Specific (SULT1E1) substrates and molecular imaging probe candidates

Graham B. Colea, Gyochang Keuma, Jie Liua, Gary W. Smallb, Nagichettiar Satyamurthya, Vladimir Kepea, and Jorge R. Barrioa,1

aDepartments of Molecular and Medical Pharmacology and bPsychiatry and Biobehavioral Sciences, David Geffen School of Medicine at University of California, Los Angeles, CA 90095-6948

Communicated by Michael E. Phelps, University of California, Los Angeles, Los Angeles, CA, December 28, 2009 (received for review July 24, 2009)

This work focuses on the development of specific substrates for estrogen sulfotransferase (SULT1E1) to produce molecular imaging probes for this . SULT1E1 is a key enzyme in estrogen homeostasis, playing a central role in the prevention and develop- ment of human disease. In vitro sulfation assays showed alkyl and aryl substitutions to a fused heterocyclic system modeled after β- naphthol (βN), based on compounds that interact with the , rendered several molecules with enhanced specificity for SULT1E1 over SULT1A1*1, SULT1A1*2, SULT1A3, and SULT2A1. Sev- eral 6-hydroxy-2-arylbenzothiazoles tested demonstrated excel- —V ∕K — K lent affinity max m ratios and specificity for SULT1E1. m values ranged from 0.12–2.36 μM. A strong correlation was ob- served between polarity of the 4′-sustituent on the 2-aryl moiety σ ðV ∕K Þ r ¼ 0 964 Fig. 1. (A) Interplay of estrogen sulfation and estrogen receptor (ER). Tissue (Hammett p) and the log max m ( . ). sensitiv- levels of E2 are regulated by SULT1E1 and STS, thus, modulating E2 interac- ity is influenced by the acidity of the 6-phenolic group tion with the estrogen receptor (ER). (B): of human SULT1E1 with 1 δ demonstrated by correlating its H NMR chemical shift ( OH) with E2 (Green) bound in the active site. Amino acid residues that form the binding ðV ∕K Þ r ¼ 0 963 the log max m ( . ). Acidity is mediated by the electron pocket are show in Beige (oxygen, Red; nitrogen, Blue; , Red; and free withdrawing capacity of the 4′-substituent outlined by the correla- water molecures, Cyan) [Modified with permission from (25).© National 13 δ ðV ∕K Þ tion of the C-2 C NMR chemical shift ( C2) with the log max m Institute of Environmental Health Sciences, 2003]. (r ¼ 0.987). 2-[4-(Methylamino)phenyl]-6-hydroxybenzothiazole (2b) was radiolabeled with carbon-11 (11C-ð2bÞ) and used in vivo Three hSULTs can metabolize : SULT1A1 (EC2.8.2.1), for microPET scanning and tissue metabolite identification. High SULT1E1 (EC2.8.2.4), and SULT2A1 (EC2.8.2.2) (1), but only PET signal was paralleled with the presence of radiolabeled SULT1E1 possesses sufficient affinity for sulfoconjugation of 11 C-ð2bÞ-6-O- and the SULT1E1 protein detected by western 17β- (E2)[Km ¼ 5 nM (4)] to allow its inactivation at a blot. Because this and other members of this family presenting concentration range that competes with binding to estrogen specificity for SULT1E1 can be labeled with carbon-11 or fluorine-18, receptors(ER)[KD ¼ 1 nM (4)]. E2 is released from its inactive in vivo assays of SULT1E1 functional activity are now feasible in 3-O-sulfated (E2S) form by (STS) (Fig. 1A), an humans. enzyme also widely distributed throughout the body (3). E2 and other estrogens interact with target tissues (or immune molecular imaging probes ∣ positron emission tomography ∣ Pittsburgh cell types) primarily via estrogen receptors (ERα and ERβ). They Compound B ∣ PIB ∣ 18F-Flutemetamol have tissue dependent effects, for example, contributing to bone homeostasis (5), cardiovascular system protection (6), and regu- ytosolic (SULT ; EC 2.8.2) are lation of inflammation in autoimmune, neurodegenerative, and CPhase II metabolic enzymes that mediate sulfation from a do- other diseases (7). Therefore, estrogen sensitivity depends on nor (3′-phosphoadenosine 5′-phosphosulfate, PAPS) to a variety both the concentration of estrogen receptors as well as of endogenous and exogenous