Fragment Screening of Human Aquaporin 1

International Journal of Molecular Sciences

Article Fragment Screening of Human Aquaporin 1

Janet To and Jaume Torres *

School of Biological Sciences, Nanyang Technological University, 60 Nanyang Drive, Singapore 637551, Singapore; [email protected] * Correspondence: [email protected]; Tel.: +65-6316-2857; Fax: +65-6791-3856

Academic Editor: Kenichi Ishibashi Received: 5 February 2016; Accepted: 18 March 2016; Published: 25 March 2016

Abstract: Aquaporins (AQPs) are membrane proteins that enable water transport across cellular plasma membranes in response to osmotic gradients. Phenotypic analyses have revealed important physiological roles for AQPs, and the potential for AQP water channel modulators in various disease states has been proposed. For example, AQP1 is overexpressed in tumor microvessels, and this correlates with higher metastatic potential and aggressiveness of the malignancy. Chemical modulators would help in identifying the precise contribution of water channel activity in these disease states. These inhibitors would also be important therapeutically, e.g., in anti-cancer treatment. This perceived importance contrasts with the lack of success of high-throughput screens (HTS) to identify effective and speciﬁc inhibitors of aquaporins. In this paper, we have screened a library of 1500 “fragments”, i.e., smaller than molecules used in HTS, against human aquaporin (hAQP1) using a thermal shift assay and surface plasmon resonance. Although these fragments may not inhibit their protein target, they bound to and stabilized hAQP1 (sub mM binding afﬁnities (KD), ˝ with an temperature of aggregation shift ∆Tagg of +4 to +50 C) in a concentration-dependent fashion. Chemically expanded versions of these fragments should follow the determination of their binding site on the aquaporin surface.

Keywords: human aquaporin 1; fragment based drug discovery; surface plasmon resonance; thermal shift; membrane protein

1. Introduction Aquaporins (AQPs) are membrane proteins that permeabilize cellular membranes to water [1–3]. In humans, 13 AQPs have been identified [4], but only AQP0, AQP1, AQP2, AQP4, AQP5 and AQP8 are water-specific (i.e., “orthodox” aquaporins), while the rest are also permeable to glycerol (referred to as “aquaglyceroporins”), although many other substrates for transport across membranes have been reported [5–14]. The structure of AQPs is well conserved across species; each AQP monomer has six α–helical transmembrane (TM) domains, with two “loops” that enter the membrane and face each other through a conserved NPA motif, and where both N- and C-terminal tails are cytoplasmically oriented [15–21]. AQPs form homotetramers, and each monomer in the tetramer functions independently as a water channel [22]. Over the last 18 years, phenotypic analyses have identified physiological and pathological implications for AQP function (reviewed in [23]), e.g., in brain swelling or glandular fluid secretion. The potential for AQP water channel inhibitors in cerebral edema [24–28], water retention [9,29–32], or regulation of eye intraocular pressure (IOP) [33] associated with glaucoma [34], and others, has also been discussed [35–38]. Recently, it has been shown that AQPs are new players in angiogenesis and cancer biology [39]. Indeed, a growing number of reports over the last 10 years hypothesize that it is the increased water flow through AQP channels, induced by local osmotic gradients, that is responsible for the formation of membrane protrusions during cell migration [37,39,40], a general phenomenon

Int. J. Mol. Sci. 2016, 17, 449; doi:10.3390/ijms17040449 www.mdpi.com/journal/ijms Int. J. Mol. Sci. 2016, 17, 449 2 of 20 observed for several cell types and different AQPs, e.g., AQP4 [41], AQP1 [42], AQP3 [43,44] or AQP9 [45–48]. Several AQPs are upregulated in many types of cancer [49], and expression correlates with a higher tumor grade [50]. Specifically for AQP1, its expression in liver endothelial cells contributes to amoeboid invasion in liver angiogenesis and cirrhosis [51]. AQP1 is abnormally expressed in tumor endothelial cells [52,53], whereas AQP1 deletion in knockout mice greatly impaired tumor growth, angiogenesis and cell migration [52]. AQP1-expressing tumors have more metastatic potential versus AQP1-null tumors [54]. siRNA against AQP1 reversed tumor growth in a murine model of melanoma [55]. Despite these supporting data, the precise role of AQP1 is not yet clear, as interactions of AQPs with the cytoskeleton or other proteins [56,57] may be at least as important as increased water flow in the formation of cell protrusions. The precise role of aquaporins in cell migration is still elusive without the availability of specific, potent and non-toxic water channel inhibitors of AQPs. In addition, hAQP1 inhibitors would have obvious potential therapeutic implications to complement current anti-cancer treatments, e.g., anti-VEGF therapy, where resistance to anti-VEGF treatment has been described [58–61]. Despite the important therapeutic implications, to our knowledge, not a single organic molecule has been reported in the literature to be effective in the inhibition of AQP1 using standard assays, e.g., red blood cells, transepithelial assays or AQP1-proteoliposomes, even less at nM concentrations. For example, current reported AQP1 or AQP4 inhibitors are sulfhydryl-reactive mercurials, e.g., HgCl2 [51,62,63], heavy metals [63–65], quaternary ammonium salts, e.g., tetraethylammonium (TEA+) [66–68], inorganic salts [69], loop diuretic co-transporter blockers [70], pan-inhibitor of carbonic anhydrase acetazolamide [71–73], or TGN-020 (2-(nicotinamoyl)-1,3,4-thiadiazole) [74]. However, many of these compounds are either toxic, not potent, lack specificity, are unsuitable for drug discovery or their inhibitory effect has been heavily disputed [75,76]. Conventional high-throughput screening (HTS) campaigns to search for inhibitors have not been successful. In a reported 100,000 compound screen by the Verkman lab, no significant water channel inhibitors were found [39]. Compared to other common membrane proteins, such as enzymes, whose active sites can act as a binding pocket for drug modulators, the natural architecture of the aquaporin water channel is conceptualized to have a “slippery” hourglass shape that has a narrow pore diameter, restricting the anchoring of drug modulators. Despite this, some aquaporins have been shown to display gating mechanisms that could be druggable, e.g., spinach plasma membrane aquaporin SoPIP2;1 [77], where dephosphorylation of two serine residues under drought stress can trigger occlusion of the pore. In addition, AQP4 water permeability is regulated in a similar way, by changing the phosphorylation state of Ser-111 and Ser-180, on B and D loops [78–80], although these gating mechanisms have been disputed [81–83]. Finally, for AQP0, calmodulin (CaM) binds to its cytoplasmic C-terminal domain in a Ca2+-dependent manner [84], which results in the closure of its cytoplasmic gate and inhibition of water permeability [85]. An alternative to HTS is fragment-based drug discovery (FBDD) [86,87], where highly sensitive and robust biophysical techniques [88] are required for the identification of small molecule “fragments” (~120 to 250 Da), which generally bind weakly to their targets, with affinities typically in the range of 10 µM–1 mM. Although these “fragments” do not necessarily inhibit their protein target, they may constitute good starting points that can be derivatized or linked to other fragments to produce a suitable inhibitor. In addition, owing to their structural simplicity, fragments frequently participate in better-quality interactions as compared to drug-sized compounds [89]. Fragments have lead-likeness properties [90] so that properties can be easily optimized to improve drug potency, are more efficient in exploring the binding sites of proteins [86,89] and have higher binding energy per unit molecular mass than larger compounds. Finally, libraries are vastly smaller than those used in HTS campaigns. The selection of fragments is usually followed by structural determination of binding sites and the mode of action. Surface plasmon resonance (SPR) produces quantitative binding information to rank top binders by affinity [89] with a consumption of protein at least 10- to 100-fold lower than for other biochemical and biophysical screening methods for fragments [91]. Int. J. Mol. Sci. 2016, 17, 449 3 of 20

During the past five years, attempts to screen fragment ligands for integral membrane proteins using biophysical methods, e.g., SPR or target immobilized NMR screening (TINS), have been described. For example, for E. coli DsbB [92], thermostabilized (StaR) G protein-coupled receptors (GPCRs) [93,94], or fatty acid amide hydrolase (FAAH) [95] using 19F NMR (n-fluorine atoms for biochemical screening, n-FABS). However, these applications are still scarce due to the difficulty in the handling of membrane proteins, which has required in many cases the use of fusion constructs or thermostabilized mutants. The present paper uses the native unmodified form of hAQP1 and is the first reported application of biophysical fragment screening in aquaporins of any kind. We have used SPR and thermal shift [96] assays. The latter was performed as a differential static light scattering (DSLS) mode, which is suitable for studying membrane proteins in detergent micelles [97,98].

2. Results

2.1. Expression and Purification of Human Aquaporin 1 (hAQP1) from Insect Cells The recombinant hAQP1 with an N-terminal histidine tag and tobacco etch virus (TEV) protease cleavage site (6His-TEV-hAQP1) was present in the soluble fraction of the cell lysate, thereby permitting direct protein extraction in the presence of detergent n-octyl-β-D-glucopyranoside (OG). Typically, 16–18 g of cell pellets can be obtained per liter culture. Protein purification was completed using nickel-affinity chromatography followed by gel filtration using fast protein liquid chromatography (FPLC). The fractions from the nickel-affinity column were analyzed by SDS-PAGE (Figure1A). The recombinant hAQP1 (monomeric size of 31,136 Da) migrated as a monomer in SDS, although high molecular weight oligomers or contaminants were also present. Eluted fractions were concentrated and subjected to further purification by gel filtration. From the chromatogram profile obtained, hAQP1 eluted as a single major peak at 70 mL (Figure1B, arrow), which represents the tetrameric state of the hAQP1, i.e., a molecular weight of ~124 kDa. Therefore, the recombinant hAQP1 obtained is monodisperse and mainly exists as a tetramer in presence of OG. A minor peak in the chromatogram, representing higher molecular weight aggregates of the protein, eluted at the void volume (V0) of the gel filtration column, around 45 mL. All the fractions collected from the major eluted peak contained hAQP1 (Figure1C). In this SDS-PAGE gel, hAQP1 migrated as a monomer, with estimated sample purity of ~95%. Typically, a yield of 5–6 mg of purified hAQP1 could be obtained per liter of culture. The sample in OG was analyzed by blue-native polyacrylamide gel electrophoresis (BN-PAGE) (Figure1D) [ 99]. In this case, interference of the dye and the presence of detergent contribute to the overall size of the protein complex and can result in an additional mass with a factor of ~1.8 [100]. For comparison, we included as internal membrane protein standard purified E. coli aquaporin Z (AqpZ), shown to form tetramers in OG with apparent molecular weight of ~170 kDa (expected 100 kDa) [101]. Based on the above, the hAQP1 band at ~242 kDa should correspond to its tetramer (expected 124 kDa). The hAQP1 band at ~480 kDa is assigned to its octamer (expected 248 kDa), since an octamer is also observed for AqpZ at ~400 kDa.

2.2. Water Permeability of Puriﬁed hAQP1

In response to an osmotic shock, the calculated osmotic water permeability coefﬁcient (P f ) of the hAQP1-proteoliposomes (P f = 0.053 cm/s) was almost 40 times greater than that of the control liposomes (P f = 0.0015 cm/s) (Figure2). This enhancement is comparable to that described in the literature [102], demonstrating that hAQP1 puriﬁed from insect cells is functional. hAQP1 was also sensitive to inhibition by mercuric chloride, a known potent inhibitor of AQP1 [63]. Osmotic water permeability of hAQP1 proteoliposomes pre-incubated with 300 µM HgCl2 was reduced by almost 15 times (P f = 0.0041 cm/s) (Figure2). A similar value has been observed in the literature [ 75], supporting that the hAQP1 produced here is active and can be inhibited. Int. J. Mol. Sci. 2016, 17, 449 4 of 20

Int. J. Mol. Sci. 2016, 17, 449 4 of 19

Figure 1. Purification and oligomeric status of human aquaporin 1 (hAQP1). (A) fraction analysis of FigureNi-NTA 1. Purification purification and profile oligomeric in SDS-PAGE status ofof humanhAQP1 aquaporinfrom Sf9 cells: 1 (hAQP1). Lys, whole (A cell) fraction lysate; analysisFT, of Ni-NTAflow-through; purification W, wash; profile E, eluent. in SDS-PAGE The arrow ofpoin hAQP1ts to the from purified Sf9 cells:hAQP1 Lys, monomer whole atcell ~31 lysate; kDa; FT, flow-through;(FigureB) gel 1.filtration Purification W, wash; profile E,and eluent. ofoligomeric hAQP1 The purification.status arrow of pointshuman The toaquaporin major the purifiedpeak 1 (hAQP1). eluted hAQP1 at (A 70) monomerfraction mL (arrow) analysis at ~31and of kDa;

(B) gelcorrespondsNi-NTA filtration purificationprofile to the oftetrameric profile hAQP1 in purification.hAQP1 SDS-PAGE (~124 Theof kD hAQP majora). The1 peakfrom minor elutedSf9 peak cells: at at 70Lys, 45 mL mLwhole (arrow) (void cell andvolume, lysate; corresponds FT,V0) to thecontainedflow-through; tetrameric aggregates hAQP1 W, wash; (~124of hAQP1;E, kDa).eluent. (C The The) SDS-PAGE minorarrow peakpoin of tsthe at to 45fractions the mL purified (void from volume, hAQP1 62 to 78 monomerV mL0) contained eluted at in~31 the aggregates kDa; gel of hAQP1;filtration(B) gel (C filtration )chromatography. SDS-PAGE profile of theof The hAQP1 fractions protein purification. am fromount 62 loaded to The 78 mLper major lane eluted peakwas in 5 eluted theµg; geland at filtration(D 70) Blue mL Native chromatography.(arrow) PAGE and The protein(BN-PAGE)corresponds amount ofto purified the loaded tetrameric hAQP1 per lane inhAQP1 100 was mM (~124 5 nµ-octyl-g; kD anda).β -D (TheD-glucopyranoside) Blueminor Native peak at PAGE(OG) 45 mL and (BN-PAGE) (void purified volume, AqpZ of purifiedV as0) ancontained internal aggregates control. The of hAQP1hAQP1; band (C) SDS-PAGE at ~242 kDa of corresponds the fractions to from tetramers 62 to 78(tet, mL expected eluted in124 the kDa) gel hAQP1 in 100 mM n-octyl-β-D-glucopyranoside (OG) and purified AqpZ as an internal control. The andfiltration band chromatography. at ~480 kDa corresponds The protein to its am octamerount loaded (oct, expected per lane 248was kDa). 5 µg; and (D) Blue Native PAGE hAQP1 band at ~242 kDa corresponds to tetramers (tet, expected 124 kDa) and band at ~480 kDa (BN-PAGE) of purified hAQP1 in 100 mM n-octyl-β-D-glucopyranoside (OG) and purified AqpZ as corresponds to its octamer (oct, expected 248 kDa). an internal control. The hAQP1 band at ~242 kDa corresponds to tetramers (tet, expected 124 kDa) and band at ~480 kDa corresponds to its octamer (oct, expected 248 kDa).

