Fragment-based nuclear magnetic resonance screen against a Regulator of G Protein Signaling identifies a binding “hot spot”

Abstract

Regulator of G Protein Signaling (RGS) proteins have attracted attention as a result of their primary role in directing the specificity as well as temporal and spatial aspects of G protein coupled receptor signaling. In addition, alterations in RGS protein expression have been observed in a number of disease states, including certain cancers. In this area, RGS17 is of particular interest. It has been demonstrated that while RGS17 is normally expressed primarily in the central nervous system, it has been found to be inappropriately expressed in lung, prostate, breast, cervical, and hepatocellular carcinomas. Overexpression of RGS17 leads to dysfunction in inhibitory G protein signaling and an overproduction of the intracellular second messenger cAMP, which in turn alters the transcription patterns of proteins known to promote various cancer types. Suppressing RGS17 expression with RNA interference (RNAi) has been found to decrease tumorigenesis and sufficiently prevents cancer cell migration, leading to the hypothesis that pharmacological blockage of RGS17 function could be useful in anticancer therapies. Here, we identified small molecule fragments capable of binding the RGS homology (RH) domain of RGS17 using a nuclear magnetic resonance (NMR) fragment-based screening approach. By chemical shift mapping of the two-dimensional (2D) ¹⁵N/¹H-heteronuclear single quantum coherence (HSQC) spectra of the backbone-assigned ¹⁵N-labeled RGS17-RH, we determined the fragment binding sites to be distant from the Gα interface. Thus, our study identifies a putative fragment binding site on RGS17 that was previously unknown.

Graphical Abstract

Altered expression of RGS17 is leads to dysfunction in inhibitory G Protein signaling results in an overproduction of the intracellular second messenger cAMP, which in turn alters transcription for proteins known to promote various cancer types. Decreasing RGS17 expression leads to a reduction in tumorigenesis and can sufficiently prevent cell migration. Therefore, inhibition of RGS17 represents a potential avenue for anticancer therapies. Here we used an NMR FBLD approach to identify fragments capable of binding RGS17. Subsequent chemical shift mapping of the two-dimensional (2D) ¹⁵N/¹H-heteronuclear single quantum coherence (HSQC) spectra of the backbone-assigned ¹⁵N-labeled RGS17-RH, led to the identification of a putative fragment binding site on RGS17 that was previously unknown.

Introduction

Regulators of G Protein signaling (RGS) are attractive targets for chemical intervention as they are downstream of G Protein-coupled receptors (GPCRs) and spatially and temporally regulate GPCR signaling pathways. GPCRs, represent one of the largest classes of proteins in the human proteome, and as such they mediate a multitude of physiological processes. In their role as regulators of the GPCR cycle, RGS proteins decrease the half-life of GTP-bound active Gα subunits, therefore the inhibition of RGS proteins has the marked effect of enhancing GPCR signaling. A number of RGS proteins have been identified as being overactive or overexpressed in certain disease states ^[1]. One such RGS protein that is an appealing drug target is RGS17, a member of the RZ subfamily of RGS proteins. Despite low levels of Rgs17 expression outside the CNS, an increasing body of evidence implicates elevated RGS17 activity in a number of cancer types. Overexpression of RGS17 is observed in some lung and prostate tumors by over 7.5 fold as compared to normal tissue ^[2], and a susceptibility locus for familial lung cancer was tracked to chromosome 6q 23–25, the location of Rgs17 ^[3]. Further, RNA inference mediated reduction in RGS17 protein levels led to a loss of 59–75% tumor volume in a mouse xenograft cancer model ^[2]. Not surprisingly, alterations in RGS17 protein expression levels have a cascading effect. Overexpression of RGS17 has been found to alter the transcription of CRE-regulated genes by decreasing Gα_i/o/z activity ^[2]. This finding helps to explain how RGS17 knockdown leads to anti-proliferative effects, as RGS17 knockdown will lead to prolonged inhibitory Gα subunit signaling, consequently inhibiting the activation of the adenylyl cyclase-protein kinase A-CREB (AC-PKA-CREB) pathway. Published data support the regulation of RGS17 protein levels in cancer by short non-coding RNAs (microRNAs) that regulate translation of their specific target by RNA silencing ^[4]. For example, the overexpression microRNA-182 and microRNA-203 have both been found to reduce RGS17 levels and suppress proliferation, invasion, and migration in lung cancer cell lines ^[5]. Further examples of microRNA mediated regulation of RGS17 protein levels can be found in hepatocellular carcinomas (microRNA-199) ^[6], cervical cancer (miRNA-508–3p/Linc00483) ^[7], and breast cancer (microRNA-32) ^[8]. These examples showcase the role of RGS17 in cancer progression. As RGS17 is characteristically upregulated in several cancers, the discovery of compounds capable of inhibiting RGS17 activity has significant potential clinical importance.

