significant drop in potency (24), while for SNM1C all three compounds (22, 23 and 24) were similarly potent.
The overall results identify quinazoline–hydroxamic acids as broadly effective inhibitors of the human SNM1 family of nucleases. Importantly, the SAR studies show how small changes in active site binding elements can make large differences in the relative potency versus SNM1A/B/C, likely due to subtle differences in the precise active site architectures of the three enzymes.
The R1 methyl group (Fig. 4) of the inhibitors is directed towards the side chains of H734 and S880. Two of the inhibitors (13 and 19) have a chlorine atom at the R3 position, which is positioned close (∼3.2 Å) to the side chain of S735 (Fig. 5B and C). The binding modes of the R2 substituents are more diverse, but all of them occupy a similar position close to K883, Y841 and G963. These three residues form the binding pocket of the 5′ phosphate of the substrate, which has been shown to be a key determinant of DNA binding and exonuclease activity for SNM1B.26 With all three inhibitors, the R2 groups are positioned in an equivalent position to the ribose ring of the nucleotide in the SNM1B nucleotide complex, but do not extend into the phosphate binding pocket itself (Fig. 5D). In the SNM1A complex with 19, weak electron density corresponding to apparent binding of a second molecule of 19 was observed; the second molecule of 19 is located within stacking distance of the first molecule and is positioned to form additional contacts with a symmetry related molecule in the crystal. Although intriguing given the observations with 1 and 2 (Fig. 2), this interaction is likely at least in part due to crystal lattice interactions.
To explore this point more directly, we investigated whether U2OS cells treated with cisplatin and 19 exhibit the hallmarks of defects in SNM1A-mediated repair. To this end, cells stably expressing N-terminally EGFP-tagged SNM1A (EGFP-SNM1A) were treated with 50 μM cisplatin alone, 19 (50 μM), or with a combination of both compounds. It is known that on treatment of cells with DNA crosslinking agents, SNM1A forms subnuclear foci that are associated with ongoing sites of DNA repair.37 Treatment with 50 μM 19 alone did not induce EGFP-SNM1A repair-associated foci, or foci of a common marker, i.e. a phosphorylated form of histone variant H2AX (γH2AX) used to mark sites of DNA breakage and repair (Fig. 6C, quantified in Fig. 6D and E). As previously established, cisplatin treatment efficiently induces co-localisation of EGFP-SNM1A foci and γH2AX foci within 4 hours following treatment, marking sites of cisplatin damage and its repair (Fig. 6C–E). These largely resolved 24 hours post-treatment, again consistent with previous studies.16,37 Co-treatment of cells with 19 and cisplatin saw a robust induction of SNM1A foci, consistent with the SNM1A activity being dispensable for its localisation to sites of crosslink repair, and γH2AX foci formation (Fig. 6C–E). In the presence of 19 both classes of colocalising foci persisted at 24 hours. This clear persistence of repair intermediates and trapping of EGFP-SNM1A at the sites of these repair intermediate implies that 19 engages with SNM1A in a cellular context, delaying the completion of cisplatin crosslink repair, which, in turn, results in the sensitisation of cells to cisplatin.
Finally, we explored the response of genetically stable and karyotypically normal immortalised (non-cancer) cells to 19, in the presence and absence of cisplatin. As for the U2OS cells, RPE-1 cells (hTert-immortalised retinal pigment epithelial cells) treated with 19 alone exhibited a mild toxicity only at the highest dose employed (100 μM), where a negligible reduction in survival was observed at 50 μM (ESI Fig. 12†). For RPE-1 cells treated with cisplatin, the inclusion of 50 μM 19 did not substantially increase their dose-dependent sensitivity of cisplatin (Fig. 6F), compared to the vehicle-alone control. Moreover, while the number of γH2AX foci persisting in RPE-1 cells 24 hours after 50 mM cisplatin treatment was elevated in the presence of 50 μM 19 relative to the vehicle-alone control (Fig. 6G), this increase was less dramatic than that observed with U2OS cells. The average number of γH2AX foci in cisplatin treated U2OS at 24 hours increases 4-fold following pretreatment with 19 (21.5 ± 0.5 to 81.6 ± 1.7 foci). By contrast in RPE-1 cells, the average γH2AX foci count only increased 2-fold under the same conditions (10.4 ± 0.3 to 21.2 ± 0.7 foci). Together, these observations imply a cancer cell-selective increase in cisplatin sensitivity associated with DNA damage persistence might be achievable through inhibition of SNM1A by small molecule inhibitors.