Figure 2. hAQP1 water permeability functional assay. Normalized light scattering intensity changes of hAQP1—E. coli lipid proteoliposomes (blue trace) compared with no-AQP liposome control (black trace) in response to an osmotic shock, and inhibition of hAQP1 proteoliposomes by mercuric chlorideFigure 2. ( redhAQP1 trace). water The permeability average fitted functional rate constants assay. ( Normalizedk) (means ± S.E.,light n scattering = 6) measured intensity at 23 changes °C for Figure 2. hAQP1 water permeability functional assay. Normalized light scattering intensity changes hAQP1-proteoliposomes,of hAQP1—E. coli lipid proteoliposomesproteoliposomes (inblue presence trace) comparedof 300 µM withHgCl no-AQP2 and no-AQP liposome liposomes control E. coli of hAQP1—were(black 109.4 trace) (±1.9), in responselipid 8.56 (±0.03) proteoliposomes to an and osmotic 2.27 (±0.04) shock, (blue sand−1, respectively.trace) inhibition compared of hAQP1 with proteoliposomes no-AQP liposome by mercuric control (blackchloridetrace) in(red response trace). The to anaverage osmotic fitted shock, rate constants and inhibition (k) (means of hAQP1 ± S.E., n proteoliposomes= 6) measured at 23 by °C mercuric for ˝ chloride hAQP1-proteoliposomes, (red trace). The average proteolipos ﬁttedomes rate in constants presence (kof) (means300 µM ˘HgClS.E.,2 andn = no-AQP 6) measured liposomes at 23 C for hAQP1-proteoliposomes,were 109.4 (±1.9), 8.56 (±0.03) proteoliposomes and 2.27 (±0.04) s in−1, respectively. presence of 300 µM HgCl2 and no-AQP liposomes were 109.4 (˘1.9), 8.56 (˘0.03) and 2.27 (˘0.04) s´1, respectively.

Int. J. Mol. Sci. 2016, 17, 449 5 of 20

2.3. Hit Binding Measured by Surface Plasmon Resonance (SPR) Once validated the purity and activity of the recombinant hAQP1 obtained from the insect cell expression, the fragment library was screened against hAQP1 using SPR. As described in the Methods section, hAQP1 was immobilized by amine coupling to the carboxymethylated dextran matrix, thereby enabling the protein to move freely [103]. The SPR response depends on the change in mass concentrationInt. J. Mol. Sci. 2016,on 17, 449 surface of the sensor. The latter, in turn, depends on the molecular5 of 19 weight of the analyte in relation to the number of ligands on the sensor surface. Rmax describes the maximal 2.3. Hit Binding Measured by Surface Plasmon Resonance (SPR) binding capacity of a captured ligand (hAQP1) for the analyte (fragment in the library) in response Once validated the purity and activity of the recombinant hAQP1 obtained from the insect cell units, and the theoretical Rmax can be calculated according to: expression, the fragment library was screened against hAQP1 using SPR. As described in the Methods section, hAQP1 was immobilized by amine coupling to the carboxymethylated dextran matrix, thereby enablingRmax the“ proteinMW analyteto move{M freelyWligand [103].ˆ TheRligand SPR responseˆ Vligand depends on the change (1) in mass concentration on surface of the sensor. The latter, in turn, depends on the molecular weight where MWofanalyte the analyteand inM relationWligand to theare numb theer molecular of ligands on weights the sensor of surface. the analyteRmax describes (fragments) the maximal and ligand (hAQP1) molecules,binding capacity respectively; of a captured Rligandligand (hAQP1)is the responsefor the analyte obtained (fragment from in the hAQP1; library) in andresponseVligand is the valency ofunits, hAQP1/proposed and the theoretical stoichiometry Rmax can be calculated of the according interaction to: [104]. To generate good and reliable SPR data for studies of protein:small moleculesRmax = MW interactions,analyte/MWligand × the RligandR max× Vligandshould ideally be ~25 response(1) units (RU), assumingwhere 100%MWanalyte protein and M biologicalWligand are the activity molecular and 1:1weights binding of th mode.e analyte Taking (fragments) a 120-Da and fragmentligand as the smallest analyte(hAQP1) in molecules, this experiment, respectively; the R theoreticalligand is the response RU of obtained hAQP1 from to be hAQP1; immobilized and Vligand is is approximately the valency of hAQP1/proposed stoichiometry of the interaction [104]. To generate good and reliable 6500 RU (calculated using the above equation: 25 RU = (120/31,136) ˆ Rligand ˆ 1). However, since it is SPR data for studies of protein:small molecules interactions, the Rmax should ideally be ~25 response unlikely thatunits the (RU), full assuming protein surface100% protein will bebiological active, activity we immobilized and 1:1 binding the proteinmode. Taking in excess, a 120-Da to 12,000 RU, to ensure thatfragment we detectas the smallest meaningful analyte bindingin this experime signals.nt, the SPR theoretical requires RU very of hAQP1 high to coupling be immobilized densities of the target, typicallyis approximately 5000 to 10,0006500 RU RU (calculated [103]. using the above equation: 25 RU = (120/31,136) × Rligand × 1). However, since it is unlikely that the full protein surface will be active, we immobilized the protein In our experiment, the hAQP1 protein immobilized onto the SPR biosensor chip yielded 9000 RU. in excess, to 12,000 RU, to ensure that we detect meaningful binding signals. SPR requires very high During thecoupling course densities of screening, of the target, control typically injections 5000 to 10,000 of RU two [103]. positive control binders of hAQP1 [97], NSC624169 (MInW our: 357) experiment, and NSC226080 the hAQP1 (proteinMW: 914) immobili werezed conducted onto the SPR at biosensor regular chip intervals yielded (approximately 9000 once everyRU. 20 During compounds, the coursei.e. of ,screening, the start, control mid injections and end of cyclestwo positive relative control to binders each assay of hAQP1 plate). [97], This was NSC624169 (MW: 357) and NSC226080 (MW: 914) were conducted at regular intervals done to monitor possible surface deterioration and subsequent non-speciﬁc binding, which may result (approximately once every 20 compounds, i.e., the start, mid and end cycles relative to each assay in false positives.plate). This These was done positive to monitor controls possible have surface molecular deterioration weights and subsequent comparable non-specific to the average binding, molecular weight of thewhich fragments may result to in befalse screened, positives. andThese also positive gave controls reproducible have molecular binding weights afﬁnities comparable (KD) to values. The percentagethe occupancy average molecular of each weight fragment of the fragments to the immobilized to be screened, protein and also was gave plotted reproducible against binding the injection affinities (KD) values. The percentage occupancy of each fragment to the immobilized protein was cycle number (Figure3). plotted against the injection cycle number (Figure 3).

Figure 3. Fragment screening against hAQP1 protein using Surface plasmon resonance (SPR). Figure 3. FragmentScreening data screening adjusted against from response hAQP1 unit protein to percentage using surface Surface occupancy, plasmon based resonance on equilibrium (SPR). Screening data adjustedbinding from response response values unit(Req) toobtained percentage from the surface Biacore occupancy,T-200 instrument based for the on 1500-fragment equilibrium binding response valueslibrary. (EachReq )cycle obtained number from represen thets Biacore a different T-200 fragment. instrument Each color for group the 1500-fragmentrepresents individual library. Each cycle numberscreening represents plates (total a different of 19). Screening fragment. proceeded Each for color 15 days, group and represents data obtained individual per screening screening plate plates was adjusted so that the baseline scattered evenly around zero percent occupancy. The relative (total of 19).binding Screening occupancy proceeded of the positive for 15 control, days, NSC624169, and data obtainedinjected at per200 µM screening every 20 platecycles, was is also adjusted so that the baselineshown (black scattered triangles). evenly around zero percent occupancy. The relative binding occupancy of the positive control, NSC624169, injected at 200 µM every 20 cycles, is also shown (black triangles). Int. J. Mol. Sci. 2016, 17, 449 6 of 20

Int. J. Mol. Sci. 2016, 17, 449 6 of 19 The data in Figure3 was used to identify top fragments with good and clean binding to hAQP1. Int. J. Mol. Sci. 2016, 17, 449 6 of 19 These primaryThe data hits in Figure were conﬁrmed3 was used to using identify secondary top fragments SPR with experiments good and clean in order binding to eliminateto hAQP1. false positivesTheseThe [ 103primary data]. A in totalhits Figure were of 223 wasconfirmed compounds used to usingidentify were secondary top chosen fragments SPR for aexperiments with second good stage and in cleanorder dose-response binding to eliminate to hAQP1. assay false to be performedpositivesThese atprimary 25,[103]. 50, Ahits 100 total were and of 20022confirmed compoundsµM. This using secondary were secondary chos assayen forSPR identiﬁeda experimentssecond stage 10 binders indose-response order (Figure to eliminate assay4), which tofalse be were furtherperformedpositives assayed [103]. at in 25, a A concentration 50, total 100 of and 22 compounds200 range µM. This from were secondar 6.25 chos toen 800y assayforµ Ma secondidentified in 2-fold stage dilutions10 dose-responsebinders to (Figure study assay 4), their which to bindingbe wereperformed further at assayed 25, 50, 100in a and concentration 200 µM. This range secondar from 6.25y assay to 800 identified µM in 2-fold 10 binders dilutions (Figure to study 4), which their kinetics and KD values (Figure5). bindingwere further kinetics assayed and K inD valuesa concentration (Figure 5). range from 6.25 to 800 µM in 2-fold dilutions to study their binding kinetics and KD values (Figure 5).

Figure 4. Top 10 fragment binders against hAQP1 discovered using SPR screening. Figure 4. Top 10 fragment binders against hAQP1 discovered using SPR screening. Figure 4. Top 10 fragment binders against hAQP1 discovered using SPR screening.

Figure 5. SPR sensorgrams of the top 10 hit fragments against hAQP1 shown in Figure 4. The y-axis representsFigure 5. SPR response sensorgrams unit (RU) of andthe topthe 10x-axis hit fragmentrepresentss againsttime (s). hAQP1 Each fragment shown in was Figure tested 4. inThe a 2-foldy-axis Figure 5. SPR sensorgrams of the top 10 hit fragments against hAQP1 shown in Figure4. The y-axis dilutionrepresents from response 6.25 to unit 800 (RU) µM. and The the steady-state x-axis represents signals time plotted (s). Eachas a fragmentfunction ofwas concentration tested in a 2-fold was representsgloballydilution response fittedfrom to6.25 unita typicalto (RU)800 1:1µM. and steady-state The the steady-statex-axis bi representsnding si model,gnals timeplotted giving (s). asthe Each a binding function fragment affinities of concentration was ( testedKD) based in was aon 2-fold 50% occupancy. Experimentalµ data has a different color for each sample, whereas fitted curve is dilutionglobally from fitted 6.25 to to a 800typicalM. 1:1 Thesteady-state steady-state binding signals model, plotted giving the as binding a function affinities of concentration (KD) based on was globallyshown50% fitted occupancy. in orange. to a typical Experimental 1:1 steady-state data has bindinga different model, color givingfor each the sample, binding whereas affinities fitted (K curveD) based is on 50% occupancy.shown in orange. Experimental data has a different color for each sample, whereas fitted curve is shown

in orange. Int. J. Mol. Sci. 2016, 17, 449 7 of 20 Int. J. Mol. Sci. 2016, 17, 449 7 of 19

TheThe KKD values were in the range of high micromolar to low millimolar affinity, affinity, whichwhich isis typicaltypical ofof small molecule fragments. AC42399, AC42399, MO00656 MO00656 and and CC53013 CC53013 exhibited exhibited stoichiometric stoichiometric binding binding to tothe the target target (Figure (Figure 5),5), with with concentration-dependen concentration-dependentt binding. binding. The The traces traces were were fittedfitted toto aa one-siteone-site bindingbinding modelmodel and and were were characterized characterized by a fast-on,by a fast-on, fast-off fast-off kinetics. kinetics. These are These well-behaved are well-behaved fragments thatfragments are good that candidates are good for candidates lead generation for lead [103 generation]. In particular, [103]. both In particular, MO00656 andboth CC53013 MO00656 have and a trifluoromethylCC53013 have a functional trifluoromethyl group functional (Figure4), suggestinggroup (Figure that 4), this suggesting chemical moietythat this may chemical be beneficial moiety formay the be interaction beneficial withfor the hAQP1. interaction For the with rest ofhAQP the fragments,1. For the bindingrest of the was fragments, generally observedbinding aswas a slowgenerally rise toobserved equilibrium as a inslow most rise cases to equilibrium and a slow in decay most back cases to and the a baseline. slow decay However, back to in the some baseline. cases, resultsHowever, were in compatiblesome cases, with results concentration-dependent were compatible with aggregation. concentration-dependent For example, foraggregation. CC40813 andFor AC34875,example, bindingfor CC40813 did not and achieve AC34875, saturation binding equilibrium did not achieve at higher saturation concentrations, equilibrium and dissociation at higher wasconcentrations, not complete and at dissociation the end of the was run not (Figure complete5). at the end of the run (Figure 5).