Previous high throughput screening (HTS) efforts by our lab have identified a small number of RGS17 inhibitors ^[9]. However, each molecule was found to lack significant inhibition against a cysteine-null RGS17 mutant (C117A), indicating that inhibition was cysteine-dependent ^[9–10]. As we have previously established, compounds that exhibit cysteine-dependent inhibition of RGS proteins are commonly found to lack selectivity for one RGS protein over another ^[11]. Given the limited success of traditional HTS as an approach to identifying RGS17/Gα interaction inhibitors, we turned to fragment-based lead discovery (FBLD). The main goal of FBLD is the same as that of HTS, both approaches aim to identify compounds that elicit a desired response. HTS however employs large compound libraries (>50,000 compounds), commonly with unique and complex structures. FBLD, uses a small number of fragments (250–5000 fragments) that have low-molecular weights (<300 Da) and usually exhibit lower affinities for the target ^[12]. FBLD sacrifices throughput and affinity for access to potential small ligand binding pockets on drug targets using information rich screening methods such as nuclear magnetic resonance (NMR), surface plasmon resonance (SPR), or X-ray crystallography ^[13]. We chose NMR for this screen as it provides definitive evidence of direct binding of test compounds with the protein of interest while maintaining a level of throughput that is superior to other biophysical methods ^[14]. The sensitivity of NMR, which is capable of identifying weak interactions, serves to counteract the characteristically low affinities of fragments used in FBLD. NMR screen for FBLD could employ ligand-detected methods such as WaterLOGSY and saturation transfer difference (STD) NMR that measure the changes in the NMR spectra of a fragment in the presence and absence of the particular protein target ^[15], and/or protein-detected methods that monitor the spectra of the protein of interest in the presence or absence of fragments. One advantage of using protein-detected methods of NMR is the opportunity to obtain binding site information, if the chemical shifts for the protein of interest have been assigned to individual residues^[16]. While this information is invaluable, it is commonly limited by the fact that this method requires significant amounts of purified protein and also the target protein is typically smaller than 50 kD ^[17]. Fortunately, the RGS homology (RH) domain, which forms the complex with the Gα protein, is approximately 17 kD in size, and isotope-¹⁵N-labeled RGS17-RH can be purified in large quantities (>100 mg/L culture from E. Coli). Further, the advantages of NMR based methods have been established for small molecule interactions on RGS4^[18]. Taking into consideration the challenges of conventional HTS to identify RGS17 inhibitors, and the prospect of uncovering binding site information for RGS17 using NMR, we pursued FBLD with NMR as a screening approach, as presented here.

Results

RGS17 ¹⁵N-¹H HSQC Chemical Shift Assignment

To facilitate the identification of fragment binding sites on RGS17, we purified a previously established RGS17 (Asn72 to Ser206) construct that had been crystallized by the Structural Genomics Consortium and optimized for expression in E. coli ^[19]. This 17 kD portion of RGS17 retains the RH domain and it is the key component responsible for forming the complex with Gα proteins. Using purified labeled sample and a suite of multidimensional NMR experiments, we successfully assigned greater than 93% (120/129) of the RGS17-RH backbone amide resonances ^[20]. Figure 1 shows the assignments of the RGS17 ¹H-¹⁵N heteronuclear single quantum correlation (HSQC) spectrum.

Figure 1.

RGS17 Chemical Shift Assignments. A) ¹H-¹⁵N HSQC spectrum of RGS17 RH domain with assigned peaks labeled. B) Boxed region in panel A is expanded for clear labelling of assigned peaks. Red peaks indicated folded peaks.

Assessing the DMSO Stability of the RGS17-RH Domain

The presence of dimethyl sulfoxide (DMSO) is a characteristic of screening campaigns as HTS and fragment libraries are typically dissolved in the solvent. Further, high concentrations of fragments are generally necessary to overcome the weak affinity they exhibit for their target. We assessed the stability of RGS17-RH in the presence of increasing concentrations of DMSO to determine the threshold for our screening strategy. RGS17-RH was found to be quite stable, with only small changes in chemical shifts (Figure 2) observed in the ¹H-¹⁵N HSQC at 5% and 10% DMSO.

Figure 2.

DMSO Stability of RGS17. A) ¹H-¹⁵N HSQC in the presence of 0% (black), 5% (red), and 10% (green) DMSO. B) Boxed region in panel A is expanded for clear view.

Maybridge Ro3 Core Fragment Library Screening

The stability of RGS RH in the presence of high concentrations of DMSO enabled the use of a fragment pooling strategy, allowing for a more efficient screen of the 1,000-fragment library. Apart from saving considerable time and resources, the fragment pooling strategy increased the throughput for this NMR screening method. The fragment pools were made up of 6 fragments at a concentration of 1 mM, a 10-fold molar excess over RGS17-RH, with 6% DMSO in the final sample. Thus, this reduced the number of NMR samples from 1,000 (1 for each fragment) to approximately 170. ¹H-¹⁵N HSQC spectra was obtained for 100 μM RGS17-RH with each fragment pool, and a vehicle sample with 6% DMSO served as a control that was analyzed at the beginning and end of each batch of samples. Examining the spectra for fragment-induced chemical shift perturbations (CSP) identified 17 pooled fragment samples as containing an RGS17-binding compound. ¹H-¹⁵N HSQC spectra were then obtained for the 102 fragment samples (6 fragments per pool x 17 pools identified) and inspection of CSPs led to identification of 22 fragments capable of binding RGS17-RH. To ensure that binding was achieved by the fragments identified and not degradation products, a common pitfall of chemical library storage ^[21], aliquots of fragment were prepared fresh from powder stocks upon reorder. Several fragments after being freshly prepared only caused a shift in a single residue in the RGS17-RH ¹H-¹⁵N HSQC, or binding was lost completely. This post-screen filtering process excluded 15 of the initial 22 hits, resulting in 7 hit fragments capable of RGS17-RH binding (Figure 3).

Figure 3.

RGS17 NMR Fragment Screening Results. A) Scheme for the NMR fragment screen against RGS17 RH domain. B) Chemical structure of hit fragments identified in NMR screen.