substrates containing aryl alcohol, SULT1E1/STS regulation of the free estrogen in these tissues alkyl alcohol, hydroxylamino, or amino groups (Fig 1A) (1, 2). (8, 9). This role of SULT1E1 (estrogen sulfotransferase or SULTs play an important role in the homeostasis of the body EST) as a “molecular switch” (8) is evident in estrogen sensitive in two ways. First, by sulfoconjugating drugs and xenobiotic com- cancers such as breast and endometrial cancer. In re- pounds for their removal via hepatobilliary and urinary systems; sponsive cancers, an increase in E2 (10), in part resulting from second, by sulfoconjugating and deactivating endogenous active down regulation or loss of SULT1E1, potently stimulates tumor substances [e.g., dopamine (DA), thyroid , and estro- growth. Alternatively, presence of SULT1E1 in the absence of gens], therefore, participating in their regulation both systemically STS leads to lower recurrence rate and longer survival (10–13). and locally in different organs. In mammalian species, SULTs are Several high-impact neurodegenerative diseases are also being found throughout the body in the gut, liver, kidneys, adrenal and examined for interactions with E2 and other estrogen analogs thyroid glands, lungs, reproductive organs, breast tissue, brain, and based on the assumed interaction of the estrogens with the blood (1, 3). More than 65 distinct SULT enzymes, spanning 33 isoforms, have been identified and characterized, 11 of which Author contributions: G.B.C., V.K., and J.R.B. designed research; G.B.C., G.K., J.L., N.S., are human sulfotransferases (hSULTs) (2) found in three families: and V.K. performed research; G.B.C., N.S., V.K., and J.R.B. analyzed data; and G.B.C., SULT1, SULT2, and SULT4. Although human SULTs sulfoconju- G.K., J.L., G.W.S., N.S., V.K., and J.R.B. wrote the paper. gate multiple classes of hydroxylated molecules, they show in- The authors declare no conflict of interest. creased affinity for specific endogenous substrates, for example, 1To whom correspondence should be addressed. E-mail: [email protected]. SULT1A3 for DA, SULT1B1 for thyroid hormones, SULT1E1 This article contains supporting information online at www.pnas.org/cgi/content/full/ for estrogens, and SULT2A and SULT2B for hydroxysteroids. 0914904107/DCSupplemental.

6222–6227 ∣ PNAS ∣ April 6, 2010 ∣ vol. 107 ∣ no. 14 www.pnas.org/cgi/doi/10.1073/pnas.0914904107 Downloaded by guest on September 25, 2021 ER (14, 15), however, estrogen sulfating enzymes (e.g., (Km ¼ 0.03 μMand0.17μM, resp.) all benzothiazoles, SULT1E1) are rarely, if ever, invoked in these experiments. benzoxazole 1e and benzimidazole 1f showed one to two orders In this work a family of SULT1E1 substrates were synthesized of magnitude higher Km values (0.42–2.75 μM and 2.08–19.3 μM, and investigated for sensitivity and specificity by examining their resp.). Methyl (1g) or phenyl substitution (2a) at carbon-2 of the ability to act as substrates with four other common sulfotrans- 6-hydroxybenzothiazole ring lead to 7- and 60-fold increases in ferases: SULT1A1*1, SULT1A1*2, SULT1A3, and SULT2A1. affinity (lower Km) for SULT1E1, resp., over the parent com- It is of great interest to assess the in vivo status of SULT1E1 pound 1d. However, these compounds have parallel Km effects in estrogen target tissues and to determine changes in SULT1E1 with both SULT1A1 isoforms detracting from any possible spec- expression as a result of inflammation, cancer, and neurodegen- ificity trend. eration. Molecular imaging with PET using optimum SULT1E1 specific substrates radiolabeled with a short-lived positron emit- 2-Aryl 6-(and 5-)hydroxybenzothiazoles. The effect of 4′-substitution ting radioisotope such as carbon-11 (t1∕2 ¼ 20.4 min) or fluorine- was examined in enzyme assays with 2-phenyl-6-(and 5-) 18 (t1∕2 ¼ 109.8 min) would offer a sensitive tool for this assess- hydroxybenzothiazoles (Scheme 1, Table 1) against