The results presented here demonstrate the viability of developing highly potent and selective inhibitors of the human SNM1A–C nucleases which play vital roles in DDR, and which are targets for cancer treatment. To identify scaffolds suitable for SNM1 inhibition, we carried out a HTS employing a fluorescence-based assay38 and the ELF compound collection. This approach has previously delivered new types of inhibitors for the clinically relevant family of di Zn(II) ion dependent B1 subfamily of the true MBL NDM-1 and related B1 subfamily MBLs;11 MBL inhibitor types identified include the indole carboxylates, which bind to the active site Zn(II) ions in a manner that stabilises, but which does not displace the hydrolytic water, which (at least in the resting enzyme) bridges between the two Zn(II) ions; the indole carboxylate scaffold, however, does not inhibit the SNM1 nucleases.11 The ELF screen identified several SNM1A hit pharmacophores that we considered suitable for further exploration as SNM1A inhibitors, including by structural studies involving both the mono- and di Zn(II) forms of SNM1A, which were used because the precise nature of the metal ions at the SNM1 active sites in vivo is unknown.
Initially, we focused on the cyclic hydroxyimide hits 1–3 as SNM1A inhibitors, in part because related compounds have been shown to inhibit the nuclease FEN1 (flap structure specific endonuclease 1), which like SNM1A is also involved in DNA damage repair pathways.35,36 The N-hydroxyimides 1–3 were shown by crystallography to bind to the active site metal ions in a manner that displaces the hydrolytic water/hydroxide and which will compete with the substrate (at least in the active site region). Although, there is likely scope for future development of cyclic N-hydroxyimides as SNM1 inhibitors, the insolubility of the tested compounds coupled with the mechanistically interesting, but complicating, crystallographic observation that, at least in cases, two molecules of the cyclic N-hydroxyimide can bind at the active site, prompted us to explore other pharmacophores from the ELF screen, in particular quinazoline-containing inhibitors (6, and related compounds), where 6 was a moderately potent inhibitor, but which appeared more amenable to SAR studies than the cyclic N-hydroxyimides. Based on analysis of the SNM1A crystal structures obtained with 1–3, we substituted the C2 carboxylic acid of the hit quinazoline inhibitors with a hydroxamic acid, a known pharmacophore for the true MBLs,39 a modification that led to increased potency of SNM1A inhibition. Subsequent SAR studies involving modifications at the C4 quinazoline position enabled the identification of potent SNM1A inhibitors, as shown by both fluorescence and gel-based assays (Fig. 4B). Interestingly a combination of a thioxodihydroquinazolinone and cisplatin has been reported to work synergistically in cisplatin resistant cells and shows promising results in a mouse model.40,41 Although, the thioxodihydroquinazolinones are structurally distinct from the quinazoline-containing inhibitors reported here, it cannot be ruled out that they, or metabolites of them, act as SNM1 inhibitors in vivo.
Crystallographic analyses on three of the quinazoline–hydroxamate inhibitors (13, 19 and 20) reveal that they bind in similar manner to β-lactam antibiotics binding to true MBLs, with the quinazoline ring binding in a manner equivalent to that of an adenine ring observed in an SNM1B nucleotide complex structure26 (ESI Fig. 13†). The quinazoline C4 linked hydroxamate is positioned to interact with the M1 and M2 metal ions of SNM1A, with its carbonyl oxygen coordinating to M1 and its hydroxyl/hydroxide group occupying the M1:
M2 bridging position which is occupied by the hydrolytic water/hydroxide during catalysis.
Hydroxamic acids have been developed for clinical use as histone deacetylase inhibitors,42 though such compounds also have potential to inhibit other human metallo-enzymes43 and to act as non-selective metal ion chelators. Hence appropriate derivatisation of them is required to enable selectivity, which in the case of SNM1 inhibitors includes against other human MBL fold nucleases such as CPSF73.28 Although there is clearly scope for further optimisation of the inhibitors reported here (e.g. cyclic hydroxamic acids and use of other Zn(II) chelating pharmacophores), the results comparing inhibition of all three of SNM1A, B and C are interesting with respect to selectivity. Whilst many of the hydroxamic acids inhibited all three human SNM1 nucleases with comparable efficiency, the results clearly indicate developing inhibitors selective for the individual isoforms should be possible, including via modification of the quinazoline 2′ and 6′ positions. Thus, SNM1B and SNM1C can tolerate larger groups at the 2′ position than SNM1A and SNM1A and SNM1B are less tolerant of substitution at the 6′ position than is SNM1C.