2.4. Hit Confirmation Using Thermal Aggregation Shift Assay 2.4. Hit Confirmation Using Thermal Aggregation Shift Assay InIn aa complementarycomplementary strategy,strategy, hitshits werewere testedtested forfor theirtheir effectseffects onon thethe thermalthermal stabilitystability ofof hAQP1 usingusing differentialdifferential staticstatic lightlight scatteringscattering (DSLS), which measures the aggregation of the proteinprotein inin responseresponse toto increasingincreasing temperature.temperature. SinceSince thisthis assayassay cancan bebe performedperformed inin aa 384-well384-well plateplate format,format, thethe 6666 toptop bindersbinders identifiedidentified fromfrom SPRSPR experimentsexperiments werewere tested.tested. As thethe fragmentsfragments are expectedexpected toto bindbind withwith weakweak affinityaffinity toto thethe targettarget protein,protein, thethe primaryprimary screeningscreening concentrationconcentration was set to 11 mM,mM, andand ˝ fragments that led to an increase of the aggregation temperature (Tagg) of hAQP1 of 1 C or more fragments that led to an increase of the aggregation temperature (Tagg) of hAQP1 of 1 °C or more werewere consideredconsidered forfor furtherfurther examination.examination. SevenSeven fragmentsfragments improvedimproved thethe thermostabilitythermostability of hAQP1hAQP1 (Figure(Figure6 6A,A, red red dots), dots), and and are are shown shown in in Figure Figure6 B.6B. Out Out of of these these seven seven hits, hits, CC38113, CC38113, MO08607 MO08607 and and CC46109CC46109 werewere foundfound toto overlapoverlap withwith thethe toptop 1010 hitshits inin thethe SPRSPR experimentexperiment (Figure(Figure6 B,6B, boxed). boxed).

Figure 6. Tagg of hAQP1 in the presence of hit fragments. (A) scatterplot of Tagg values of hAQP1 in Figure 6. Tagg of hAQP1 in the presence of hit fragments. (A) scatterplot of Tagg values of hAQP1 in the presence of each of the 66 top fragment binders identified from SPR assay, screened here at 1 mM the presence of each of the 66 top fragment binders identified from SPR assay, screened here at 1 mM byby differentialdifferential staticstatic lightlight scattering.scattering. Each point is an average of three independent measurements. FragmentsFragments which stabilized hAQP1 by 1 ˝°CC or more are indicated as red dots, whereas the rest areare shownshown inin blue;blue; ((B)) structuresstructures ofof thethe fragmentsfragments identifiedidentified inin ((A).). CompoundsCompounds alsoalso identifiedidentified asas thethe toptop 1010 bindersbinders byby SPRSPR appearappear insideinside aa box.box.

The thermal denaturation curves in presence of these seven fragments (Figure 7A) show that The thermal denaturation curves in presence of these seven fragments (Figure7A) show that they they have a stabilizing effect, i.e., a positive shift in the Tagg of hAQP1. A non-binder fragment have a stabilizing effect, i.e., a positive shift in the T of hAQP1. A non-binder fragment included included as a negative control, TL00706, did not haveagg a significant effect on hAQP1 thermal stability. as a negative control, TL00706, did not have a signiﬁcant effect on hAQP1 thermal stability. One of One of the fragments, CC46109, increased the Tagg of the protein by ~50 °C when tested at 1 mM (Figure 7B). However, the increase in Tagg in this case is an estimate, since the protein denaturation curve did not achieve saturation in the temperature range of the experiment.

Int. J. Mol. Sci. 2016, 17, 449 8 of 20

˝ the fragments, CC46109, increased the Tagg of the protein by ~50 C when tested at 1 mM (Figure7B). However, the increase in Tagg in this case is an estimate, since the protein denaturation curve did not achieve saturation in the temperature range of the experiment. Int. J. Mol. Sci. 2016, 17, 449 8 of 19

Figure 7. Thermal denaturation curves of hAQP1. (A) assays performed in the presence of 1 mM of Figure 7. Thermal denaturation curves of hAQP1. (A) assays performed in the presence of 1 mM of each of the seven stabilizing hit fragments shown in (B). Light scattering intensities are plotted as a each of the seven stabilizing hit fragments shown in (B). Light scattering intensities are plotted as a function of temperature and fitted to the Boltzmann equation by non-linear regression to obtain the function of temperature and ﬁtted to the Boltzmann equation by non-linear regression to obtain the temperature of aggregation, Tagg. Each curve is a representative of three independent experiments temperature of aggregation, Tagg. Each curve is a representative of three independent experiments with with similar results; (B) average Tagg values of purified hAQP1 in the absence and presence of similar results; (B) average Tagg values of puriﬁed hAQP1 in the absence and presence of hit fragments. hit fragments.

Next, we examinedexamined whetherwhether these sevenseven hitshits stabilizestabilize hAQP1hAQP1 inin aa dose-dependentdose-dependent manner.manner. InIn thisthis secondarysecondary assay,assay, the the protein protein was was incubated incubated with wi increasingth increasing concentration concentration of each of fragmenteach fragment prior toprior the to protein the protein thermostability thermostability assay. Forassay. five For fragments, five fragments, the increase the increase in thermal in thermal stability stability of the hAQP1 of the proteinhAQP1 wasprotein clearly was dependentclearly dependent on fragment on fragment concentration concentration (Figure (Figure8). These 8). These results results confirmed confirmed the validitythe validity of these of these fragments fragments as true as binders true binders to hAQP1. to hAQP1. The results Theof results the other of the two other fragments, two fragments, MO08607 andMO08607 SB01744, and were SB01744, less consistent were less and consistent are not shown.and are Innot the shown. case of In MO08607, the case the of stabilizationMO08607, the of ˘ ˝ hAQP1stabilization was muchof hAQP1 better was than much in the better initial than screening in the whereinitial Tscreeningagg was 59.0 where0.3 TaggC was at 1 59.0 mM ± (Figure 0.3 °C7 at). ˝ µ Here,1 mM T(Figureagg of hAQP17). Here, was Tagg 61.8 of hAQP1˘ 0.6 Cwas in presence61.8 ± 0.6 of°Cjust in presence 50 M of of the just compound, 50 µM of the and compound, increased ˝ ˝ toand 66.3 increased˘ 0.1 C to and 66.3 86.7 ± 0.1˘ 7.2 °C Cand in the86.7 presence ± 7.2 °C ofin 400 the and presence 2000 µ M.of 400 Therefore, and 2000 although µM. Therefore, showing inconsistentalthough showing results, inconsistent MO08607 can results, be considered MO08607 a ca truen be binder. considered For SB01744, a true thebinder.Tagg Forof hAQP1 SB01744, at differentthe Tagg of concentrations hAQP1 at different of the fragment concentrations could not of be th interpretede fragment due could to inconsistencies not be interpreted in the results,due to suggestinginconsistencies that in this the fragment results, issugges a falseting positive, that this or thatfragment its interaction is a false withpositive, hAQP1 or that is non-specific. its interaction with hAQP1 is non-specific. As indicated in the introduction, the binders found should not necessarily show inhibition. Nevertheless, we performed stopped-flow assays to test the effects of the seven fragments in Figure 6 on hAQP1 water permeability using human red blood cells. Based on the values of the rate constants of the scattering increase, none of these fragments produced a significant reduction in water permeability when tested at 1 mM (not shown). Several factors may account for this. For example, binding does not occlude the pore as it may take place in extramembrane regions, or far from the lumen of the pore. In addition, hAQP1 in red blood cells is highly glycosylated compared to the hAQP1 purified from insect cells, and this may affect binding affinity. Another factor is the

Int. J. Mol. Sci. 2016, 17, 449 9 of 19

orientation of the protein in the membrane, as in theory only one side of the protein (extracellular) is exposed to the drug in RBC assays, whereas both extramembrane domains are exposed when the protein is solubilized in micelles. Another factor, also noted by Seeliger et al. [76], is the high density of AQP1 at the membrane of red blood cells. However, there was a large amplitude reduction in the traces for compounds CC38113 and CC46109, both highlighted in Figure 6A. It is interesting that we found previously a drug-like compound, NSC658139, that led to a strong reduction in overall amplitude [97]. This compound did not reduce the rate constant in the RBC assays but stabilized hAQP1 by as much as 13 °C at 30 µM. We are at present unable to explain this behavior or relate it to Int. J. Mol. Sci. 2016, 17, 449 9 of 20 inhibition of hAQP1.

FigureFigure 8. 8.Dose-dependency Dose-dependency ofof hit hit fragments fragments on onT Taggagg ofof hAQP1.hAQP1. ((AA)) thermalthermal denaturationdenaturation curvescurves ofof hAQP1hAQP1 inin thethe presence presence ofof increasing increasing concentrations concentrations ofof eacheach ofof the the five five selected selected hit hit fragments. fragments. LightLight scatteringscattering intensitiesintensities areare plotted as a a function function of of te temperaturemperature and and fitted fitted to to the the Boltzmann Boltzmann equation equation by bynon-linear non-linear regression regression to to obtain obtain the the temperature temperature of of aggregation, TTaggagg;;( (B) tabletable ofof thethe average averageT aggTagg valuesvalues ofof purifiedpurified hAQP1hAQP1 inin thethe presence of hit fr fragmentsagments at at increasing increasing concentrations. concentrations. Each Each value value is isthe the average average of of three three independen independentt measurements. measurements. These These values values are are plotted plotted as aa histogramhistogram ((rightright),), wherewhere colors colors represent represent drug drug concentrations concentrations shown shown in in Figure Figure8A. 8A.

3. DiscussionAs indicated in the introduction, the binders found should not necessarily show inhibition. Nevertheless,To our weknowledge, performed the stopped-flow data reported assays herein to test is the the effects first offragment-based the seven fragments drug in discovery Figure6 onapproach hAQP1 waterto identify permeability binders usingto an humanaquaporin red bloodusing cells.SPR and Based thermal on the valuesshift assays. of the rateThe constantsuse of these of thebiophysical scattering techniques increase, none that of can these detect fragments weak produced intermolecular a significant interactions reduction is injustified water permeabilitybecause the whenfragments tested screened at 1 mM (notin this shown). work Severalare expected factors to may be accountweak binders for this. due For to example, their small binding size, does and notare occludeunlikely the to poreshow as inhibitory it may take activi placety when in extramembrane tested in biological regions, assa orys. far Indeed, from the the lumen affinity of of the the pore. top In10 addition, fragment hAQP1 binders in identified red blood using cells SPR is highly was in glycosylated the millimolar compared to high tomicromolar the hAQP1 range. purified from insect cells, and this may affect binding affinity. Another factor is the orientation of the protein in the membrane, as in theory only one side of the protein (extracellular) is exposed to the drug in RBC assays, whereas both extramembrane domains are exposed when the protein is solubilized in micelles. Another factor, also noted by Seeliger et al. [76], is the high density of AQP1 at the membrane of red blood cells. However, there was a large amplitude reduction in the traces for compounds CC38113 and CC46109, both highlighted in Figure6A. It is interesting that we found previously a drug-like compound, NSC658139, that led to a strong reduction in overall amplitude [97]. This compound did Int. J. Mol. Sci. 2016, 17, 449 10 of 20

not reduce the rate constant in the RBC assays but stabilized hAQP1 by as much as 13 ˝C at 30 µM. We are at present unable to explain this behavior or relate it to inhibition of hAQP1.