Evaluating Fragment Affinity for the RGS17-RH Domain

The 7 hit fragments identified in the NMR screen were prioritized by affinity of each fragment for RGS17-RH. ¹H-¹⁵N HSQC spectra obtained for RGS17-RH with increasing fragment concentrations yielded increasing chemical shifts (Figure 4A). To avoid precipitation of fragments, the solubility limits for each of the 7 hit fragments was determined experimentally by monitoring the height of fragment peaks in ¹H NMR spectra (Figure 4B). The fragment solubility limit was reached when the peak height no longer increased as fragment concentration increased. Dissociation constants (K_D) were determined by plotting the CSPs observed as a function of the fragment concentration. This curve was then fit to a rectangular hyperbolic, one-site binding model as shown in Figure 5. This analysis was performed for each residue that exhibited concentration-dependent peak shifts. To obtain an overall measure of affinity for each fragment the calculated K_D values obtained for each residue were averaged (K_{D, Avg}) (Figure 5 and Table 1). As shown in table 1 the calculated K_{D, Avg} values ranged from 0.5 ± 0.4 mM for fragment 4 to 5.5 ± 4.7 mM for fragment 2, while the number of residues analyzed for each fragment ranged from three to six (Table 1).

Figure 4.

Representative Fragment Titration. A) Changes in chemical shifts of backbone amide peaks in the presence of 0 (red) to 5 mM (blue) fragment (Fragment concentrations: 0.1, 0.4, 0.75, 1, 2, and 5 mM). B) Monitoring of fragment solubility at concentrations ranging from 0 (green) up to 5 mM (brown) fragment.

Figure 5.

RGS17 Affinity for Hit Fragments. Determination of CSP at various increasing concentrations of fragment allows K_D determination. Fragments being analyzed is indicated above plot, and residues being analyzed are labeled. Data was analyzed using one-site specific binding model using GraphPad Prism.

Table 1.

Fragment Screening Summary Results of screening Maybridge Ro3 library against RGS17 using protein-detected 1H-15N HSQC NMR. L.E. Ligand efficiency.

Fragment	CAS#	KD, avg (mM)	Residues Analyzed	L.E.	RGS4 Binding	RGS C117S Binding
1	348134-09-9	3.3 ± 2.0	K121, L163, N167, Y171	0.20	Yes	Yes
2	21928-09-8	5.5 ± 4.7	L163, H169, Q175	0.18	Yes	Yes
3	220863-87-3	n.d.	n.d.	n.d.	Yes	No
4	2664-63-3	0.5 ± 0.4	D139, N167, Q175	0.31	Yes	Yes
5	175137-20-9	5.2 ± 8.5	E109, L163, N167, Q175	0.23	No	No
6	883061-57-2	1.5 ± 0.3	E109, Y137, L163, N167	0.25	Yes	Yes
7	60222-71-1	2.9 ± 2.4	E109, I159, L163, N167, A174, Q175	0.25	Yes	Yes

Covalent Modification of RGS17-RH C117 Residue by Fragment 3

During the course of titration experiments it was observed that fragment 3 produced significant changes in the ¹H-¹⁵N HSQC spectra in a manner that was concentration independent (Figure 6A,B), indicating a potential mechanism of irreversible binding. As many efforts to identify RGS inhibitors have resulted in compounds that covalently modify cysteine residues, it was necessary to assess the method of binding for fragment 3. To assess the covalent binding of fragment 3 to RGS17-RH, a known thio reactive compound N-(7-dimethylamino-4-methylcoumarin-3-yl) maleimide (DACM) was used as a control. DACM is a thiol reactive dye that increases in fluorescence intensity after covalently labeling cysteine residues on proteins. It was found that pre-incubating RGS17-RH with 10-fold molar excess of fragment 3 or N-ethyl maleimide (NEM) (another cysteine reactive compound) decreased DACM labeling, as shown in Figure 6C and 6D. Reduced DACM labeling is attributed to the decrease in available cysteine residues, as pre-incubation with either agent leads to irreversible modification of cysteine residues prior to DACM addition. Further, the cysteine-dependence of fragment 3 was confirmed when it failed to bind to an RGS17-RH mutant with the cysteine-117 residue mutated to serine (C117S) (Figure 7F).

Figure 6.

Covalent Modification of RGS17 by Hit Fragment 3. A,B) ¹H-¹⁵N HSQC of RGS17 in the absence (black) and increasing fragment 3 concentration up to 5 mM (yellow to green). B) Expanded view of boxed region of 6A. C) Kinetic trace of 2 μM DACM reaction with buffer alone or 2 μM RGS17 or 20 μM indicated compound. D) Calculated slope from DACM kinetic traces in 6C. Data represents mean of n=3 ± S.D. ** p ≤0.01 one-way ANOVA with multiple comparisons.

Figure 7.

Structural Perturbations Induced by Fragments 3 (D,E,F) and 4 (A,B,C) at 1 mM in 1% DMSO. ¹H-¹⁵N HSQC spectra of RGS17 in the presence of 4 (A) and 3 (D), RGS4 in the presence of 4 (B) and 3 (E), and RGS17 C117S in the presence of 4 (C) and 3 (F). Red spectra indicate compound treated and black are 1% DMSO control.

Prioritization of Binding Fragments

Ligand efficiency (LE) was used to prioritize the hit fragments identified, the equation for LE can be found in the methods section of this paper. LE is a metric used to correlate the size of a compound with the observed binding energy, thereby assigning a higher priority for the fragments that make the most productive binding contacts per heavy (non-hydrogen) atom. Prioritizing these fragments by LE was done to aid future medicinal chemistry efforts. An acceptable LE is considered to be 0.3 as a starting point for FBLD ^[22]. The LE for the fragments discovered in this screen ranged from 0.18 for fragment 2 to 0.31 for fragment 4, as shown in Table 1.