SULT1E1, ment. Identification of patient populations in which SULT1E1 SULT1A1*1, SULT1A1*2, SULT1A3, and SULT2A1 enzymes. expression is elevated or absent would help define unique adju- With the 6-hydroxybenzothiazole derivatives (2a–2d, 2f–2l), there vant cancer therapies symbiotic to estrogen therapies currently in was a correlation between the polarity of the 4′-substitution on use. Considering the antiinflammatory role of estrogen and the the 2-phenyl ring (Hammett σp) and SULT1E1 substrate sensitiv- role of SULT1E1 in regulating estrogen levels (7), it can also V ∕K A ity ( max m) (Fig. 2 ) (17). This is due to the electron-donating be anticipated that identification of SULT1E1 in patients with or withdrawing effects of the 4′-substituents on the acidity of the neurodegenerative diseases would have significant diagnostic 6-arylhydroxyl moiety. To corroborate this interpretation, the 1H value and potential therapeutic connotations. δ NMR chemical shift of the hydroxyl group ( OH) is also tightly ðV ∕K Þ C correlated with the log max m kinetic parameter (Fig. 2 ). Results Furthermore, there is an analogous correlation of the 13C Enzyme Assay Validation. Control reactions were used to confirm δ NMR chemical shift of C-2 ( C2) in these benzothiazoles with the activity and identity of the commercially purchased enzymes. ðV ∕K Þ B the log max m (Fig 2 ). Exceptions to these correlations p-Nitrophenol (PNP) and SULT1A1*1 or SULT1A1*2, E2 and included compounds with large 4′-substituents, such as sulfone SULT1E1, (DHEA) and SULT2A1, V ∕K (2k)oracetamide(2g), which have lower max m values and DA and SULT1A3 gave Km values of 0.1 μM, 0.8 μM, (Table 1) than expected from their electronic character. 0.006 μM, 0.8 μM, and 2.1 μM, respectively, in agreement with All 4′-substituted 6-hydroxybenzothiazoles 2b-n, had no previously published values (4, 16). reactivity with SULT1A3, SULT2A1, or SULT1A1*2 and only acetamido (2g), fluoro (2h), iodo (2i), and nitro (2l) derivatives Fused heterocyclic substrates. None of the fused heterocyclic sub- showed reactivity with SULT1A1*1, the most common SULT1A1 strates (1a–1k) assayed (Scheme 1 and Table S1) provided a pre- allozyme. The 5-hydroxybenzothiazoles 2m and 2n displayed ference for SULT1E1 over either SULT1A1 isoform with Km similar Km for SULT1E1 as the 6-hydroxybenzothiazole counter- values in the μM range (5.91–177 μM). However, they showed parts (2b and 2c); but surprisingly they showed good affinities for significantly lower affinity for SULT1A3 (Km values higher than SULT1A1*1 and SULT1A1*2. Benzothiazoles that were non- 150 μM) or no activity for SULT2A1. Heterocyclic moieties (1) reactive with both SULT1A1 allozymes (2b–2f, 2j–2k) were tested accepted as substrates by SULT1E1 include: benzofurans (1b and from low nanomolar to low picomolar concentrations to confirm 1i), benzothiophenes (1c and 1j), benzoxazole (1e), and ben- that these are not high-affinity substrates undergoing substrate zothiazoles (1d, 1g,and1k). Heterocycles with a proton on inhibition, a hallmark of sulfotransferase enzymes (1, 18). the ring nitrogen (e.g., indoles 1a and 1h and benzoimidazole 1f) are not tolerated well because only negligible sulfation activity In vivo data. Representative microPET images of rat (brain) and was observed. All heterocycles 1 were efficient substrates for mouse (whole body) obtained after IV administration with SULT1A1*1 and SULT1A1*2, Km values in the μMorsub- 11C-ð2bÞ are shown in Fig. 3. There was very little variability μMrange(0.023–6.62 μM), yet when compared with βN in uptake and spatial distribution observed between animals with- in the species. SUVR values for rat brain were normalized to the cerebellum. The majority of the SULT1E1 activity was observed in the frontal cortical area (SUVR 1.77 0.35), superior cortical areas (1.72 0.28), subcortical areas (2.39 0.45), and brain stem (1.58 0.18). Areas of high 11C-ð2bÞ retention also