3. Discussion To our knowledge, the data reported herein is the first fragment-based drug discovery approach to identify binders to an aquaporin using SPR and thermal shift assays. The use of these biophysical techniques that can detect weak intermolecular interactions is justified because the fragments screened in this work are expected to be weak binders due to their small size, and are unlikely to show inhibitory activity when tested in biological assays. Indeed, the affinity of the top 10 fragment binders identified using SPR was in the millimolar to high micromolar range. Int.Int. J. Mol. Sci.Sci. 20162016, ,17 17, ,449 449 10 of 1910 of 19 Int. J. Mol.In Sci. the 2016 SPR, 17, 449 screening, more than 100 fragments (out of 1500) achieved surface10 of occupancy19 above Int. J. Mol. Sci. 2016, 17, 449 10 of 19 theIn baseline the SPR values.screening, This more apparent than 100 fragments hit rate is(out much of 1500) higher achieved than surface that observedoccupancy inabove conventional HTS InIn the SPR screening,screening, more more than than 100 100 fragments fragments (out (out of 1500)of 1500) achieved achieved surface surface occupancy occupancy above above the baseline values. This apparent hit rate is much higher than that observed in conventional HTS thethescreenings baselinebaselineInt.In J.the Mol. values.SPR ofSci. 2016drug-sizedscreening, This,This 17, 449 apparent apparent more compounds than hit hit rate 100 rate isfragments ismuch because much higher (outhigher the ofthan complexity1500)than that achievedthat observed observed of surface thein conventional in drugs occupancy conventional and10 HTSabove of the 19 HTSprobability of screenings of drug-sized compounds because the complexity of the drugs and the probability of screeningsscreeningscomplementaritythe baseline of drug-sizeddrug-sizedvalues. with This compounds compounds theapparent target hit because are becauserate inversely is themuch the complexity highercomplexity correlated than of thattheof [105 the drugsobserved]. drugs The and second in andthe conventional probabilitythe method probability HTS measuredof of protein complementarityIn the SPRwith screening, the target more are than inversely 100 fragments correlated (out of[105]. 1500) Theachieved second surface method occupancy measured above complementaritycomplementaritythermostabilityscreenings of drug-sized withwith under thethe compounds buffertargettarget are conditionsare inversely becauseinversely the identicalcorrelated correlated complexity to[105]. [105]. thoseof Thethe The useddrugssecond second inand method the the method SPRprobability measured assay. measured of Out of the 66 proteinthe thermostability baseline values. under This bufferapparent conditions hit rate is identi muchcal higher to those than used that inobserved the SPR in assay. conventional Out of HTSthe proteinproteincomplementarity thermostability with under under the targetbuffer buffer areconditions conditions inversely identi identi correlatedcalcal to thoseto [105].those used usedThe in the secondin SPRthe SPR assay.method assay. Out measured ofOut the of the 66top top binders screeningsbinders detected detected of drug-sized in inthe the SPRcompounds SPRassay, assay, seven because sevenwere the found complexity were to foundalso of increase the to drugs also the and increasethermal the probability stability the thermal of of stability of 6666 protein toptop binders thermostability detecteddetected in in under the the SPR SPRbuffer assay, assay, conditions seven seven were identiwere found calfound to to those alsoto also increaseused increase in thethe SPRthermalthe assay.thermal stability Out stability of of the of hAQP1,complementarity where most with presented the target a dose-dependentare inversely correlated behavior. [105]. The This second result method suggests measured that good binders to hAQP1,66 top where binders most detected presented in the a dose-dependentSPR assay, seven behavi were or.found This to result also increasesuggests thethat thermal good binders stability to of hAQP1,hAQP1,protein where thermostability mostmost presented presented under a adose-dependent dose-dependent buffer conditions behavi behaviidentior.cal or.This to This those result result used suggests insuggests the thatSPR thatgoodassay. good binders Out bindersof the to to hAQP1hAQP1hAQP1, may maywhere not notnecessarily most necessarily presented stabilize a stabilize dose-dependent the protein, the protein,at least behavi when ator. leastThis tested result when in these suggests tested conditions. inthat these good Conversely, conditions.binders to Conversely, hAQP1hAQP166 may top notbinders necessarilynecessarily detected stabilize stabilizein the SPR the the assay, protein, protein, seven at atleast were least when found when tested to tested also in increase these in these conditions. the conditions. thermal Conversely, stability Conversely, of fragmentsfragmentshAQP1 maywere were not found necessarily found that produced that stabilize produced good the protein, dose-res good atponse dose-responseleast when data intested thermal datain these shift in conditions. thermalbut did not shiftConversely, have but a did not have a fragmentsfragmentshAQP1, were where foundfound most thatthat presented produced produced a dose-dependent good good dose-res dose-res ponsebehaviponse or.data dataThis in resultthermalin thermal suggests shift shift butthat did butgood notdid binders havenot havetoa a goodfragments dose-response were found behavior that producedin SPR, despite good appearingdose-response as binders data in in thermal the primary shift but SPR did screen not have(see a goodgoodgood dose-responsehAQP1 dose-response may not behavior behaviornecessarily behavior in in stabilizeSPR, SPR, in despite SPR,despite the protein, despiteappearing appearing at least appearing as when asbinders binders tested asin in the bindersin these theprimary conditions.primary in SPR the SPRscreen primaryConversely, screen (see SPR (see screen (see summarizedgood dose-response results in Tablebehavior 1). in SPR, despite appearing as binders in the primary SPR screen (see summarizedsummarizedsummarizedfragments results resultswere in in found Table Table inTable that 1). 1). produced1). good dose-response data in thermal shift but did not have a summarizedgood dose-response results in Table behavior 1). in SPR, despite appearing as binders in the primary SPR screen (see Table 1. Summary of the main fragments identified in this work. Binding and stability data Tablesummarized 1. Summary results of in the Table main 1). fragments identified in this work. Binding and stability data correspondingTableTable 1. 1.SummarySummary to 14 compounds of the of main the (Figuresmain fragments 4 fragmentsand 6B). identified Note identified that, in for this AC15153, inwork. this thermalBinding work. shift and Binding data stability could and data stability data correspondingTable 1. Summary to 14 compounds of the main (Figures fragments 4 and 6B).identified Note that, in thisfor AC15153, work. Binding thermal and shift stability data could data notcorrespondingcorresponding be obtained tobecause 14 to compounds 14 compoundsthe protein (Figures did (Figures not 4 aggr and4egate and6B). 6 NotewithinB). Note that, the that, forrange AC15153, for of AC15153,temperatures thermal thermal shifttested. data shiftKD could data could not notcorresponding be Tableobtained 1. Summarybecause to 14 compounds the of proteinthe main (Figures did fragments not 4 aggrand egate6B).identified Note within that,in thethis for range work.AC15153, of Binding temperatures thermal and shift stability tested. data coulddata KD valuesnotbe be obtained forobtained the last because because four compounds thethe proteinprotein could did not not not be aggr aggregateobtainedegate due within within to the the thebad range rangequality of oftemperaturesof temperaturesthe sensorgrams tested. tested. KD K values valuesnot be correspondingfor obtained the last fourbecause to 14compounds compounds the protein could (Figures did not not 4be aggrand ob tained6B).egate Note withindue that, to thefor AC15153,badrange quality of temperaturesthermal of the shift sensorgrams data tested. could K D D performedvalues for theat increasing last four compounds concentrations. could However, not be obtheytained were due clearly to the identified bad quality as bindersof the sensorgrams in the performedforvalues thenot for lastbe at theobtained fourincreasing last compoundsfour because compoundsconcentrations. the couldprotein could notdidHowever, not benot be obtainedaggr ob theytainedegate were within due due clearly toto the thethe range bad badidentified qualityof quality temperatures as of of bindersthe the sensorgrams tested. sensorgrams in the KD performed primary screening performed at 200 µM (Figure 3). Highlighted are the three compounds with good performedprimaryatperformed increasingvalues screening at forat increasing theincreasing concentrations. performed last four concentrations.concentrations. compounds at 200 However, µM could(Figure However, not they 3). be Highlighted wereob theytainedthey clearly were weredue are toclearly identified clearlythe badthr identified ee identifiedquality compounds as binders of as the asbinders sensorgrams withbinders in the good in primarythein the screening dose-responsesprimaryperformed screening in atboth performedincreasing SPR and concentrations. at positive 200 µM (+) (Figure Thermal However, 3). Shift Highlighted they experiments. were clearly are Valuesthe identified thr eeof Tcompoundsagg as correspond binders within tothe good dose-responsesprimary screening in both performedµ SPR and atpositive 200 µM (+) (Figure Thermal 3). ShiftHighlighted experiments. are the Values three ofcompounds Tagg correspond with good to proteinperformed incubated at 200in 1 mMM (Figureof each fragment;3). Highlighted errors are are only the given three for compounds positive thermal with shifts. good n.a., dose-responses in dose-responsesprimary screening in both performedSPR and positive at 200 µM (+) (Figure Thermal 3). Highlighted Shift experiments. are the thr Valuesee compounds ofagg Tagg correspondwith good to proteindose-responses incubated inin both 1 mM SPR of andeach positive fragment; (+) Thermalerrors are Shift only experiments. given for positive ValuesT thermal of T correspond shifts. n.a., to notboth applicable.dose-responses SPR and positive in both (+)SPR Thermal and positive Shift (+) Thermal experiments. Shift experiments. Values of Valuesagg ofcorrespond Tagg correspond to proteinto incubated proteinnotprotein applicable. incubated incubated in in 1 1mM mM of of eacheach fragment;fragment; errors areare only only given given for for positive positive thermal thermal shifts. shifts. n.a., n.a., in 1 mMprotein of incubated each fragment; in 1 mM errorsof each arefragment; only errors given are for only positive given for thermal positive shifts. thermal n.a., shifts. not n.a., applicable. notMaybridgenot applicable. applicable. Code Structure SPR (KD) Tagg (°C) ΔTagg (+/−) Maybridgenot applicable. Code Structure SPR (KD) Tagg (°C) ΔTagg (+/−) MaybridgeMaybridge Code Code Structure Structure SPR SPR ( K(KD)D) TaggTagg(°C)(°C) Δ˝TΔaggT agg(+/ −(+/) −) MaybridgeMaybridge Code Code Structure Structure SPR SPR (K (KDD)) Tagg (°C)T aggΔ(TaggC) (+/−) ∆Tagg (+/´) AC42399 546 µM 48.0 − AC42399 546 µM 48.0 − AC42399AC42399AC42399AC42399 546546546µ µM µM MµM 48.0 48.0 48.0 −− − ´

MO00656 865 µM 51.8 − MO00656 865 µM 51.8 − MO00656MO00656MO00656 865865865µ µM µMM 51.8 51.8 −− ´ MO00656 865 µM 51.8 −

CC53013 1.31 mM 48.6 − CC53013 1.31 mM 48.6 − CC53013CC53013CC53013 1.311.311.31 mMmM mM 48.6 48.6 −− ´ CC53013 1.31 mM 48.6 −

CC38113 62 µM 63.3 ± 0.1 + CC38113 62 µMµ 63.3 ± 0.1 + CC38113CC38113 6262 µMM 63.3 ± 0.1 63.3 ˘ 0.1 + + CC38113 62 µM 63.3 ± 0.1 + CC38113 62 µM 63.3 ± 0.1 +

AC15153AC15153 337337 µMµM n.a. n.a. n.a. n.a. AC15153AC15153 337337 µM µM n.a. n.a. n.a. n.a. AC15153 337 µM n.a. n.a. AC15153 337 µM n.a. n.a. SB00353 902 µM 51.1 − SB00353SB00353 902902 µM µM 51.1 51.1 −− SB00353 902 µM 51.1 − SB00353 902 µM 51.1 − MO08607 962 µM 59.0 ± 0.3 + MO08607MO08607 962962 µM µM 59.0 59.0 ± 0.3± 0.3 + + MO08607 962 µM 59.0 ± 0.3 +

MO08607 962 µM 59.0 ± 0.3 +

Int. J. Mol. Sci. 2016, 17, 449 10 of 19 Int. J. Mol. Sci. 2016, 17, 449 10 of 19 In the SPR screening, more than 100 fragments (out of 1500) achieved surface occupancy above In the SPR screening, more than 100 fragments (out of 1500) achieved surface occupancy above the baseline values. This apparent hit rate is much higher than that observed in conventional HTS the baseline values. This apparent hit rate is much higher than that observed in conventional HTS screenings of drug-sized compounds because the complexity of the drugs and the probability of screenings of drug-sized compounds because the complexity of the drugs and the probability of complementarity with the target are inversely correlated [105]. The second method measured complementarity with the target are inversely correlated [105]. The second method measured protein thermostability under buffer conditions identical to those used in the SPR assay. Out of the protein thermostability under buffer conditions identical to those used in the SPR assay. Out of the 66 top binders detected in the SPR assay, seven were found to also increase the thermal stability of 66 top binders detected in the SPR assay, seven were found to also increase the thermal stability of hAQP1, where most presented a dose-dependent behavior. This result suggests that good binders to hAQP1, where most presented a dose-dependent behavior. This result suggests that good binders to hAQP1 may not necessarily stabilize the protein, at least when tested in these conditions. Conversely, hAQP1 may not necessarily stabilize the protein, at least when tested in these conditions. Conversely, fragments were found that produced good dose-response data in thermal shift but did not have a fragments were found that produced good dose-response data in thermal shift but did not have a good dose-response behavior in SPR, despite appearing as binders in the primary SPR screen (see good dose-response behavior in SPR, despite appearing as binders in the primary SPR screen (see summarized results in Table 1). summarized results in Table 1). Table 1. Summary of the main fragments identified in this work. Binding and stability data Table 1. Summary of the main fragments identified in this work. Binding and stability data corresponding to 14 compounds (Figures 4 and 6B). Note that, for AC15153, thermal shift data could corresponding to 14 compounds (Figures 4 and 6B). Note that, for AC15153, thermal shift data could not be obtained because the protein did not aggregate within the range of temperatures tested. KD not be obtained because the protein did not aggregate within the range of temperatures tested. KD values for the last four compounds could not be obtained due to the bad quality of the sensorgrams values for the last four compounds could not be obtained due to the bad quality of the sensorgrams performed at increasing concentrations. However, they were clearly identified as binders in the performed at increasing concentrations. However, they were clearly identified as binders in the primary screening performed at 200 µM (Figure 3). Highlighted are the three compounds with good primary screening performed at 200 µM (Figure 3). Highlighted are the three compounds with good dose-responses in both SPR and positive (+) Thermal Shift experiments. Values of Tagg correspond to dose-responses in both SPR and positive (+) Thermal Shift experiments. Values of Tagg correspond to protein incubated in 1 mM of each fragment; errors are only given for positive thermal shifts. n.a., protein incubated in 1 mM of each fragment; errors are only given for positive thermal shifts. n.a., not applicable. not applicable. Maybridge Code Structure SPR (KD) Tagg (°C) ΔTagg (+/−) Maybridge Code Structure SPR (KD) Tagg (°C) ΔTagg (+/−)

AC42399 546 µM 48.0 − AC42399 546 µM 48.0 −

MO00656 865 µM 51.8 − MO00656 865 µM 51.8 −

CC53013 1.31 mM 48.6 − CC53013 1.31 mM 48.6 −

CC38113 62 µM 63.3 ± 0.1 + CC38113 62 µM 63.3 ± 0.1 + Int. J. Mol. Sci. 2016, 17, 449 11 of 20

AC15153 Table 1. 337Cont µM. n.a. n.a. AC15153 337 µM n.a. n.a. ˝ Maybridge Code Structure SPR (KD) Tagg ( C) ∆Tagg (+/´)

SB00353 902 µM 51.1 − SB00353SB00353 902902 µMµM 51.1 51.1− ´

Int. J. Mol. Sci. 2016, 17, 449 11 of 19 Int. J. Mol. Sci. 2016, 17, 449 11 of 19 Int. J. Mol. Sci. 2016, 17, 449 11 of 19 Int. J. Mol. Sci. 2016, 17, 449 11 of 19 Int. J. Mol. Sci.MO08607 2016MO08607, 17, 449 962962 µMµM 59.0 ± 0.3 59.0 ˘ 0.3 + 11 of 19 + MO08607 962 µM 59.0 ± 0.3 + CC46109 982 µM 103.4 ± 4.6 + CC46109 982 µM 103.4 ± 4.6 + CC46109 982 µM 103.4 ± 4.6 + CC46109CC46109CC46109 982982982 µM µMµM 103.4 103.4 ± ± 4.6 4.6 103.4 ˘ +4.6 + +