Fragment Selectivity and Cysteine Dependence

To evaluate the selectivity and cysteine dependence of the fragments identified in this screen, ¹H-¹⁵N HSQC spectra were obtained for fragments with RGS4-RH or the RGS17-RH C117S mutant. RGS4 was chosen in part because there is significant expression in regions where RGS17 is prevalent, specifically within the brain^[23]. Further, RGS4 belongs to the R4 subfamily of RGS proteins and has little complexity outside of its RGS-RH domain^{[1, 11]}. As a result, RGS4 has been used to compare RGS-RH sequence homology across all RGS subfamilies^[24]. Of the seven identified fragments, only fragment 5 did not bind to RGS4-RH. Fragments 1, 2, 4, 6, and 7 were able to bind to the mutant RGS17-RH C117S (Table 1). During the course of these experiments, it was observed that at high concentrations of fragment 4 there was a considerable loss of observable spectra for RGS17-RH, RGS17-RH C117S, and RGS4-RH (Figure 7A-C). This result is indicative of a loss in structural integrity for these proteins.

Functional Impact of Fragments on RGS17 GAP Activity and Thermal Stability

To this point, efforts have focused on identifying compounds capable of binding RGS17-RH. To translate binding to functional impact, it is necessary to assess the capability of these fragments to inhibit RGS17-RH GAP activity. To assess the correlation of fragment binding to inhibition of RGS17-RH GAP activity, a previously developed phosphate detection assay was used ^[25]. Testing with compounds at 1 mM resulted in only fragment 3 inhibiting the RGS17-RH GAP activity to a statistically significant degree (p ≤ 0.0001, one-way ANOVA with multiple comparisons), although fragment 4 did result in nearly 20% inhibition at 1 mM (Figure 8A). The fragments were further analyzed by differential scanning fluorimetry (DSF) to assess their ability to affect the thermal stability of RGS17-RH. Only fragment 3 was found to alter the melting temperature of RGS17-RH to a statistically significant degree (Figure 8B). Addition of Fragment 4 resulted in a curve shape for which melting temperature could not be accurately determined, therefore, it was removed from analysis.

Figure 8.

GAP Activity and Thermal Stability in Presence of Hits. A) Malachite green GAP analysis of 4.5 μM RGS17 in the absence and presence of 1 mM hit fragment. B) Differential scanning fluorimetry using Sypro Orange to determine melting temperature of 10 μM RGS17 in the presence of 1 mM indicated fragment. Data represent mean of n=3 ± S.E.M. **** p ≤ 0.0001, one-way ANOVA with multiple comparisons.

Identification of a Small-Molecule Binding Site on RGS17

A major advantage of using protein-detected NMR as a screening modality, along with assignment of RGS17-RH ¹H-¹⁵N HSQC backbone resonances, is the determination of fragment binding sites. Mapping the residues that had observable fragment-induced chemical shifts onto the previously solved RGS17-RH crystal structure (1ZV4) revealed most fragments to bind at a site on RGS17-RH that was distant from the Gα-interaction face. As shown in Figure 9, these fragments make contacts with an allosteric region that is near the loop between the α6 and α7 helices of RGS17-RH. It is interesting that all the fragments bind in this general region, and yet within the terminal subdomain of RGS17-RH (α1–3, α8–9) there were no observable chemical shift perturbations.

Figure 9.

The Gα interface and the α helices (α1-α9) are labeled in the top center panel (outlined) and the predicted Gα binding site residues on RGS17 (From left to right) are Ile-143 to Serine-145, Ser-150, His-183 to Arg-184, and Asn-192 shown as blue spheres ^[39]. The lower panels show fragment Binding Sites on RGS17 (PDB:1ZV4) revealed by CSPs indicated as red spheres in the presence of indicated hit fragment as determined by ¹H-¹⁵N HSQC NMR.

Fragment Structure and Purity

The chemical identity and purity of the hit fragments was assessed using 1D NMR and HPLC/MS. The 1D NMR and MS spectra were consistent with the reported fragment structures. The HPLC experiments found the purity for all the fragments was above ~95% (fragment 1 ~96%, fragment 2 ~ 98%, fragment 3 ~99%, fragment 4 ~98%, fragment 5 ~97%, fragment 6 ~98%, fragment 7 ~95%). These data are attached as supplementary figures 1–3.

Discussion

The discovery of RGS17-RH inhibitors has been limited to traditional HTS and has achieved moderate success. Therefore, FBLD was pursued as an alternative approach to HTS. FBLD as an alternative method to HTS sacrifices throughput for quality and quantity of information collected from a single experiment. Use of triple resonance NMR experiments that were performed on ¹⁵N/¹³C-labeled RGS17-RH resulted in the assignment of over 93% of the backbone amide resonances. This in turn allowed for the identification of RGS17-RH binding fragments, as well as their binding sites using chemical shift mapping technique after acquisition of ¹H-¹⁵N HSQC spectra in the presence of individual fragments. Using this approach, 1,000 fragments from the Maybridge Ro3 Core library were screened against RGS17-RH. As noted above, pooling fragments saved considerable amounts of sample and instrumentation costs, all while increasing the assay throughput. Ultimately, seven fragments were identified as capable of binding to RGS17-RH, with a hit rate of 0.7% (Figure 3, Table 1). This hit rate is encouraging as hit rates from fragment screens, greater than 0.1%, serve as an indicator for the druggability of a protein of interest ^[26]. Further, a hit rate of 0.7% only refers to RGS17-RH binding for fragments. Incorporating the GAP activity assay for measuring RGS17-RH inhibition, the hit rate is lowered to 0.2% (Fragments 3 and 4), which is still above the 0.1% threshold. This result, and our pooling approach, should encourage future efforts to expand the number of fragments screened.