corre- sponded with areas showing high proportions of sulfated probe (11C-ð2bÞ-6-O-sulfate) and confirmed presence of the SULT1E1 protein by western blot (Fig. 3 and Table 2). Peripheral organs in mice showed more variability in probe retention, although the spatial distribution was essentially identical between animals. SUVR values were normalized by a 3D isocontour region of in- terest covering the whole mouse. SUVR values (n ¼ 3) included 2.2 0.6 (liver), 8.5 2.5 (gall bladder), 0.37 0.13 (lung), 0.17 0.05 (brain), and 1.8 0.4 (kidney). Organs presenting prominently high expression of SULT1E1 were the liver and kid- ney, in agreement with published reports (1, 3). TLC autoradio- graphs showed a single polar metabolite in all tissue samples tested. This polar metabolite had the same Rf value as that of enzymatically produced 35S-ð2bÞ-6-O-sulfate. Its desulfation could be achieved by treatment with the enzyme (STS) resulting in regeneration of the parent 11C-ð2bÞ. This polar metabolite remained essentially unaffected by the glucuronidase Scheme 1. enzyme (GLU) consistent with a sulfate structure. Occasionally,

Cole et al. PNAS ∣ April 6, 2010 ∣ vol. 107 ∣ no. 14 ∣ 6223 Downloaded by guest on September 25, 2021 Table 1. Michaelis–Menten kinetic parameters for 2-aryl substituted hydroxybenzothiazoles SULT1E1 SULT1A1*1 K V V ∕K K V V ∕K m max max m m max max m 2a 0.99 ± 0.03 3.18 ± 0.87 3.22 ± 0.9 0.018 ± 0.006 4.97 ± 0.57 275 ± 97 2b 1.42 ± 0.12 1.99 ± 0.07 1.40 ± 0.13 * * * 2c 1.42 ± 0.32 2.18 ± 0.08 1.54 ± 0.35 * * * 2d 2.11 ± 0.20 2.88 ± 0.14 1.36 ± 0.15 * * * 2e 1.00 ± 0.06 1.98 ± 0.05 1.97 ± 0.08 * * * 2f 0.90 ± 0.11 1.97 ± 0.08 2.19 ± 0.26 * * * 2g 2.36 ± 0.14 2.04 ± 0.04 0.87 ± 0.05 0.025 ± 0.005 0.88 ± 0.05 36 ± 7 2h 0.56 ± 0.09 4.08 ± 0.25 7.29 ± 1.25 0.029 ± 0.008 0.91 ± 0.09 31 ± 9 2i 0.52 ± 0.05 2.95 ± 0.08 5.64 ± 0.56 0.01 ± 0.003 0.67 ± 0.05 67 ± 21 2j 0.25 ± 0.02 4.33 ± 0.12 17.1 ± 1.4 * * * 2k 1.32 ± 0.08 3.57 ± 0.06 2.69 ± 0.16 * * * 2l 0.12 ± 0.02 3.33 ± 0.19 27.75 ± 4.89 0.058 ± 0.019 1.24 ± 0.09 21 ± 7 2m 0.77 ± 0.08 1.74 ± 0.05 2.26 ± 0.24 ††† 2n 0.43 ± 0.10 1.05 ± 0.08 2.44 ± 0.60 †††

K ¼ 0 045 0 01 V ¼ 5 25 0 55 V ∕K SULT1A1*2 assay with compound 2a yielded m . . and max . . ( max m 117 ± 23); sulfate formation was seen for compounds 2m and 2n, however, kinetic parameters could not be reliably measured. All 2-aryl substituted hydroxybenzothiazoles were assayed against SULT1A3 and SULT2A1 with no detectable activity. Compound 2b was additionally assayed against the SULT1B1 enzyme from 20 nM to 3 μM with no detectable activity. All experiments were performed in K μ V ∕ ∕ V ∕K ð ∕ ∕ Þ∕μ triplicate and all values are expressed as mean ± s.d. Units: m as M, max as nmol min mg, and max m as nmol min mg M. *No detectable activity. †Activity was seen, however kinetic parameters could not be reliably measured.

rat plasma at late time points (30 min after IV injection) showed a shows substantial preference for SULT1A1 (Km values, second, more polar minor metabolite that is consistent with Table S1). Substitutions on ring-A (the ring having the hydroxy 11C-ð2bÞ-6-O-glucuronidate. This second metabolite was not group) serve to decrease catalytic efficiency or convert a substrate found in mouse plasma or any other specimen tested. into a SULT1E1 inhibitor (18, 25). Introduction of an ethyl chain at position 6 (ring-B) on the βN substructure results in a 5-fold Discussion increase in affinity (24) pointing to ring-B modification and elon- Therapeutic agents such as raloxifene and 4-hydroxytamoxifen gation as a favorable structural modification for increased (Scheme 2) (19), as well as naturally occurring dietary flavonoids, SULT1E1 activity. are substrates or inhibitors of sulfotransferases (SULT1A1 and One structural element that is a substrate for both SULT1E1 SULT1E1) to varying degrees (16, 20). Until now, no family of and SULT1A1 is the 2-phenyl-6-hydroxybenzothiophene