CC40813 133 µM 50.3 − CC40813 133 µM 50.3 − CC40813CC40813 133133 µMµM 50.3 50.3 − ´ CC40813CC40813 133133 µM µM 50.3 50.3 −−

AC34875 225 µM 50.9 − AC34875 225 µM 50.9 − AC34875AC34875 225225 µMµM 50.9 50.9 − ´ AC34875AC34875 225225 µM µM 50.9 50.9 −−

CC47846 n.a. 70.0 ± 1.2 + CC47846CC47846 n.a.n.a. 70.0 ± 1.2 70.0 ˘ 1.2 + + CC47846 n.a. 70.0 ± 1.2 + CC47846CC47846 n.a.n.a. 70.0 70.0 ± ± 1.2 1.2 + +

CC55714 n.a. 63.6 ± 0.4 + CC55714CC55714 n.a.n.a. 63.6 ± 0.4 63.6 ˘ 0.4 + + CC55714 n.a. 63.6 ± 0.4 + CC55714CC55714 n.a.n.a. 63.6 63.6 ± ± 0.4 0.4 + +

AC36708AC36708 n.a.n.a. 62.9 ± 0.4 62.9 ˘ 0.4 + + AC36708 n.a. 62.9 ± 0.4 + AC36708 n.a. 62.9 ± 0.4 + AC36708AC36708 n.a.n.a. 62.9 62.9 ± ± 0.4 0.4 + +

SB01744SB01744 n.a.n.a. 58.7 ± 0.2 58.7 ˘ 0.2 + + SB01744 n.a. 58.7 ± 0.2 + SB01744 n.a. 58.7 ± 0.2 + SB01744SB01744 n.a.n.a. 58.7 58.7 ± ± 0.2 0.2 + + Overall, the fragments selected from the two independent studies are anticipated to serve as Overall,Overall, the the fragments fragments selected selected from fromthe two the in twodependent independent studies studiesare anticipated are anticipated to serve as to serve as good good Overall,starting thematerials fragments for the selected future from design the of two more independent complex molecules studies are that anticipated can become to serveclinical as good Overall,starting thematerials fragments for the selected future from design the of two more independent complex molecules studies are that anticipated can become to serveclinical as candidatesstartinggoodOverall, starting materials for the materialshAQP1 fragments for inhibition. for the selectedthe future future The from design next design the step of twoof morewill more in dependentbe complex complexto map the studiesmolecules molecules binding are thatsitesanticipated that canof can these become becometo fragments serve clinical clinicalas candidates candidatesgood starting for materialshAQP1 inhibition. for the future The next design step of will more be complexto map the molecules binding thatsites can of these become fragments clinical goodoncandidates the starting hAQP1 for materials hAQP1protein inhibition.forsurface the futureusin Theg designnextbiophysical step of willmore or be structuralcomplex to map themolecules techniques. binding that sites High-resolution can of becomethese fragments clinical X-ray onforcandidates the hAQP1 hAQP1 for inhibition. hAQP1protein inhibition.surface The nextusin Theg step biophysicalnext will step be will toor be mapstructural to map the bindingthetechniques. binding sites sitesHigh-resolution of of these these fragments fragments X-ray on the hAQP1 candidatesoncandidates the hAQP1 for for hAQP1 hAQP1protein inhibition. inhibition.surface usin The Theg next biophysicalnext step step will will or be be structural to to map map the thetechniques. binding binding sites sitesHigh-resolution of of these these fragments fragments X-ray crystallographyproteinon the hAQP1 surface mayprotein using be difficultsurface biophysical usinfor theseg orbiophysical structurallow affinity or techniques. structuralcompounds techniques. and High-resolution may leadHigh-resolution to low X-ray occupancy crystallographyX-ray may crystallographyon the hAQP1 proteinmay be difficultsurface usinfor theseg biophysical low affinity or structuralcompounds techniques. and may leadHigh-resolution to low occupancy X-ray onandcrystallography the low hAQP1 electron protein maydensity. be surface difficult The binding usin forg these biophysicalepitope low of affinity the or drug structural compounds can also techniques. be and studied may High-resolution bylead solution to low NMR,occupancy X-ray e.g., andbecrystallography difficultlow electron for maydensity. these below difficult The affinity binding for these compoundsepitope low of affinity the anddrug compounds maycan also lead be and to studied low may occupancy bylead solution to low NMR, andoccupancy low e.g., electron density. crystallographySTDand low(Saturation electron may density.Transfer be difficult The Difference) binding for these epitopeexperiments, low affinityof the drugwhich compounds can rely also on beand the studied maytransfer lead by solution ofto saturationlow occupancy NMR, from e.g., STDand low(Saturation electron Transferdensity. TheDifference) binding epitopeexperiments, of the whichdrug can rely also on be the studied transfer by solutionof saturation NMR, from e.g., andSTDTheand low low(Saturation binding electron electron epitope density. Transferdensity. ofThe TheDifference) the binding binding drug epitope canepitopeexperiments, also of of thebe the drug studied whichdrug can can rely also by also on solutionbe be the studied studied transfer NMR, by by solution ofsolution e.g.,saturation NMR, STD NMR, (Saturationfrome.g., e.g., Transfer proteinSTD (Saturation to ligand Transfer[106,107]. Difference) Direct protein experiments, resonance which perturbation rely on theusing transfer labeled of proteins saturation is morefrom proteinSTD (Saturation to ligand Transfer[106,107]. Difference) Direct protein experiments, resonance which perturbation rely on theusing transfer labeled of proteins saturation is morefrom STDdifficultDifference)protein (Saturation todue ligand to experiments, theTransfer [106,107]. large Difference)complex Direct which formed protein experiments, rely by onresona tetr theameric ncewhich transfer perturbation AQP rely ofandon saturationthe theusing transfer detergent labeled fromof micelles.saturationproteins protein Finally,is frommore to ligand [106,107]. difficultprotein todue ligand to the [106,107]. large complex Direct formedprotein by resona tetramericnce perturbation AQP and theusing detergent labeled micelles.proteins Finally,is more proteintargetdifficultDirect perturbation to proteindue ligand to the resonance[106,107]. canlarge be complextested Direct perturbation using proteinformed solid resonaby usingstat tetre NMRnceameric labeled perturbation in AQPlipid proteins andbilayers, usingthe is detergentbut more labeled resonance difficult proteinsmicelles. assignments due isFinally, more to the large complex targetdifficult perturbation due to the can large be complextested using formed solid by stat tetre NMRameric in AQPlipid bilayers,and the detergentbut resonance micelles. assignments Finally, difficulttargetdifficult perturbation due due to to the the largecan large be complex complextested using formed formed solid by by stat tetr tetre americNMRameric in AQP AQPlipid and bilayers,and the the detergent detergentbut resonance micelles. micelles. assignments Finally, Finally, offormedtarget hAQP1 perturbation by are tetrameric not yet can available. be AQP tested and using the solid detergent state NMR micelles. in lipid Finally,bilayers, targetbut resonance perturbation assignments can be tested using oftarget hAQP1 perturbation are not yet can available. be tested using solid state NMR in lipid bilayers, but resonance assignments targetof hAQP1In perturbation principle, are not a yetcan fragment available. be tested binding using solidclose stat to eits NMR pore in could lipid bilayers,be developed but resonance into a more assignments complex solidof hAQP1Instate principle, are NMR not ayet infragment available. lipid bilayers, binding close but resonanceto its pore assignmentscould be developed of hAQP1 into a are more not complex yet available. ofmolecule hAQP1In principle, arethat not could yet a fragmentavailable.physically binding occlude close the tochannel. its pore Possible could bechemical developed modifications into a more include complex the moleculeIn principle, that could a fragmentphysically binding occlude close the channel.to its pore Possible could chemicalbe developed modifications into a more include complex the moleculeInInIn principle, principle, principle,that could a a fragment afragmentphysically fragment binding binding occlude binding close close the close to channel.to its its to pore pore its Possiblepore could could could be chemicalbe developed developed be developed modifications into into a a intomore more include a morecomplex complex the complex molecule linkagemolecule of thattwo couldor more physically fragment occludes or the the customization channel. Possible of functional chemical groupsmodifications to build include lead-like the linkagemolecule of thattwo couldor more physically fragment occludes or the the customization channel. Possible of functional chemical groupsmodifications to build include lead-like the moleculecompounds.thatlinkage could ofthat two physicallyIt could is or these morephysically lead-like occludefragment occlude compounds thes or channel. the customizationchannel.that Possibleare expectedPossible of chemical functionalchemicalto be active modifications modifications groups in biological to build include assays lead-like andthe the linkage of two compounds.linkage of two It is or these more lead-like fragment compoundss or the customization that are expected of functionalto be active groups in biological to build assays lead-like and linkagecancompounds.or morethus of be fragmentstwo tested It isor these usingmore or lead-like aquapofragment the customizationrin compoundss functionalor the customization that assays of functionalare to expected asse ofss functionaltheir groups to be inhibitory active to groups build in effects.biological lead-liketo build assays lead-like compounds. and It is these cancompounds. thus be tested It is theseusing lead-likeaquaporin compounds functional thatassays are to expected assess their to be inhibitory active in effects. biological assays and compounds.cancompounds. thus be testedIt It is is these theseusing lead-like lead-likeaquapo rincompounds compounds functional that thatassays are are toexpected expected assess theirto to be be inhibitory active active in in biologicaleffects. biological assays assays and and lead-likecan thusWe proposebe compounds tested the using use aquapo of that fragments arerin expectedfunctional as building assays to be bl activetoocks asse ofss inaquaporin their biological inhibitory modu assays effects.lators. and While can there thus is be tested using can thusWe proposebe tested the using use aquapo of fragmentsrin functional as building assays bl toocks asse ofss aquaporin their inhibitory modu effects.lators. While there is canstill thus aWe long be propose waytested to using thego beforeuse aquapo of thefragmentsrin hits functional from as fragment building assays screenings bltoocks asse ssof their aquaporinare ableinhibitory to modudeliver effects.lators. potent While modulators there is stillaquaporin aWe long propose way functional to thego beforeuse assaysof thefragments hits to from assess as fragment building their inhibitory screeningsblocks of aquaporin are effects. able to modudeliverlators. potent While modulators there is ofstill aquaporins, Wea long propose way thisto the go strategy usebefore of fragmentstheshould hits befrom asfurther buildingfragment dev eloped blscreeningsocks ofas aquaporinan are alternative able to modu deliver approachlators. potent While to modulatorsthe there current is ofstill aquaporins, a Welong propose way this to go strategy the before use shouldthe of hits fragments befrom further fragment asdev buildingeloped screenings as blocksan are alternative able of to aquaporin deliver approach potent modulators.to modulatorsthe current While there is stillofstill aquaporins, a a long long way way tothis to go go strategy before before the shouldthe hits hits frombe from further fragment fragment dev elopedscreenings screenings as an are are alternative able able to to deliver deliver approach potent potent to modulators themodulators current saturationof aquaporins, in the this screening strategy shouldof drug-sized be further mole devculeseloped for as goodan alternative therapeutic approach candidates to the againstcurrent saturationstillof aquaporins, a long inway the this toscreening strategy go before shouldof drug-sized the be hits further from mole dev fragmentculeseloped for as screeningsgoodan alternative therapeutic are approach able candidates to to deliver the againstcurrent potent modulators ofaquaporins.saturation aquaporins, inSpecifically, thisthe strategyscreening AQP1 should of deletiondrug-sized be further directly moledev elopedcausescules forasreduced angood alternative vascularitytherapeutic approach and candidates hypoxia, to the current againstwhich aquaporins.saturation inSpecifically, the screening AQP1 of deletiondrug-sized directly mole causescules forreduced good vascularitytherapeutic and candidates hypoxia, againstwhich saturationincreasesaquaporins.of aquaporins, VEGFin Specifically,the expression thisscreening strategy AQP1 [55],of deletiondrug-sized shouldsuggesting directly be mole furtherthat culescauses simultaneous developedfor reduced good inhibitiontherapeuticvascularity as an alternative of andcandidates AQP1 hypoxia, and approach against whichVEGF to the current increasesaquaporins. VEGF Specifically, expression AQP1 [55], deletion suggesting directly that causes simultaneous reduced inhibitionvascularity of and AQP1 hypoxia, and VEGFwhich aquaporins.increasesaquaporins. VEGF Specifically, Specifically, expression AQP1 AQP1 [55], deletion deletion suggesting directly directly that causes causes simultaneous reduced reduced vascularity inhibitionvascularity ofand and AQP1 hypoxia, hypoxia, and which VEGFwhich pathwaysincreases couldVEGF have expression an improved [55], effect.suggesting In additi thaton, simultaneous anti-AQP1 therapies inhibition should of AQP1not show and adverse VEGF pathwaysincreases couldVEGF have expression an improved [55], effect.suggesting In additi thaton, simultaneous anti-AQP1 therapies inhibition should of AQP1not show and adverse VEGF increasespathways VEGF could haveexpression an improved [55], suggesting effect. In additi that on,simultaneous anti-AQP1 therapiesinhibition should of AQP1 not show and adverseVEGF pathways pathways could could have have an an improved improved effect. effect. In In additi addition,on, anti-AQP1 anti-AQP1 therapies therapies should should not not show show adverse adverse

Int. J. Mol. Sci. 2016, 17, 449 12 of 20 saturation in the screening of drug-sized molecules for good therapeutic candidates against aquaporins. Speciﬁcally, AQP1 deletion directly causes reduced vascularity and hypoxia, which increases VEGF expression [55], suggesting that simultaneous inhibition of AQP1 and VEGF pathways could have an improved effect. In addition, anti-AQP1 therapies should not show adverse effects on humans because “Colton-null” individuals (AQP1 null) do not display signiﬁcant clinical symptoms [108,109]. Therefore, AQP1 water channel inhibition may be a novel approach to reduce tumor vessels.