After deconvolution of the fragment pools, we prioritized the seven fragments identified by affinity/K_D (Figure 5, Table 1). Due to the limits of fragment solubility, as determined by ¹H-NMR, a K_D could not be determined for fragment 3. However, it was not discarded from further experiments as it caused significant CSPs when soluble. For the fragments where a K_D value was obtained, the values ranged from 0.5 to 5.5 mM (Table 1) which is appropriate for a fragment screen ^[26]. The LE’s for most fragments were < 0.3, which is generally considered to be an acceptable threshold for an initial fragment screening hit that is pursuable in FBLD ^[27], only fragment 4 exceeded this threshold with a LE of 0.31. However, fragment 4 was later found to non-specifically induce structural changes, observed as a loss of signal in ¹H-¹⁵N-HSQC spectra of RGS17-RH, RGS4-RH and RGS17-RH C117S at high fragment concentrations (1–2 mM) (Figure 7). It should be noted that fragment 4 was not inducing structural instability through a cysteine-dependent mechanism. In the contrary, fragment 3 was found to interact with the C117 residue of RGS17-RH in a covalent manner, as titration of fragment 3 induced CSPs in the RGS17-RH HSQC spectrum in a concentration-independent manner. No significant changes were observed when RGS17-RH C117S was titrated with fragment 3, confirming fragment 3 was binding in a cysteine dependent manner. Further, fragment 3 was found to cause structural instability in RGS4-RH, another RGS family member that is known to be sensitive to thiol modification ^[28]. After confirming RGS17-RH binding for the seven fragments, we looked to assess the functional impact for these fragments. As RGS proteins are known for their impact on Gα protein GAP activity, converting GTP to GDP + inorganic phosphate, we employed the malachite green assay, previously developed by the Roman lab, that quantifies free phosphate in solution ^[25]. We used this assay to measure the ability of the fragments to inhibit RGS17-RH’s GAP activity towards Gα_i1 (R178M, A326S). Phosphate release during GTP hydrolysis is attributed to Gα_i1 (R178M, A326S) GTPase activity. In the presence of RGS proteins this rate of hydrolysis is accelerated. Mutations to the Gα_i1 protein reduce its intrinsic GTPase activity (R178M), making GTP hydrolysis more RGS dependent ^[29], and increase Gα_i1’s rate of GDP release and subsequent GTP binding, which is the rate limiting step in Gα protein cycling thereby removing the need for guanidine nucleotide exchange factor (GEF) activity of a GPCR ^[29b]. Fragments 3 and 4 were found to reduce GAP activity of RGS17-RH in the malachite green assay, but only fragment 3 reduced GAP activity to a statistically significant degree (Figure 8A). This finding was expected as fragments 3 and 4 were found to cause considerable structural instability of RGS17-RH in solution. The remaining fragments (1,2,5,6,7) had low affinity for RGS17-RH, with K_D,Avg values ranging from 1.5 mM for fragment 6 to 5.5 mM for fragment 2. The lack of appreciable RGS17-RH GAP inhibition by these fragments (1,2,5,6, and 7) is a likely result of only modest RGS17-RH structural changes detected via NMR. Future work aimed at improving the affinity of these fragments will be critical to the development of these hit fragments into lead molecules capable of significantly inhibiting RGS17-RH GAP activity. This will also aid in the validation of the small-molecule binding pocket near the α6–7 loop as a region capable of conferring allosteric inhibition. The potential for successful improvement of these hit fragments relies on empirical data that details the fragment-protein binding nature.

To this end, the fragments identified in this screen were evaluated using DSF to assess their likelihood of success in future crystallography efforts. DSF is used to determine the melting temperature (T_m) of a protein, and it does so by measuring the change in fluorescence over a range of temperatures for a dye that displays a significant increase in fluorescence when bound to unfolded protein ^[30]. Previously, it has been established that compounds that increase the T_m of a protein have a higher likelihood of yielding high-resolution structures for crystallography, as opposed to compounds that were found to decrease T_m ^[31]. Fragments 3 and 4 were the only fragments found to reduce GAP activity of RGS17-RH, they were also the only fragments that impacted the thermal stability of RGS17-RH. Fragment 3 decreased the T_m RGS17-RH by nearly 11 ˚C. The melting curve obtained for fragment 4 was devoid of an appreciable inflection point, which indicates a decrease in the T_m. Fragments 1,2,5,6, and 7 decreased protein stability, but the decrease T_m was not statistically significant (Figure 8B). The results from our DSF experiments indicate that these fragments are unlikely to yield high resolution crystallography data.