core structurally related compounds has been reported to provide component of raloxifene (Scheme 2), an estrogen receptor (ER) substrate preference for SULT1E1 as is seen with estrogens. antagonist used as a chemotherapeutic agent (19). Therefore, 5- 2-Aryl-6-hydroxybenzothiazoles 2b-2f, 2j-2k, several of which and 6-hydroxy substituted fused heterocyclic phenols 1a-1k were have been recently reported in connection with amyloid Aβ tested first to examine the extent to which the atomic constituents aggregate detection in the brain of Alzheimer’s disease patients of ring-B would affect the affinity of the resulting fused hetero- (21, 22) or as possessing breast cancer cell cytotoxicity (23), are cyclic phenols for SULT1E1 and other SULTs (1A1*1, 1A1*2, shown here to be highly specific substrates for SULT1E1. These 1A3, 2A1) (Table S1). Comparison of KmðSULT1E1Þ∕ compounds have afforded a unique opportunity to examine the KmðSULT1A1Þ isoform ratios for all compounds 1 with affinity effect of changing substituent electronic properties on the speci- for SULT1E1 clearly shows that only benzothiazoles 1d, 1g, and ficity of SULT1E1. This was previously examined with SULT1A1 1k have Km that ratios are comparable or lower (146, 5.2, and (18) but has remained elusive for SULT1E1. 18.5 for 1A1*1; 29, 1.2, and 2.7 for 1A1*2, resp.) than that of Earlier literature reports (24) identified βN (Scheme 2) as one βN (219 and 34.8 for 1A1*1 and *2, resp.); all other com- of the smallest aromatic substrates for SULT1E1, although it pounds 1 had Kmð1E1Þ∕Kmð1A1Þ ratios above 95. Therefore,

ðV ∕K Þ σ r ¼ 0 964 Fig. 2. (A) Linear correlation between log max m and Hammett p ( . ) for 2-arylsubstituted-6-hydroxybenzothiazole derivatives. Compounds 2g 0 0 (4 ¼ NHCOCH3) and 2k (SO2-CH3) are apparent outliers due to possible steric hindrance of the large 4 -substituents affecting binding with SULT1E1. (B) Linear δ ðV ∕K r ¼ 0 987 correlation between C2 and the log max m)( . ). Compounds 2g and 2k are apparent outliers due to steric hindrance. (C) 6-Hydroxy proton δ ðV ∕K Þ r ¼ 0 963 chemical shifts ( OH) and the log max m ( . ). Compounds 2g and 2k are outliers due to steric hindrance. All numbers represent the average of triplicate values.

6224 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0914904107 Cole et al. Downloaded by guest on September 25, 2021 Scheme 2. Fig. 3. (Top) Representative microPET scans in rat brain (Top, Left) and whole-body mice (Top, Right) with 11C-ð2bÞ co-registered with microCT images. The animals were scanned in a dynamic mode for 25 min. (Bottom): X-ray crystallographic analysis of hSULT1E1 co-crystallyzed Western blot data obtained from the same tissues. Lane 1:80μg frontal cor- with E2 and PAP reveals a hydrophobic pocket with E2’s aromatic tical area (Cortex 1); Lane 2:80μg superior cortical areas (Cortex 2); Lane 3: ring inserted in the active site through a “gate” consisting of two 80 μg subcortical areas; Lane 4:80μg cerebellum; Lane 5:80μg brain stem; phenylalanine residues (F80 and F141, Fig. 1B). These features μ μ μ Lane 6:50 g testis; Lane 7 100 g testis; and Lane 8: 150 g testis control the orientation of the substrate and position the phenolic 3-OH group in the proximity of the catalytic histidine residue benzothiazoles were further explored for SULT1E1 substrate (H107) (25, 27). 