4. Materials and Methods

4.1. Purification of hAQP1 from Insect Cells The full-length hAQP1 cDNA sequence was cloned into the vector pFB-LIC-Bse downstream of an N-terminal 6His and TEV protease cleavage site coding sequences (6His-TEV-hAQP1). Insect Spodoptera frugiperda cells overexpressing the recombinant hAQP1 protein was obtained from the Protein Production Platform (NTU, Singapore). The cell pellet was resuspended in 5 mL/g lysis buffer (20 mM Tris-HCl pH 8, 300 mM NaCl, 2 mM β-mercaptoethanol (β-ME), 10% glycerol (v/v), benzonase (25 U/mL) (Novagen, Madison, WI, USA), complete protease inhibitor cocktail and 2% n-Octyl-β-D-Glucopyranoside (OG, Anagrade, Affymetrix, Singapore)). Cell lysis was achieved with 10 min of sonication followed by passing through a microfluidizer (Microfluidics, Westwood, MA, USA) 5 times. The mixture was allowed to solubilize at 4 ˝C for 2 h, followed by centrifugation at 40,000ˆ g for 30 min at 4 ˝C. The supernatant was loaded onto Ni-NTA column (packed with Bio-Rad Profinity™ IMAC resin, (Bio-Rad, Hercules, CA, USA) pre-equilibrated with lysis buffer. After overnight binding at 4 ˝C, the resin was washed with 30 resin volume of washing buffer (20 mM Tris-HCl pH 8, 300 mM NaCl, 25 mM imidazole, 2 mM β-ME and 10% glycerol). The final two resin volume wash was performed using washing buffer containing 50 mM imidazole and 1% OG (w/v) to remove additional contaminants. Recombinant hAQP1 protein was eluted with at least 5 resin volumes of elution buffer (20 mM Tris-HCl pH 8, 300 mM NaCl, 300 mM imidazole, 2 mM β-ME, 10% glycerol and 1% OG). Elution fractions were pooled and concentrated before gel filtration chromatography using HiLoad 16/600 Superdex 200 prep grade column (GE Healthcare, Uppsala, Sweden) at 4 ˝C. The protein was eluted with 20 mM Tris-HCl pH 8, 300 mM NaCl and 1% OG. The fraction size per elution is 2 mL, collected at a flow rate of 1 mL/min. E. coli AQPZ was prepared as indicated previously [101].

4.2. Gel Electrophoresis The protein concentration of samples was measured using the NanoDrop™ 1000 spectrophotometer (Thermo Scientific, Waltham, MA, USA). SDS-PAGE gels were run at 200 V for 50 min, using TGS (25 mM Tris, 192 mM glycine, 0.1% SDS, pH 8.3) running buffer. The gels were stained with Coomassie blue (Bio-Rad, Hercules, CA, USA) and destained using 30% methanol and 10% acetic acid for visualization of protein bands. Protein standards were purchased from GE Healthcare (Vendevagen, Sweden). Blue-native polyacrylamide gel electrophoresis, BN-PAGE, was performed as previously described [101]. Briefly, purified protein was incubated in sample buffer containing 750 mM aminocaproic acid, 50 mM Bis-Tris HCl pH 7.0, 0.5 mM EDTA and 100 mM OG (4 times its critical micellar concentration, CMC, of 25 mM). Coomassie brilliant blue was added to the sample to a concentration of 0.35% (w/v) immediately before gel loading. Samples were loaded into a pre-cast NativePAGE™ Novex™ 4%–16% Bis-Tris protein gel (Invitrogen, Carlsbad, CA, USA), with an inner blue cathode buffer (15 mM Bis-Tris HCl, 50 mM Tricine, and 0.02% Coomassie blue, pH 7.0) and an outer anode buffer (50 mM Bis-Tris HCl pH 7.0), and separated at a constant 150 V for approximately 70 min at 4 ˝C before replacing the blue cathode buffer with colourless cathode buffer (15 mM Bis-Tris HCl and 50 mM Tricine, pH 7.0), and allowed to run at 250 V till the dye front reached the edge of the gel. The NativeMark™ (Invitrogen) acted as the protein standard. The gels were stained with Coomassie blue. Int. J. Mol. Sci. 2016, 17, 449 13 of 20

4.3. Reconstitution of Purified Aquaporin into Liposomes. Liposomes used in this work were prepared using film rehydration [110]. In this procedure, a thin lipid film of Escherichia coli lipids is formed by drying under a nitrogen stream lipid extract powder (Avanti Polar Lipids Inc., Alabaster, AL, USA) in chloroform. After at least 2 h in a vacuum desiccator, purified aquaporin protein in detergent was added and supplemented with reconstitution buffer (20 mM Tris–HCl pH 8.0) to obtain a protein-to-lipid molar ratio of 1:400. A volume of 1 mL of this reconstitution mixture was incubated for 1 h at room temperature. Uniform sized liposomes were obtained by extrusion through a 400 nm pore-size polycarbonate membrane using an Avestin extruder. Dynamic light scattering (90 Plus Particle Size Analyzer, Brookhaven Instruments, Holtsville, NY, USA) was used to measure the diameter of the AQP proteoliposomes. A liposome control was prepared in the same way without addition of protein.

4.4. Stopped-Flow Water Permeability Assay The water permeability of proteoliposomes reconstituted with purified aquaporin was determined in a SX20 stopped-flow spectrometer (Applied Photophysics, Leatherhead, UK) with a dead time of 1.1 ms. Water permeability was assayed by rapidly mixing of a suspension of proteoliposomes with PBS containing 500 mM sucrose to establish a 250 mM inwardly directed osmotic gradient. The time course of 90˝ scattered light intensity at 500 nm was used to measure the kinetics of liposome shrinkage. Experiments were carried out at 23 ˝C. The solution osmolarities were measured using a vapor pressure osmometer (Wescor, Logan, UT, USA). Each experimental trace represents the average of five to ten replicates. The measured average diameter of control liposomes and hAQP1-proteoliposomes were 368 and 262 nm, respectively.

4.5. Light Scattering Data Analysis Data was acquired by the Pro-Data SX software and analyzed by Pro-Data Viewer software (Applied Photophysics), where the measured light scattering intensity signals were fitted to an exponential function to obtain the rate constants. Dose-response curves were fitted using OriginPro 8.5 software (OriginLab, Northampton, MA, USA) to a sigmoidal function. The osmotic water permeability coefficients (P f ) were calculated from the light scattering time course [102].

4.6. Surface Plasmon Resonance (SPR)-Protein Immobilization Surface Plasmon Resonance measurements were performed using Biacore T-200 (GE Healthcare, Uppsala, Sweden) using the running buffer PBS pH 7.4 supplemented with 1% OG and 5% dimethylsulfoxide (DMSO) at 25 ˝C. Puriﬁed recombinant hAQP1 was directly immobilized onto Biacore Sensor Chip CM5 Research Grade (GE Healthcare, Little Chalfont, UK) using standard amine-coupling chemistry at a ﬂow rate of 5 µL/min. Carboxyl derivatives on CM5 dextran surface were activated with a 10-min 1:1 mixture of 0.2 M N-ethyl-N1-(3-(diethylamino)propyl)carbodiimide (EDC) and 50 mM N-hydroxysuccinimide (NHS). The aquaporins were diluted in 10 mM sodium acetate, pH 5.0 and injected across the surface separately. Addition of 0.5 M ethanolamine–HCl at pH 8.5 for 10 min was used to block unreacted activated carboxyl derivates.

4.7. Fragment Screening Using SPR The Maybridge Ro3 Diversity Fragment Library (Thermo Fisher Scientiﬁc, Waltham, MA, USA) consisting of 1500 fragments was used for screening. These fragments have molecular weight ranging from 82.1 to 295.3 Da, with an experimental guarantee of solubility of ě1 mM in aqueous phosphate buffer and ě200 mM in DMSO. They satisfy the “rule of three” criteria, i.e., a molecular weight of ď300 Da, a cLogP of ď3.0, ď3 H-bond acceptors or H-bond donors, ď3 rotatable bonds, and a polar surface area of ď60 Å2. Lyophilized powder of each fragments were pre-dissolved in 100% DMSO prior to the screening campaign. A volume of 10 µL of a 4 mM stock solution of each fragment in Int. J. Mol. Sci. 2016, 17, 449 14 of 20

100% DMSO was added to 190 µL of a buffer composed of 1.06 ˆ PBS + 1.06% OG so as to obtain a final fragment concentration of 200 µM in PBS + 1% OG + 5% DMSO in a volume of 200 µL. We note that although a higher screening concentration in the mM range would in principle result in higher protein surface occupancy, allowing the detection of weaker interactions, this may also increase the probability of non-specific binding and/or promiscuously binding colloidal aggregates. In addition, the presence of salt and detergent in the buffer helps to stabilize the hAQP1 and also reduces the chances of non-specific binding. The DMSO solvent has a high refractive index and can be a major cause of false positives and plate-to-plate baseline variations. Therefore, it was important to accurately match its concentration in the sample buffer and the running buffer throughout the screening to minimize data inconsistencies during analysis. The fragments were injected across a reference channel and hAQP1 channel for 2 min at flow rate of 30 µL/min, and then allowed to dissociate in running buffer for another min. Each fragment was injected in duplicates. DMSO calibration was performed for DMSO correction. Reproducibility was monitored with duplicate injections of each sample. No regeneration step was necessary as all fragments were dissociated. Previously validated positive controls, NSC226080 and NSC624169 (not shown), were included over the course of screening to monitor the surface protein activity. Fragments that showed good binding response and reproducibility were further tested in a dose-dependent assay. These fragments, together with the positive controls, were diluted into 25, 50, 100 and 200 µM and tested over the same surfaces under the same conditions. DMSO calibration was performed for DMSO correction. Non-stoichiometric binders (potential fragments that aggregate, precipitate, binding non-specifically, etc.) were excluded. Good binders were assayed in a broader concentration range from 6.25 to 800 µM in 2-fold dilutions under the same conditions.

4.8. Data Processing and Curve Fitting Raw sensorgrams were corrected for DMSO and double referenced (subtraction of the reference flow cell followed by subtraction of the closest buffer blank), which responses were corrected with a reference flow cell and a blank buffer injections. Corrected sensorgrams were y-axis and x-axis corrected at the point of fragment injection. Responses for the varied concentrations for a fragment were overlaid. Resultant equilibrium responses during injection phase were plotted against the log of their respective concentrations and fit to a steady-state model to obtain their affinities.

4.9. Thermal Aggregation Shift Assay Temperature-dependent protein aggregation was measured with static light scattering (Stargazer-384, Harbinger Biotech, Markham, ON, Canada) as previously described [98] with modifications. Briefly, purified hAQP1 in PBS pH 7.4 and 1% OG at a concentration of 0.2 mg/mL, incubated with test compound or fragment at the appropriate concentration at room temperature for 15 min, and aliquoted in a clear-bottom 384-well plate (Nunc, Roskilde, Denmark) to a volume of 45 µL in triplicate. The final concentration of DMSO was kept at 5%. Mineral oil (45 µL) covered the sample to prevent evaporation. The sample was heated from 25 to 80 ˝C at a rate of 1 ˝C per minute. Incident light was shone on the protein drop from beneath at a 30˝ angle, and a CCD camera was used to measure the scattered light intensity every 30 s to monitor protein aggregation. Data collated was analyzed using the Bioactive software (Harbinger Biotech). Intensities were plotted as a function of temperature and fitted to the Boltzmann equation by non-linear regression. The point of inflection of the fitted curve defined the temperature of aggregation (Tagg).

Acknowledgments: We thank the Nanyang Technological University Protein Production Platform for the expression of hAQP1 in insect cells, Julien Lescar for sharing the small fragment library and Yin Hoe Yau, York Wieo Chuah, and Susana Geifman for help in SPR experiments. This work has been funded by Singapore Ministry of Education(MOE) Tier 1 grant RG150/14. Int. J. Mol. Sci. 2016, 17, 449 15 of 20

Author Contributions: Janet To and Jaume Torres conceived and designed the experiments; Janet To performed the experiments and analyzed the data; Jaume Torres contributed reagents/materials/analysis tools; Janet To and Jaume Torres wrote the paper. Conﬂicts of Interest: The authors declare no conﬂict of interest.