We then determined the identity and location of residues that underwent concentration dependent CSPs from NMR titration to identify a putative fragment binding site on RGS17-RH. We found that each fragment caused changes in the region of the helical domain of RGS17-RH, which is distal to the Gα interface (Figure 9). Previous work has identified this region as a putative small molecule binding site in RGS4-RH ^[32]. Residues on α5, near the small molecule binding site, have been found to mediate the interaction with phosphatidylinositol 3,4,5-triphosphate (PIP₃) and calmodulin (CaM) ^[32]. The interplay between RGS4 and these binding partners is such that PIP₃ inhibits the GAP activity of RGS4, while CaM binding to RGS4 blocks PIP₃ mediated GAP inhibition. In this study, only fragment 5 did not bind both RGS17 C117S-RH and RGS4-RH. It is possible, that the small molecule binding site identified in this study is conserved between these two RGS family members. Assessing the RGS17-RH residues that exhibited fragment-induced CSPs for multiple fragments, we found that E109 (fragments 5–7), L163 (1,2,5–7), N167 (1,4–7), and Q175 (2,4,5) were most commonly impacted by the presence of the identified fragments. The importance of theses residues identified here could be probed further by making point mutations on RGS17-RH to assess the impact their absence has on the fragment binding capabilities or functional impact on RGS-Gα inhibition. The fragments that do not make contacts with one or more of these four residues could be grown with the intention of reaching these residues and improving their inhibition of the RGS17-Gα protein-protein interaction. Evidence to support our findings comes from a recent study that produced the highest resolution crystal structure of RGS17 and identified multiple Ca²⁺ binding sites on RGS17 (PDB: 6AM3) ^[33]. Interestingly, one of the identified Ca²⁺ binding sites consisted of Y106 and E109, which is near the binding site of fragments 5–7. The study found that Ca²⁺ promoted the RGS17-Gα interaction and improved the catalytic efficiency of GTP-hydrolysis by Gα bound to RGS17. The design of fragments that take advantage of the Ca²⁺ binding site/s could yield fragments with improved RGS17-Gα inhibition.

Structural information detailing the nature of a target protein bound to a molecule of interest is a significant goal for any screen. In the context of a fragment-based screen, this information would allow for an iterative process in which fragments are tested for binding and models for fragment-bound target can be generated (experimentally or computationally). Using these models, new and larger compounds with potentially higher affinities for the target protein can be synthesized and tested, allowing the process to repeat itself. This study represents a proof of concept as the fragments identified bind to the target protein, are capable of having a functional impact on the target protein, and their binding sites were determined. As the fragment binding site identified in this study is not fully utilized by any one fragment, it follows that fragments designed to take full advantage of the binding site identified in this study will have improved RGS17-RH affinity and inhibition profiles.

Conclusion

In this study we successfully identified seven fragments capable of binding to RGS17-RH. One fragment exhibited a binding that was cysteine-dependent, and therefore nonspecific for RGS17-RH, while another fragment induced structural instability for all RGS proteins tested (RGS17-RH, RGS17-RH C117S, and RGS4). Only fragment 3 significantly inhibited the GAP activity of RGS17-RH. Using protein backbone resonance assignment experiments and chemical shift mapping via ¹H-¹⁵N HSQC spectra of RGS17-RH, we determined the fragment binding sites. Mapping the residues that underwent fragment-induced chemical shifts to the previously solved RGS17-RH crystal structure (1ZV4) showed that most fragments were binding to a site on RGS17-RH that was distant from the Gα-interaction face. We found that the fragments tested were making contacts with an allosteric region that was near the loop between the α6 and α7 helices. All of the fragments tested bound to this general region, with no appreciable chemical shifts observed within the terminal subdomain. This is a key finding that identifies a binding site for fragments that are capable of binding RGS17-RH. This screen, which consisted of only 1,000 fragments, shows that fragment screening can be used to generate interesting results. Further, with the RGS17 ¹H-¹⁵N HSQC assigned and establishment of screening variables, such as DMSO tolerance for RGS17, future work to expand screening of fragments to larger libraries is now more accessible than ever. As RGS17 remains an attractive drug target, binding information is critical to the development of fragments into larger more complex chemical analogs.

Experimental

Cloning and Molecular Biology

The wild type RGS17-RH protein construct was obtained from Addgene and was a gift from Nicola Burgess-Brown (Plasmid #39141). Using previously described ligation independent cloning techniques, RGS4 residues 51–179 (RGS4-RH) were cloned into pNIC-Bsa4, which was a gift from Opher Gileadi (Addgene #26103) ^[34]. RGS17 C117S was purchased as a G Block (Integrated DNA Technologies, Coralville, IA) and amplified using PCR before being cloned into pNIC-Bsa4 (Addgene #26103), using ligation independent cloning as previously described ^[34]. All constructs were designed with an N-terminal tobacco etch virus (TEV) protease-cleavable 6X-his tag.

Protein Purification

RGS17-RH, RGS4-RH, and RGS17-RH C117S were purified largely as described previously ^[19]. Briefly, RGS protein-coding plasmids were transformed into BL21-CodonPlus(DE3)-RIPL cells and subsequently grown up overnight at 37 ˚C on LB agar (RPMI) plates supplemented with either kanamycin or ampicillin, depending on the selection marker encoded by each plasmid, in addition to chloramphenicol and spectinomycin. The following day, bacterial colonies were selected and seeded into Terrific Broth (supplemented with the proper antibiotic and chloramphenicol) and grown at 37 ˚C with shaking of 275–300 rpm. The growth was stopped when an OD₆₀₀ of 2.0 was reached, measuring with a Biotech Synergy 2 plate reader. Protein production was induced with 0.5–1 mM isopropyl 1-thio-β-D-galactopyranoside (IPTG), incubation of the culture continued for 16 hours at 18 ˚C while shaking of 275–300 rpm. At the end of this incubation time, cells were pelleted and resuspended at 4 ˚C in buffer A (50 mM HEPES, 500 mM NaCl, 1 mM β-mercaptoethanol, 10 mM imidazole at pH 8.0). Cells were then subjected to lysis with lysosome at 1 mg per ml of cell pellet, and DNase I (~2 mg) was added. The lysate was further subjected to multiple freeze-thaw cycles in liquid N₂. Separation of the soluble lysate fraction was achieved by centrifugation at 100,000 x g for 1 hour. Separation of His-tagged proteins from the supernatant was achieved using an immobilized metal affinity chromatography column (Ni-sepharose 6 Fast Flow, GE Healthcare) that was mounted on an ÄKTA FPLC (GE Healthcare). The eluted fractions containing His-tagged RGS protein were treated with His-tagged TEV protease at a molar ratio of ~1:20 TEV:RGS. This reaction was initiated and dialyzed overnight against 5 liters of buffer A at 4 ˚C in order to cleave the His₆ tag. Samples containing RGS protein free of the His₆ tag were again subjected to the immobilized metal affinity chromatography column (Ni-sepharose 6 Fast Flow, GE Healthcare). The flow through was collected and subjected to size-exclusion chromatography (10 mM borate, 500 mM NaCl, and 1 mM DTT at pH 7.0) to obtain RGS protein that was determined to be 99+% pure by SDS-PAGE.