4′-Substitutions to the 2-aryl-6-hydroxyben- specificity and other compounds 1, that is, indoles, benzofurans, zothiazole appear to provide the necessary anchoring elements, benzothiophenes, benzoxazoles, and benzimidazoles were ex- in addition to hydrophobic interactions within the binding pocket, cluded from further examination. for a catalytically competent reaction. This adds credence to the It is notable that all 4′-substituted 2-aryl 6-hydroxybenzothia- possibility that the 4′-substituent may be interacting with amino zoles 2b-2l lacked substrate affinity for SULT1A1*2, SULT1A3, acid residues in the outer regions of the substrate binding pocket. and SULT2A1. Additionally, excluding acetamido (2g), fluoro It is known from crystal structure determinations that differences (2h), iodo (2i), and nitro derivatives (2l), they also presented a in the outer region of the substrate binding pocket, both at the very limited affinity for the SULT1A1*1 allozyme. Thus, 2-aryl amino acid level and at secondary structural levels, are respon- 6-hydroxybenzothiazoles with 4′-electron withdrawing substitu- sible for endogenous substrate specificity among different SULTs. ents (2j, 2k) and all compounds with electron-donating substitu- For example, Asn87 is in a small, two-strand β-sheet that forms a ents (2b-2d, 2f), in addition to the structurally related compound hydrogen bond with the 17-OH group of E2 on SULT1E1, Glu146 2e, exhibited complete specificity for SULT1E1 (Table 1). and Glu89, which form hydrogen bonds with the amino group of V ∕K Enzyme sensitivity (based on max m) of the analogs 2a-2d DA on SULT1A3, or two loops that close over the SULT1A1 sub- and 2f-2l with SULT1E1 was strongly correlated with the electro- strate binding pocket (25, 28). negativity of the 4′-substituent. Electron withdrawing groups The apparent lack of substrate affinity for SULT1A1 displayed σ V ∕K (higher Hammett p) provided better kinetics (higher max m) by 6-hydroxybenzothiazoles 2b-f and2j,k as opposed to high sub- with the 4′-nitro substitution (2l) being most efficient (Fig. 2A, strate affinity for SULT1A1 displayed by compounds 2a, 2g-i, and r ¼ 0 964 K V . ). Both m and max were affected by the electronega- 2l is somewhat surprising. Based on crystallographic analysis of tivity of the 4′-substitution on the 2-aryl group. SULT1A1 co-crystallized with PAP and PNP (29) or E2 (30), it δ Because OH represents the extent of the dissociation of the is evident that the SULT1A1 substrate binding pocket can under- hydroxyl group, or the pKa of the phenol, the transference of go significant plastic deformations to accommodate a variety of the sulfate group from the PAPS to the substrate is structurally different substrates. The outer loops, 146–154 and facilitated by electronic delocalization of the phenolic oxygen 84–90, of SULT1A1 that cover the substrate with BIOCHEMISTRY electron pair (Fig. 2C, r ¼ 0.963) through C-2 (Fig. 2B, PNP would assume a more open conformation to accommodate r ¼ 0.987) mediated by the 4′-substituent. The correlation be- the larger E2, albeit as a weak substrate. When bound to the bind- tween the polarity of 4′-substituents (Hammett σp) and kinetic ing pocket of SULT1A1, 2-aryl substituted 6-hydroxybenzothia- parameters for SULT1A1 has previously been shown for simple zoles (2) may also have the 4′-substituents interacting with phenol derivatives (26); the rate limiting step was said to be the other aminoacid residues in the pocket as suggested with nucleophilic attack of the phenoxide ion onto the sulfate group SULT1E1. Thus, based on their ability to fit optimally into the of PAPS. SULT1A1 binding site, these 6-hydroxybenzothiazoles (2) would

Table 2. Quantitative microPET and ground tissue metabolite data.

RATS (n ¼ 3) MICE (n ¼ 7) 11C-ð2bÞ-6-O-sulfate 11C-ð2bÞ-6-O-sulfate Brain region SUVR* (% of total activity) Organ SUVR† (% of total activity) Cortex1‡ 1.77 ± 0.35 57.4 ± 4.4 Brain 0.17 ± 0.05 3.1 ± 8.2 Cortex2§ 1.72 ± 0.28 65.6 ± 7.7 Lungs 0.37 ± 0.13 72.9 ± 18.9 Subcortical 2.39 ± 0.45 64.8 ± 1.8 Liver 2.2 ± 0.6 82.7 ± 5.1 Cerebellum 1 34.7 ± 4.3 Kidney 1.8 ± 0.4 82.2 ± 11.5 Brainstem 1.58 ± 0.18 35.5 ± 4.7 Testis ¶ 58.9 ± 25.3 Plasma — 66.3 ± 6.2∥ Plasma — 78.5 ± 6.8 *Normalized to cerebellum. †Normalized to whole body. ‡Frontal cortex. §Superior cortical regions. ¶Testis could not be measured within the field of view. ∥A minor, more polar component (7 ± 10%) consistent with 11C-ð2bÞ-6-O-glucuronidate was occasionally found only in rat plasma.