Abbreviations AQP Aquaporin DSLS Differential Static Light Scattering FBDD Fragment-based Drug Discovery HTS High-throughput Screening SPR Surface Plasmon Resonance TM Transmembrane

References

1. Benga, G.; Popescu, O.; Borza, V.; Pop, V.I.; Muresan, A.; Mocsy, I.; Brain, A.; Wrigglesworth, J.M. Water permeability in human erythrocytes: Identification of membrane proteins involved in water transport. Eur. J. Cell Biol. 1986, 41, 252–262. [PubMed] 2. Denker, B.M.; Smith, B.L.; Kuhajda, F.P.; Agre, P. Identification, purification, and partial characterization of a

novel MR 28,000 integral membrane protein from erythrocytes and renal tubules. J. Biol. Chem. 1988, 263, 15634–15642. [PubMed] 3. Preston, G.M.; Carroll, T.P.; Guggino, W.B.; Agre, P. Appearance of water channels in Xenopus oocytes expressing red cell CHIP28 protein. Science 1992, 256, 385–387. [CrossRef][PubMed] 4. Ishibashi, K.; Hara, S.; Kondo, S. Aquaporin water channels in mammals. Clin. Exp. Nephrol. 2009, 13, 107–117. [CrossRef][PubMed] 5. Yasui, M.; Hazama, A.; Kwon, T.H.; Nielsen, S.; Guggino, W.B.; Agre, P. Rapid gating and anion permeability of an intracellular aquaporin. Nature 1999, 402, 184–187. [PubMed] 6. Yool, A.J. Functional domains of aquaporin-1: Keys to physiology, and targets for drug discovery. Curr. Pharm. Des. 2007, 13, 3212–3221. [CrossRef][PubMed] 7. Saparov, S.M.; Liu, K.; Agre, P.; Pohl, P. Fast and selective ammonia transport by aquaporin-8. J. Biol. Chem. 2007, 282, 5296–5301. [CrossRef][PubMed] 8. Holm, L.M.; Jahn, T.P.; Moller, A.L.; Schjoerring, J.K.; Ferri, D.; Klaerke, D.A.; Zeuthen, T. NH3 and NH4+ permeability in aquaporin-expressing Xenopus oocytes. Pﬂugers Arch. 2005, 450, 415–428. [CrossRef] [PubMed] 9. King, L.S.; Kozono, D.; Agre, P. From structure to disease: The evolving tale of aquaporin biology. Nat. Rev. Mol. Cell Biol. 2004, 5, 687–698. [CrossRef][PubMed] 10. Wysocki, R.; Chery, C.C.; Wawrzycka, D.; Van Hulle, M.; Cornelis, R.; Thevelein, J.M.; Tamas, M.J. The glycerol channel Fps1p mediates the uptake of arsenite and antimonite in Saccharomyces cerevisiae. Mol. Microbiol. 2001, 40, 1391–1401. [CrossRef][PubMed] 11. Sanders, O.I.; Rensing, C.; Kuroda, M.; Mitra, B.; Rosen, B.P. Antimonite is accumulated by the glycerol facilitator GlpF in Escherichia coli. J. Bacteriol. 1997, 179, 3365–3367. [PubMed]

12. Cooper, G.J.; Boron, W.F. Effect of PCMBS on CO2 permeability of Xenopus oocytes expressing aquaporin 1 or its C189S mutant. Am. J. Physiol. Cell Physiol. 1998, 275, C1481–1486. 13. Herrera, M.; Hong, N.J.; Garvin, J.L. Aquaporin-1 transports NO across cell membranes. Hypertension 2006, 48, 157–164. [CrossRef][PubMed] 14. Ivanov, I.I.; Loktyushkin, A.V.; Gus’kova, R.A.; Vasil’ev, N.S.; Fedorov, G.E.; Rubin, A.B. Oxygen channels of erythrocyte membrane. Doklady 2007, 414, 137–140. [CrossRef][PubMed] 15. Gonen, T.; Sliz, P.; Kistler, J.; Cheng, Y.; Walz, T. Aquaporin-0 membrane junctions reveal the structure of a closed water pore. Nature 2004, 429, 193–197. [CrossRef][PubMed] 16. Sui, H.; Han, B.-G.; Lee, J.K.; Walian, P.; Jap, B.K. Structural basis of water-speciﬁc transport through the AQP1 water channel. Nature 2001, 414, 872–878. [CrossRef][PubMed] Int. J. Mol. Sci. 2016, 17, 449 16 of 20

17. Hiroaki, Y.; Tani, K.; Kamegawa, A.; Gyobu, N.; Nishikawa, K.; Suzuki, H.; Walz, T.; Sasaki, S.; Mitsuoka, K.; Kimura, K.; et al. Implications of the aquaporin-4 structure on array formation and cell adhesion. J. Mol. Biol. 2006, 355, 628–639. [CrossRef][PubMed] 18. Horsefield, R.; Norden, K.; Fellert, M.; Backmark, A.; Tornroth-Horsefield, S.; Terwisscha Van Scheltinga, A.C.; Kvassman, J.; Kjellbom, P.; Johanson, U.; Neutze, R. High-resolution X-ray structure of human aquaporin 5. Proc. Natl. Acad. Sci. USA 2008, 105, 13327–13332. [CrossRef][PubMed] 19. Tajkhorshid, E.; Nollert, P.; Jensen, M.Ã.; Miercke, L.J.W.; O’Connell, J.; Stroud, R.M.; Schulten, K. Control of the selectivity of the aquaporin water channel family by global orientational tuning. Science 2002, 296, 525–530. [CrossRef][PubMed] 20. Lee, J.K.; Kozono, D.; Remis, J.; Kitagawa, Y.; Agre, P.; Stroud, R.M. Structural basis for conductance by the archaeal aquaporin AqpM at 1.68 Å. Proc. Natl. Acad. Sci. USA 2005, 102, 18932–18937. [CrossRef][PubMed] 21. Savage, D.F.; Stroud, R.M. Structural Basis of Aquaporin Inhibition by Mercury. J. Mol. Biol. 2007, 368, 607–617. [CrossRef][PubMed] 22. Shi, L.B.; Skach, W.R.; Verkman, A.S. Functional independence of monomeric CHIP28 water channels revealed by expression of wild-type mutant heterodimers. J. Biol. Chem. 1994, 269, 10417–10422. [PubMed] 23. Verkman, A.S. More than just water channels: Unexpected cellular roles of aquaporins. J. Cell Sci. 2005, 118, 3225–3232. [CrossRef][PubMed] 24. Manley, G.T.; Binder, D.K.; Papadopoulos, M.C.; Verkman, A.S. New insights into water transport and edema in the central nervous system from phenotype analysis of aquaporin-4 null mice. Neuroscience 2004, 129, 983–991. [CrossRef][PubMed] 25. Verkman, A.S. Physiological importance of aquaporins: Lessons from knockout mice. Curr. Opin. Nephrol. Hypertens. 2000, 9, 517–522. [CrossRef][PubMed] 26. Badaut, J.; Brunet, J.F.; Regli, L. Aquaporins in the brain: From aqueduct to “multi-duct”. Metab. Brain Dis. 2007, 22, 251–263. [CrossRef][PubMed] 27. Nielsen, S.; Smith, B.L.; Christensen, E.I.; Agre, P. Distribution of the aquaporin CHIP in secretory and resorptive epithelia and capillary endothelia. Proc. Natl. Acad. Sci. USA 1993, 90, 7275–7279. [CrossRef] [PubMed] 28. Oshio, K.; Watanabe, H.; Song, Y.; Verkman, A.S.; Manley, G.T. Reduced cerebrospinal fluid production and intracranial pressure in mice lacking choroid plexus water channel Aquaporin-1. FASEB J. 2005, 19, 76–78. [CrossRef][PubMed] 29. Verkman, A.S. Novel roles of aquaporins revealed by phenotype analysis of knockout mice. Rev. Physiol. Biochem. Pharmacol. 2005, 155, 31–55. [PubMed] 30. Nielsen, S.; Frokiaer, J.; Marples, D.; Kwon, T.H.; Agre, P.; Knepper, M.A. Aquaporins in the kidney: From molecules to medicine. Physiol. Rev. 2002, 82, 205–244. [CrossRef][PubMed] 31. Kim, S.W.; Wang, W.; Kwon, T.H.; Knepper, M.A.; Frøkiær, J.; Nielsen, S. Increased expression of ENaC subunits and increased apical targeting of AQP2 in the kidneys of spontaneously hypertensive rats. Am. J. Physiol. Ren. Physiol. 2005, 289, F957–F968. [CrossRef][PubMed] 32. Verkman, A.S. Applications of aquaporin inhibitors. Drug News Perspect. 2001, 14, 412–420. [CrossRef] [PubMed] 33. Verkman, A.S. Role of aquaporin water channels in eye function. Exp. Eye Res. 2003, 76, 137–143. [CrossRef] 34. Zhang, D.; Vetrivel, L.; Verkman, A.S. Aquaporin deletion in mice reduces intraocular pressure and aqueous fluid production. J. Gen. Physiol. 2002, 119, 561–569. [PubMed] 35. Huber, V.J.; Tsujita, M.; Nakada, T. Aquaporins in drug discovery and pharmacotherapy. Mol. Asp. Med. 2012, 33, 691–703. [CrossRef][PubMed] 36. Verkman, A.S. Mammalian aquaporins: Diverse physiological roles and potential clinical significance. Expert Rev. Mol. Med. 2008, 10, e13. [CrossRef][PubMed] 37. Verkman, A.S. Aquaporins in clinical medicine. Annu. Rev. Med. 2012, 63, 303–316. [CrossRef][PubMed] 38. Castle, N.A. Aquaporins as targets for drug discovery. Drug Discov. Today 2005, 10, 485–493. [CrossRef] 39. Verkman, A.S.; Anderson, M.O.; Papadopoulos, M.C. Aquaporins: Important but elusive drug targets. Nat. Rev. Drug Discov. 2014, 13, 259–277. [CrossRef][PubMed] 40. Papadopoulos, M.C.; Saadoun, S.; Verkman, A.S. Aquaporins and cell migration. Pflugers Arch. 2008, 456, 693–700. [CrossRef][PubMed] Int. J. Mol. Sci. 2016, 17, 449 17 of 20

41. Saadoun, S.; Papadopoulos, M.C.; Watanabe, H.; Yan, D.; Manley, G.T.; Verkman, A.S. Involvement of aquaporin-4 in astroglial cell migration and glial scar formation. J. Cell Sci. 2005, 118, 5691–5698. [CrossRef] [PubMed] 42. Hara-Chikuma, M.; Verkman, A.S. Aquaporin-1 facilitates epithelial cell migration in kidney proximal tubule. J. Am. Soc. Nephrol. 2006, 17, 39–45. [CrossRef][PubMed] 43. Hara-Chikuma, M.; Verkman, A.S. Aquaporin-3 facilitates epidermal cell migration and proliferation during wound healing. J. Mol. Med. 2008, 86, 221–231. [CrossRef][PubMed] 44. Cao, C.; Sun, Y.; Healey, S.; Bi, Z.G.; Hu, G.; Wan, S.; Kouttab, N.; Chu, W.M.; Wan, Y.S. EGFR-mediated expression of aquaporin-3 is involved in human skin fibroblast migration. Biochem. J. 2006, 400, 225–234. [CrossRef][PubMed] 45. Loitto, V.M.; Huang, C.; Sigal, Y.J.; Jacobson, K. Filopodia are induced by aquaporin-9 expression. Exp. Cell Res. 2007, 313, 1295–1306. [CrossRef][PubMed] 46. Loitto, V.M.; Karlsson, T.; Magnusson, K.E. Water flux in cell motility: Expanding the mechanisms of membrane protrusion. Cell Motil. Cytoskelet. 2009, 66, 237–247. [CrossRef][PubMed] 47. Karlsson, T.; Lagerholm, B.C.; Vikstrom, E.; Loitto, V.M.; Magnusson, K.E. Water fluxes through aquaporin-9 prime epithelial cells for rapid wound healing. Biochem. Biophys. Res. Commun. 2013, 430, 993–998. [CrossRef] [PubMed] 48. Karlsson, T.; Bolshakova, A.; Magalhaes, M.A.; Loitto, V.M.; Magnusson, K.E. Fluxes of water through aquaporin 9 weaken membrane-cytoskeleton anchorage and promote formation of membrane protrusions. PLoS ONE 2013, 8, e59901. [CrossRef][PubMed] 49. Vandergoot, F.G.; Ausio, J.; Wong, K.R.; Pattus, F.; Buckley, J.T. Dimerization stabilizes the pore-forming toxin aerolysin in solution. J. Biol. Chem. 1993, 268, 18272–18279. 50. Verkman, A.S.; Hara-Chikuma, M.; Papadopoulos, M.C. Aquaporins—New players in cancer biology. J. Mol. Med. 2008, 86, 523–529. [CrossRef][PubMed] 51. Huebert, R.C.; Vasdev, M.M.; Shergill, U.; Das, A.; Huang, B.Q.; Charlton, M.R.; LaRusso, N.F.; Shah, V.H. Aquaporin-1 facilitates angiogenic invasion in the pathological neovasculature that accompanies cirrhosis. Hepatology 2010, 52, 238–248. [CrossRef][PubMed] 52. Saadoun, S.; Papadopoulos, M.C.; Hara-Chikuma, M.; Verkman, A.S. Impairment of angiogenesis and cell migration by targeted aquaporin-1 gene disruption. Nature 2005, 434, 786–792. [CrossRef][PubMed] 53. Vacca, A.; Frigeri, A.; Ribatti, D.; Nicchia, G.P.; Nico, B.; Ria, R.; Svelto, M.; Dammacco, F. Microvessel overexpression of aquaporin 1 parallels bone marrow angiogenesis in patients with active multiple myeloma. Br. J. Haematol. 2001, 113, 415–421. [CrossRef][PubMed] 54. Hu, J.; Verkman, A.S. Increased migration and metastatic potential of tumor cells expressing aquaporin water channels. FASEB J. 2006, 20, 1892–1894. [CrossRef][PubMed] 55. Nicchia, G.P.; Stigliano, C.; Sparaneo, A.; Rossi, A.; Frigeri, A.; Svelto, M. Inhibition of aquaporin-1 dependent angiogenesis impairs tumour growth in a mouse model of melanoma. J. Mol. Med. 2013, 91, 613–623. [CrossRef][PubMed] 56. Nicchia, G.P.; Srinivas, M.; Li, W.; Brosnan, C.F.; Frigeri, A.; Spray, D.C. New possible roles for aquaporin-4 in astrocytes: Cell cytoskeleton and functional relationship with connexin43. FASEB J. 2005, 19, 1674–1676. [CrossRef][PubMed] 57. La Porta, C. AQP1 is not only a water channel: It contributes to cell migration through Lin7/β-catenin. Cell Adhes. Migr. 2010, 4, 204–206. [CrossRef] 58. Bergers, G.; Hanahan, D. Modes of resistance to anti-angiogenic therapy. Nat. Rev. Cancer 2008, 8, 592–603. [CrossRef][PubMed] 59. Carmeliet, P. Angiogenesis in life, disease and medicine. Nature 2005, 438, 932–936. [CrossRef][PubMed] 60. Prager, G.W.; Unseld, M.; Zielinski, C.C. Angiogenesis in cancer: Anti-VEGF escape mechanisms. Transl. Lung Cancer Res. 2012, 1, 14–25. [CrossRef][PubMed] 61. Casanovas, O.; Hicklin, D.J.; Bergers, G.; Hanahan, D. Drug resistance by evasion of antiangiogenic targeting of VEGF signaling in late-stage pancreatic islet tumors. Cancer Cell 2005, 8, 299–309. [CrossRef][PubMed] 62. Zhang, R.; Van Hoek, A.N.; Biwersi, J.; Verkman, A.S. A point mutation at cysteine 189 blocks the water permeability of rat kidney water channel CHIP28k. Biochemistry 1993, 32, 2938–2941. [CrossRef][PubMed] 63. Preston, G.M.; Jung, J.S.; Guggino, W.B.; Agre, P. The mercury-sensitive residue at cysteine 189 in the CHIP28 water channel. J. Biol. Chem. 1993, 268, 17–20. [PubMed] Int. J. Mol. Sci. 2016, 17, 449 18 of 20