Isotope-labeled (¹⁵N and ¹³C) RGS proteins were purified essentially as above with a few exceptions. Namely, when the culture reached an OD₆₀₀ of 1.5, cells were pelleted at 3,500 x g at 4 ˚C for 15 minutes before being resuspended in an equal volume of M9 minimal medium that was supplemented with 2 g/L D-[¹³C₆] glucose and 1 g/L ¹⁵NH₄Cl for the ¹³C/¹⁵N-labeled sample. For the ¹⁵N labeled sample only 1 g/L of ¹⁵NH₄Cl was added to the M9 minimal medium. The isotope-labeled samples were concentrated to > 1 mM in a buffer containing 20 mM K₂HPO₄, 100 mM NaCl, 0.5 mM β-mercaptoethanol, and 2 mM NaN₃ at a pH of 7.6.

¹H-¹⁵N HSQC Assignment of RGS17

Using a 600 MHz Varian INOVA NMR spectrometer equipped with a triple resonance gradient probe, the following triple resonance experiments: HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO) were completed at 25 ˚C to assign the RGS17-RH backbone (and Cβ) chemical shifts. The collected data was then processed and analyzed using NMR Pipe and CCPNAnalysis, respectively ^[35]. The sample contained 750 μM RGS17-RH in a buffer containing 20 mM K₂HPO₄, 100 mM NaCl, 0.5 mM β-mercaptoethanol, and 2 mM NaN₃ at a pH of 7.6.

Liquid Handling for the NMR Screen: Fragment Pooling

Fragments in the Maybridge Ro3 core set (Thermo Fisher Scientific) had a stock concentration of 100 mM in 100% DMSO. Six fragments at 5 μL each were pooled (30 μL pool final volume) and added to 96 DeepWell (Nunc) plates already containing 337.5 μL of 22 mM potassium phosphate, 110 mM NaCl, 0.55 mM β-mercaptoethanol and 2.2 mM NaN₃ at pH 7.6 using a Hamilton Microlab Star liquid handling system. Plates containing diluted fragment pools were sealed and stored at −80 ˚C.

NMR Fragment Screen with RGS17-RH

On the day of screening, 50 μL D₂O and 82.5 μL of 606 μM ¹⁵N-RGS17-RH were added to each well to achieve a final concentration of 1 mM for each of the six fragments, and 100 μM RGS17-RH for each sample. Samples were transferred to 5 mm NMR tubes (Deuterotubes) and analyzed using a Bruker Avance II 500 MHz spectrometer equipped with a triple resonance gradient probe and B-ACS 60 sample changer. The DMSO tolerance experiments were completed as above but with 0%, 5%, and 10 % final concentration of DMSO. After the initial screen of pooled fragments, deconvolution experiments were performed as above but with 5 μL of individual fragment with 25 μL of DMSO. Collected data was processed and analyzed using NMRPipe and CCPNAnalysis respectively. Spectra were manually examined for fragment-induced changes in chemical shifts.

Assessing Fragment Affinity with NMR and Calculating Chemical Shift Perturbations (CSP)

To determine dissociation constants (K_D), 100 μM ¹⁵N-RGS17-RH was incubated with fragment concentrations from 100 μM up to the fragments limit of solubility (generally 2–10 mM). CSPs were determined by measuring the distance between the centers of the peaks for the control and fragment-treated samples using the following equation,

$$CSP=\sqrt{{\left(\text{Δ}{\delta }_H\right)}^2+(0.101\times {\left(\text{Δ}{\delta }_N\right)}^2}$$

Δδ_H and Δδ_N= difference in chemical shift in the absence and presence of fragment in the indicated dimension (H for hydrogen and N for nitrogen). Using GraphPad Prims 7, CSP values for varying concentration of fragment were fit to a one-site binding model with a correction for ligand depletion in order to determine K_D values for individual residues that exhibited CSP ≥2 S.D. from the mean. The equation that was used is as follows,

$$CSP=\frac{{CSP}_{max}\left(\right(\left[P\right]+\left[F\right]+K_D-\sqrt{({\left(\left[P\right]+\left[F\right]+K_D\right)}^2-4\times [P\left]\right[L]}}{2\left[P\right]}$$

CSP_max is the maximum CSP value observed for the fragment, [F] is the concentration of fragment, [P] is the concentration of protein, and K_D is the dissociation constant. K_D values obtained for residues were averaged to determine the K_D,avg of fragment binding sites.