Cole et al. PNAS ∣ April 6, 2010 ∣ vol. 107 ∣ no. 14 ∣ 6225 Downloaded by guest on September 25, 2021 either assume a catalytically competent binding orientation or be fluorine-18, resp., (35, 36) and used for brain imaging Aβ aggre- forced into a catalytically noncompetent binding orientation as gates in Alzheimer’s disease, in vitro as well as in vivo (21, 22). seen in the case of E2.E2 binding to SULT1A1 in a catalytically Similarly, and based on the results of this work, it would be ex- noncompetent orientation results in a dead-end complex hy- pected that other PET probes proposed for amyloid plaque de- pothesized as the cause of the substrate inhibition at higher con- tection in humans having the 6-hydroxybenzothiazole, for centrations (30). example, 2-[6-(methylamino)pyridin-3-yl]-6-hydroxybenzothia- V ∕K σ The two exceptions to the max m versus Hammet p trend zole (37), and the 6-hydroxybenzoxazole skeletons (38) would (Fig. 2) are compounds having sterically demanding groups, for also be good substrates for SULT1E1. The newly discovered example, acetamide (2g) and methylsulfone (2k) (Table 1). This SULT1E1 substrate activity for the family of compounds pre- suggests that the acidity of the phenolic group is not the sole de- sented in this investigation may have important implications re- terminant of substrate efficiency. The poor SULT1E1 substrate garding the previous interpretation of their in vivo behavior as ability of 2g and 2k can be rationalized on the basis that these sub- imaging agents considering that SULT1E1 is widely distributed stituents produce significant steric hindrance to fit optimally in the in humans and its presence has been also demonstrated in the binding pocket of SULT1E1 (27). human brain (3). Further work to this end is warranted. In contrast to 6-hydroxy derivatives, 5-hydroxybenzothiazoles are substrates for both SULT1A1 allozymes at very low substrate Materials and Methods β concentrations, effectively removing the SULT1E1 specificity ob- Materials. N, PNP, E2, DHEA, DA, dithiothreitol (DTT), 2-methyl-5-hydroxy- served with many of the 6-hydroxy counterparts (Table 1). This benzothiazole (1k) Mgcl2, and cytosolic extracts of sf-9 cells infected with SULT1A1 interaction of 5-hydroxybenzothiazoles is remarkable a baculovirus containing SULT cDNAs (SULT1A1*2, SULT1A3, SULT1E1, SUL- considering that the 6-hydroxy counterparts are not SULT1A1 T2A1) were purchased from Sigma. 5-Hydroxyindole (1h) was purchased from Alfa-Aesar and 6-hydroxy-indole (1a) from Acros. E. coli cell lysates contain- substrates. It is not possible, based on the data available, to ing SULT1A1*1 were purchased from Cypex. ½35S-30-Phosphoadenosine determine the exact reasons for this disparity, but NMR data 5′-phosphosulfate (PAPS) (1–3 Ci∕mmol) was purchased from Perkin Elmer. (Table S2) provides a clue by showing significantly different Normal phase, aluminum backed TLC plates were purchased from Whatman. π δ -electronic configurations of 5-hydroxy compounds (e.g., C6 δ and C2) as compared with that of the 6-hydroxy counterparts in- Chemistry. All synthesis methods are described in SI Text and outlined in dicating different electronic interactions with specific amino acid Schemes S1–S3. residues in the binding pocket of SULT1A1, transforming unpro- ductive binding (as is the case with the 6-hydroxy counterparts) In Vitro Enzymatic Assays. Assays performed using each SULT expressed in sf-9 into a binding configuration that would position the 5-OH group cytosolic extracts (SULT1A1*2, SULT1A3, SULT1E1, and SULT2A1) or E. Coli into a catalytically competent orientation. extracts (SULT1A1*1) that were used as provided without additional purifi- To establish a preliminary demonstration of the ability of cation. Initially, compounds were screened using a wide range of concentra- tions (10 nM–25 μM). Next, kinetic parameters were determined (19). 6-hydroxy benzothiazoles to act as in vivo SULT1E1 probes, mi- 11 reactions included the substrate initially dissolved in DMSO (final concentra- croPET imaging was performed in rodents with C-ð2bÞ (Fig. 3). μ 11 tion 0.2%), 50 mM Tris-HCl, pH 7.4, 0.1% BSA, 7.5 mM DTT, and 1 M The SULT1E1-mediated accumulation of substrates [e.g., C-ð2bÞ] ½35S-PAPS (39) in a 50 μL volume. SULT1E1 and SULT2A1 reactions were sup- 11 is based on the principle of metabolic trapping of C-ð2bÞ-6-O- plemented with 7 mM MgCl2; the DA control reaction with SULT1A3 included 18 sulfate analogous to the use of 2-deoxy-2-½ Ffluoro-D-glucose 1 mM pargyline to inhibit endogenous monoamine oxidase activity from the (2-FDG), whose affinity for hexokinase permits tissue formation sf-9 cells. Control reactions contained only the solvent vehicle. Reactions were of its 6-phosphate (31). In agreement with this premise, biochemical carried out at 37 °C for 15 min then