64. Niemietz, C.M.; Tyerman, S.D. New potent inhibitors of aquaporins: Silver and gold compounds inhibit aquaporins of plant and human origin. FEBS Lett. 2002, 531, 443–447. [CrossRef] 65. Yukutake, Y.; Tsuji, S.; Hirano, Y.; Adachi, T.; Takahashi, T.; Fujihara, K.; Agre, P.; Yasui, M.; Suematsu, M. Mercury chloride decreases the water permeability of aquaporin-4-reconstituted proteoliposomes. Biol. Cell 2008, 100, 355–363. [CrossRef][PubMed] 66. Brooks, H.L.; Regan, J.W.; Yool, A.J. Inhibition of aquaporin-1 water permeability by tetraethylammonium: Involvement of the loop E pore region. Mol. Pharmacol. 2000, 57, 1021–1026. [PubMed] 67. Brooks, H.L.; Regan, J.W.; Yool, A.J. Inhibition of aquaporin-1 water permeability by TEA. FASEB J. 1999, 13, A394–A394. 68. Detmers, F.J.M.; de Groot, B.L.; Müller, E.M.; Hinton, A.; Konings, I.B.M.; Sze, M.; Flitsch, S.L.; Grubmüller, H.; Deen, P.M.T. Quaternary ammonium compounds as water channel blockers: Speciﬁcity, potency, and site of action. J. Biol. Chem. 2006, 281, 14207–14214. [CrossRef][PubMed] 69. Yukutake, Y.; Hirano, Y.; Suematsu, M.; Yasui, M. Rapid and reversible inhibition of aquaporin-4 by zinc. Biochemistry 2009, 48, 12059–12061. [CrossRef][PubMed] 70. Migliati, E.; Meurice, N.; DuBois, P.; Fang, J.S.; Somasekharan, S.; Beckett, E.; Flynn, G.; Yool, A.J. Inhibition of aquaporin-1 and aquaporin-4 water permeability by a derivative of the loop diuretic bumetanide acting at an internal pore-occluding binding site. Mol. Pharmacol. 2009, 76, 105–112. [CrossRef][PubMed] 71. Xiang, Y.; Ma, B.; Li, T.; Gao, J.W.; Yu, H.M.; Li, X.J. Acetazolamide inhibits aquaporin-1 protein expression and angiogenesis. Acta Pharmacol. Sin. 2004, 25, 812–816. [PubMed] 72. Ma, B.; Xiang, Y.; Mu, S.M.; Li, T.; Yu, H.M.; Li, X.J. Effects of acetazolamide and anordiol on osmotic water permeability in AQP1-cRNA injected Xenopus oocyte. Acta Pharmacol. Sin. 2004, 25, 90–97. [PubMed] 73. Huber, V.J.; Tsujita, M.; Yamazaki, M.; Sakimura, K.; Nakada, T. Identiﬁcation of arylsulfonamides as Aquaporin 4 inhibitors. Bioorg. Med. Chem. Lett. 2007, 17, 1270–1273. [CrossRef][PubMed] 74. Huber, V.J.; Tsujita, M.; Kwee, I.L.; Nakada, T. Inhibition of aquaporin 4 by antiepileptic drugs. Biorg. Med. Chem. 2009, 17, 418–424. [CrossRef][PubMed]

75. Yang, B.; Kim, J.K.; Verkman, A.S. Comparative efficacy of HgCl2 with candidate aquaporin-1 inhibitors DMSO, gold, TEA+ and acetazolamide. FEBS Lett. 2006, 580, 6679–6684. [CrossRef][PubMed] 76. Seeliger, D.; Zapater, C.; Krenc, D.; Haddoub, R.; Flitsch, S.; Beitz, E.; Cerda, J.; de Groot, B.L. Discovery of novel human aquaporin-1 blockers. ACS Chem. Biol. 2013, 8, 249–256. [CrossRef][PubMed] 77. Tornroth-Horsefield, S.; Wang, Y.; Hedfalk, K.; Johanson, U.; Karlsson, M.; Tajkhorshid, E.; Neutze, R.; Kjellbom, P. Structural mechanism of plant aquaporin gating. Nature 2006, 439, 688–694. [CrossRef][PubMed] 78. Gunnarson, E.; Zelenina, M.; Aperia, A. Regulation of brain aquaporins. Neuroscience 2004, 129, 947–955. [CrossRef][PubMed] 79. Gunnarson, E.; Zelenina, M.; Axehult, G.; Song, Y.; Bondar, A.; Krieger, P.; Brismar, H.; Zelenin, S.; Aperia, A. Identification of a molecular target for glutamate regulation of astrocyte water permeability. Glia 2008, 56, 587–596. [CrossRef][PubMed] 80. Zelenina, M.; Zelenin, S.; Bondar, A.A.; Brismar, H.; Aperia, A. Water permeability of aquaporin-4 is decreased by protein kinase C and dopamine. Am. J. Physiol. Ren. Physiol. 2002, 283, F309–F318. [CrossRef] [PubMed] 81. Ho, J.D.; Yeh, R.; Sandstrom, A.; Chorny, I.; Harries, W.E.C.; Robbins, R.A.; Miercke, L.J.W.; Stroud, R.M. Crystal structure of human aquaporin 4 at 1.8 A and its mechanism of conductance. Proc. Natl. Acad. Sci. USA 2009, 106, 7437–7442. [CrossRef][PubMed] 82. Assentoft, M.; Kaptan, S.; Fenton, R.A.; Hua, S.Z.; de Groot, B.L.; Macaulay, N. Phosphorylation of rat aquaporin-4 at Ser111 is not required for channel gating. Glia 2013, 61, 1101–1112. [CrossRef][PubMed] 83. Sachdeva, R.; Singh, B. Phosphorylation of Ser-180 of rat aquaporin-4 shows marginal affect on regulation of water permeability: Molecular dynamics study. J. Biomol. Struct. Dyn. 2014, 32, 555–566. [CrossRef] [PubMed] 84. Reichow, S.L.; Gonen, T. Noncanonical Binding of Calmodulin to Aquaporin-0: Implications for Channel Regulation. Structure 2008, 16, 1389–1398. [CrossRef][PubMed] 85. Reichow, S.L.; Clemens, D.M.; Freites, J.A.; Németh-Cahalan, K.L.; Heyden, M.; Tobias, D.J.; Hall, J.E.; Gonen, T. Allosteric mechanism of water-channel gating by Ca2+-calmodulin. Nat. Struct. Mol. Biol. 2013, 20, 1085–1092. [CrossRef][PubMed] Int. J. Mol. Sci. 2016, 17, 449 19 of 20

86. Rees, D.C.; Congreve, M.; Murray, C.W.; Carr, R. Fragment-based lead discovery. Nat. Rev. Drug Discov. 2004, 3, 660–672. [CrossRef][PubMed] 87. Larsson, A.; Jansson, A.; Aberg, A.; Nordlund, P. Efficiency of hit generation and structural characterization in fragment-based ligand discovery. Curr. Opin. Chem. Biol. 2011, 15, 482–488. [CrossRef][PubMed] 88. Siegal, G.; Ab, E.; Schultz, J. Integration of fragment screening and library design. Drug Discov. Today 2007, 12, 1032–1039. [CrossRef][PubMed] 89. Carr, R.A.E.; Congreve, M.; Murray, C.W.; Rees, D.C. Fragment-based lead discovery: Leads by design. Drug Discov. Today 2005, 10, 987–992. [CrossRef] 90. Wenlock, M.C.; Austin, R.P.; Barton, P.; Davis, A.M.; Leeson, P.D. A comparison of physiochemical property profiles of development and marketed oral drugs. J. Med. Chem. 2003, 46, 1250–1256. [CrossRef][PubMed] 91. Navratilova, I.; Hopkins, A.L. Fragment screening by surface plasmon resonance. ACS Med. Chem. Lett. 2010, 1, 44–48. [CrossRef][PubMed] 92. Fruh, V.; Zhou, Y.; Chen, D.; Loch, C.; Ab, E.; Grinkova, Y.N.; Verheij, H.; Sligar, S.G.; Bushweller, J.H.; Siegal, G. Application of fragment-based drug discovery to membrane proteins: identification of ligands of the integral membrane enzyme DsbB. Chem. Biol. 2010, 17, 881–891. [CrossRef][PubMed] 93. Chen, D.; Errey, J.C.; Heitman, L.H.; Marshall, F.H.; Ijzerman, A.P.; Siegal, G. Fragment screening of GPCRs

using biophysical methods: identiﬁcation of ligands of the adenosine A2A receptor with novel biological activity. ACS Chem. Biol. 2012, 7, 2064–2073. [CrossRef][PubMed] 94. Chen, D.; Ranganathan, A.; AP, I.J.; Siegal, G.; Carlsson, J. Complementarity between in silico and biophysical

screening approaches in fragment-based lead discovery against the A2A adenosine receptor. J. Chem. Inf. Mod. 2013, 53, 2701–2714. [CrossRef][PubMed] 95. Lambruschini, C.; Veronesi, M.; Romeo, E.; Garau, G.; Bandiera, T.; Piomelli, D.; Scarpelli, R.; Dalvit, C. Development of fragment-based n-FABS NMR screening applied to the membrane enzyme FAAH. ChemBioChem 2013, 14, 1611–1619. [CrossRef][PubMed] 96. Brandts, J.F.; Lin, L.N. Study of strong to ultratight protein interactions using differential scanning calorimetry. Biochemistry 1990, 29, 6927–6940. [CrossRef][PubMed] 97. To, J.; Yeo, C.Y.; Soon, C.H.; Torres, J. A generic high-throughput assay to detect aquaporin functional mutants: Potential application to discovery of aquaporin inhibitors. Biochim. Biophys. Acta Gen. Subj. 2015, 1850, 1869–1876. [CrossRef][PubMed] 98. Vedadi, M.; Niesen, F.H.; Allali-Hassani, A.; Fedorov, O.Y.; Finerty, P.J.; Wasney, G.A.; Yeung, R.; Arrowsmith, C.; Ball, L.J.; Berglund, H.; et al. Chemical screening methods to identify ligands that promote protein stability, protein crystallization, and structure determination. Proc. Natl. Acad. Sci. USA 2006, 103, 15835–15840. [CrossRef][PubMed] 99. Wittig, I.; Braun, H.P.; Schägger, H. Blue native PAGE. Nat. Protocols 2006, 1, 418–428. [CrossRef][PubMed] 100. Heuberger, E.H.M.L.; Veenhoff, L.M.; Duurkens, R.H.; Friesen, R.H.E.; Poolman, B. Oligomeric state of membrane transport proteins analyzed with blue native electrophoresis and analytical ultracentrifugation. J. Mol. Biol. 2002, 317, 591–600. [CrossRef][PubMed] 101. Gan, S.W.; Vararattanavech, A.; Nordin, N.; Eshaghi, S.; Torres, J. A cost-effective method for simultaneous homo-oligomeric size determination and monodispersity conditions for membrane proteins. Anal. Biochem. 2011, 416, 100–106. [CrossRef][PubMed] 102. Zeidel, M.L.; Ambudkar, S.V.; Smith, B.L.; Agre, P. Reconstitution of functional water channels in liposomes containing purified red cell CHIP28 protein. Biochemistry 1992, 31, 7436–7440. [CrossRef][PubMed] 103. Giannetti, A.M. From eperimental design to validated hits: A comprehensive walk-through of fragment lead identification using surface plasmon resonance. Methods Enzymol. 2011, 493, 169–218. [PubMed] 104. Altintas, Z.; Uludag, Y.; Gurbuz, Y.; Tothill, I. Development of surface chemistry for surface plasmon resonance based sensors for the detection of proteins and DNA molecules. Anal. Chim. Acta 2012, 712, 138–144. [CrossRef][PubMed] 105. Hann, M.M.; Leach, A.R.; Harper, G. Molecular complexity and its impact on the probability of finding leads for drug discovery. J. Chem. Inf. Comput. Sci. 2001, 41, 856–864. [CrossRef][PubMed] 106. Klein, J.; Meinecke, R.; Mayer, M.; Meyer, B. Detecting binding affinity to immobilized receptor proteins in compound libraries by HR-MAS STD NMR. J. Am. Chem. Soc. 1999, 121, 5336–5337. [CrossRef] 107. Mayer, M.; Meyer, B. Characterization of ligand binding by saturation transfer difference NMR spectroscopy. Angew. Chem. Int. Ed. 1999, 38, 1784–1788. [CrossRef] Int. J. Mol. Sci. 2016, 17, 449 20 of 20

108. Agre, P.; Smith, B.L.; Preston, G.M. ABH and Colton blood group antigens on aquaporin-1, the human red cell water channel protein. Transfus. Clin. Biol. 1995, 2, 303–308. [CrossRef] 109. Chou, C.L.; Knepper, M.A.; Van Hoek, A.N.; Brown, D.; Yang, B.; Ma, T.; Verkman, A.S. Reduced water permeability and altered ultrastructure in thin descending limb of Henle in aquaporin-1 null mice. J. Clin. Investig. 1999, 103, 491–496. [CrossRef][PubMed] 110. Olson, F.; Hunt, C.A.; Szoka, F.C.; Vail, W.J.; Papahadjopoulos, D. Preparation of liposomes of deﬁned size distribution by extrusion through polycarbonate membranes. BBA Biomembr. 1979, 557, 9–23. [CrossRef]

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons by Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).