Prioritizing Fragments with Ligand Efficiency

Ligand efficiency (LE) is a metric used to rank fragments for future efforts for the development of hit compounds into lead molecules through further screening and/or medicinal chemistry. LE may be viewed as the change in free energy per heavy atom in a compound or binding affinity per heavy atom ^[36]. The equation for LE is shown below, where R= ideal gas constant, T= temperature, HA= number of heavy atoms (non-hydrogen), K_D= dissociation constant ^[37].

$$LE=\left(\frac{-2.3039\mathrm{*}R\mathrm{*}T}{HA}\right)\log \left(K_D\right)$$

Determination of Cysteine Reactivity of Fragment 3 with DACM

In black non-binding 384-well plates (Corning), 10 μL of 8 μM RGS17-RH (In buffer: 20 mM K₂HPO₄, 100 mM NaCl, and 2 mM NaN₃ at a pH of 7.6) or buffer alone was mixed with 10 μL buffer, 10 μL of 80 μM fragment 3 or NEM, followed by an incubation at ambient temperature for 30 minutes. Post incubation, 20 μL of 40 μM DACM (Anaspec) was added, and fluorescence was read for 1,000 seconds with a Perkin-Elmer Envision using excitation and emission wavelengths of 385 and 440 nm, respectively. The final mixture consisted of 2 μM RGS17-RH, 20 μM fragment, and 20 μM DACM in the aforementioned phosphate buffer. To determine reactivity the slope of DACM fluorescence for all conditions tested was analyzed. Fluorescence in the absence of protein was normalized to 0% and fluorescence of RGS17-RH with no fragment added was normalized to 100%. One-way ANOVA with multiple comparisons was used to determine statistical significance, analysis was done using GraphPad Prism.

Differential Scanning Fluorimetry analysis of RGS17-RH with Fragments

Experiments were carried out in white 384-well PCR plates (Sorenson) with final concentrations of 10 μM RGS17-RH, 1 mM of indicated fragment or DMSO control, and 2X Spyro Orange dye (ThermoFisher Scientific) in a total reaction volume of 20 μL. Prior to the addition of Spyro Orange, RGS17-RH and fragment were incubated for 30 minutes at ambient temperature. Fluorescence was then measured from 20–80 ˚C using a Roche 480 LightCycler II. Melting temperatures (T_m) were calculated using DMAN, a java tool used for analysis of multi-well DSF experiments ^[38].

GAP Activity assay: Malachite Green

RGS GAP activity in the presence of hit fragments was assessed with the malachite green free PO₄ detection assay as described previously ^[25]. Final concentrations of 300 μM GTP, 4.5 μM RGS17-RH, and 1 mM fragment were used. RGS-RH proteins were serially diluted in half logarithmic concentrations to cover three orders of magnitude (3.16 μM to 3.15 nM), these dilutions allowed for the assessment of RGS-RH domain GAP activity and the determination of each proteins EC₈₀ (concentration of protein that results in 80% of maximal response) concentration. These EC₈₀ concentration were determined immediately after protein purification. Concentration response experiments were then performed with fragments at each proteins determined EC₈₀ concentration. Data for malachite green assays was analyzed using GraphPad Prism 7, data represents mean of n=3 independent experiments ± SEM.

Confirmation of Fragment Structure: 1D NMR

On the day of the run, 5 μL of 200 mM Fragment in deuterated DMSO, 50 μL D2O, and 445 μL buffer (50 mM HEPES, 200 mM NaCl, 1 mM β-mercaptoethanol pH 8.0) were mixed. Then, samples with a final fragment concentration of 2 mM in 500 μL were transferred to 5 mm NMR tubes (Deuterotubes) and analyzed using a Bruker Avance II 500 MHz NMR spectrometer equipped with a triple resonance gradient probe and an automatic sample changer. Individual 1D 1H NMR spectra were collected and analyzed for all fragments with the exception of fragment 3 which was not soluble in this buffer. These acquired 1D 1H NMR spectra are consistent with their respective chemical structures and are shown as supplementary figures.

Fragment Purity: HPLC/MS

On the day of the run, 200 mM Fragment stocks in deuterated DMSO were diluted to 100 μM with 0.1% UHPLC grade formic acid in UHPLC grade water. LC-QTOF-MS analysis was performed on an Agilent 6530 quadrupole time-of-flight mass spectrometer interfaced with a 1260 Series Agilent capillary HPLC system (CA, USA). The separation was carried out on a Zorbax StableBond (SB-C18) column (150mm x 0.5mm, 5μm) (Agilent, CA, USA) at room temperature with mobile phases consisting of 0.1% UHPLC grade formic acid in UHPLC grade water (solvent A) and 0.1% UHPLC grade formic acid in UHPLC grade acetonitrile (solvent B) in the following gradient (A:B): 0 min (97:3) → 10 min (40:60) → 13 min (5:95) → 17 min (5:95) → 17.01 min (97:3) → 30 min (97:3). The flow rate was set at 15μL/min and injection volume was 5 μL. After separation, analytes were eluted into the dual ESI source of the QTOF, operating in positive ionization mode for all analytes with the exception of fragment 4, which was operated in negative ionization mode. The ion source parameters were as follows: nitrogen drying gas temperature 300 °C, nebulizer pressure 35 psig, fragment voltage 175 V, sheath gas temperature 320 °C, capillary voltage 3500 V. The m/z scanning range was 100 to 600 at 1 spectra/s. All acquisition and analysis were performed using Agilent MassHunter B.06.00 software (CA, USA). Data was analyzed using GraphPad Prism 8. The base peak chromatograms (BPC) and MS are shown as supplementary figures. Compound purity was determined by integration of peaks in the BPC of each sample using the equation shown below. In the equation, AUC is area under the curve and AP is all peaks in the chromatogram in question.

$$\mathrm{\%}Purity=100\times \left(\frac{AUCforfragmentpeak)}{\left(AUCforAP\right)-\left(AUCforAPinblank\right)}\right)$$