We tested selected compounds for evidence that they can inhibit SNM1A in cells. It is well-established that SNM1A is a key mediator of DNA crosslink repair, and that SNM1A deficient cells are sensitive to drugs that induce DNA crosslinks, such as cisplatin.16,37 We observed cancer cell sensitisation when treatment with cisplatin treatment was combined with 19. A chemical-genetic approach, employing matched, engineered SNM1A− cells demonstrated that the sensitisation of U2OS cells is dependent on the presence of SNM1A, implying that the observed sensitisation is (at least in part) due to on-target engagement by 19. To more directly examine this, we employed imaging approaches, the results of which revealed that EGFP-SNM1A is efficiently recruited to sites cisplatin of damage and repair as marked by γH2AX foci. Importantly, upon co-administration of cisplatin and 19, SNM1A persisted at such sites, indicating that repair is compromised. The recruitment of SNM1A to crosslinks is known to be independent of its catalytic activity and to be mediated by interaction of SNM1A with PCNA (proliferating cell nuclear antigen) via the conserved SNM1A PIP box motif and with the ubiquitin-modified form of PCNA (PCNAUb) mediated by a ubiquitin-binding zinc finger in the N-terminus of SNM1A. Together, our cellular data indicate that 19 likely competes with repair substrates for SNM1A, stalling the step in crosslink repair that is mediated by SNM1A as evidenced by persistence of SNM1A at sites of cisplatin damage that are also marked by the DNA repair intermediate marker, γH2AX.
Interestingly, compared with U2OS cells an immortalised (non-cancer) cell line (RPE-1) exhibited a less dramatic senstisation to cisplatin in combination with 19 treatment. Moreover, a concomitantly lower fold increase in persistent γH2AX foci was observed. This observation suggests that when additional DNA damaging genomic stress is simultaneously administered pharmacologically, tumour-derived cells might be more vulnerable to SNM1A loss than non-cancer cells. The reasons for this require further investigation with one possibility being additive effects of SNM1A inhibition and underlying defects in the machinery maintaining genome stability in cancer cells, and/or the presence of more robust cell cycle checkpoints in non-cancer cells preventing unrestrained entry into S-phase in the presence of cisplatin damage.
Confocal images were obtained with a Plan APO 63X 1.40NA oil immersion objective, a pinhole setting of 1 AU, bandpass emission settings of 410–468 nm for Hoechst, 490–544 nm for EGFP, 579–624 nm for RFP or Alexa 568, and 633–695 nm for Alexa 647, a projected pixel dimension of around 110 nm × 110 nm, a pixel dwell time of 1.35 μs, and with a line averaging setting of 2. In order to ensure sufficient cell numbers (N > 300), images were acquired in a tiled 5 × 5 format corresponding to an image area of about 0.65 mm × 0.65 mm. Images were imported into ImageJ and foci were counted using a macro adjusting for staining levels between experiments. Anti-γH2AX antibodies were a mouse monoclonal (Millipore; JBW301).
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4sc00367e |
‡ The contributions of these authors should be considered equal. |
§ Current address: School of Biological Sciences, Building 110, 3A Symonds Street, Auckland 1010, New Zealand. |
¶ Current address: Calico Life Sciences, 1170 Veterans Blvd, South San Francisco, CA 94080 USA. |
|| Current address: Enzymology and Applied Biocatalysis Research Center, Faculty of Chemistry and Chemical Engineering, Babes-Bolyai University, Str. Arany Janos, nr. 11. Cluj-Napoca, RO-400028. |
** Current Address: BioAscent Discovery, Newhouse, ML1 5UH, United Kingdom. |
†† Current address: SGC Karolinska, Center for Molecular Medicine (CMM), Karolinska University Hospital, 171 76 Stockholm, Sweden. |
This journal is © The Royal Society of Chemistry 2024 |