were terminated with 50 μL of ice cold analysis on all mice scanned has demonstrated significant levels of MeOH and placing at −80 °C. Twenty μL aliquots were spotted on silica gel 11 ð Þ 6 TLC plates followed by development in 1-BuOH:HOAc:water (8∶1∶1 by C- 2b - -O-sulfate in organs with high PET signal (Table 2). R ¼ 0 66 0 05 Radioactivity accumulation, as observed with microPET in mice, volume) allowing separation of sulfated products ( f . . ) from un- reacted PAPS (Rf ¼ 0) (40). TLC plates were optimally exposed on phosphor is also consistent with areas known to have SULT1E1 expression, imaging plates and read by using a Fuji BAS-5000 digital autoradiography especially in the kidney and liver (1) (Fig. 3). Similar observations plate reader. Spots were localized; the TLC plates scraped into liquid were made with the regional distribution of brain activity upon scintillation vials and counted using MP − Ecoliteþ LS fluid and a Packard Tri- 11 intravenous administration of C-ð2bÞ in rats, also demonstrating carb 2300TR LSC for 5 min per vial. Reactions were monitored for linearity itscapacitytobesulfatedinthebraininvivobecauseno with time, enzyme concentration, and the amount of each substrate used 11C-ð2bÞ-6-O-sulfate of peripheral origin is expected to cross the to ensure the enzyme was the limiting factor. blood brain barrier (32). Although the low affinity glucuronidation pathway (33) could also be involved, the high affinity sulfation path- Data Analysis. Michaelis–Menten kinetics. The sulfate transfer mechanism – way seems to be predominant. This is demonstrated by the presence (18) is assumed to follow Michaelis Menten kinetics: 11 V ½S of C-ð2bÞ-6-O-sulfate as the main radiolabeled component in both max vo ¼ [1] brain and peripheral tissues. Km þ½S K V In summary, this work presents a unique systematic attempt to The curve was fitted to Eq. 1,and m and max values were calculated by design, synthesize, and characterize specific SULT1E1 substrates. using GraphPad PRISM4 statistical analysis software. The electron withdraw- 11 The successful in vivo utilization of C-ð2bÞ also provides direct ing capacity of the 4′-substituent in the 2-aryl moiety of the benzothiazoles V ∕K evidence of the feasibility to determine local tissue measures of was correlated with max m. SULT1E1 activity in intact animals. Added significance of this finding is provided by the fact that ex vivo determination of MicroPET imaging. All animal experiments were performed under the strict SULT1E1 activity is highly problematic due to the post mortem guidelines of the University of California, Los Angeles Animal Research inactivation of cytosolic sulfotransferases (32, 34). Committee. Male Sprague–Dawley rats (n ¼ 3, age 1 yr, weight The use of SULT1E1 specific PET molecular imaging probes 360 g 40) and male Balb/c mice (n ¼ 7, age 3 mo, weight 29 g 2) (Charles has exciting implications for the biochemical characterization of River) were anesthetized via 2% isoflurane (IsoFlo; Abbot Laboratories) in major diseases where inflammation is known to play significant 100% oxygen in an induction box equipped with an inductive heating pad. Animals were weighed and transferred to an imaging chamber roles. Two specific SULT1E1 substrates presented in this equipped with a nose cone for gas delivery and inductive heater (41, 42). work, 2-[4-(methylamino)phenyl]-6-hydroxybenzothiazole (2b, MicroPET scans were acquired with a Focus 220 microPET scanner (Con- Pittsburgh Compound B or PIB) and 2-[3-fluoro-4-(methylami- corde Microsystems, Inc.). Animals were injected with 1.5 mCi (for rats) or 18 no)phenyl]-6-hydroxybenzothiazole (2e, F-flutemetamol or 200 μCi (for mice) of 11C-ð2bÞ at the start of the scan that was performed 18 F-GE067) have been radiolabeled with carbon-11 and for 25 min when the time activity curves (TAC) in all organs of interest ceased

6226 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0914904107 Cole et al. Downloaded by guest on September 25, 2021 to show rapid changes (43) (Figs. S1B and S2B). Images were reconstructed fuged as above. One hundred μL aliquots of supernatants were added to using filtered back projection and a ramp filter. Images were corrected for 1,000 units of steroid sulfatase (STS, EC3.1.6.2) or glucuronidase (GLU, attenuation by using microCT (42). At the termination of the experiment, EC3.2.1.31) and placed at 37 °C for 30 min. Next, samples were diluted 1∶2 1 mL of whole blood was drawn and animals were euthanized via intra ve- with acetonitrile and 20 μL aliquots of each sample were spotted on reverse nous pentobarbital overdose. Blood samples were centrifuged for 5 min to phase TLC plates and developed as indicated above. separate blood plasma. The rat brain was divided in half sagitally and then into five regions (Fig. 3). Mouse tissues were harvested and divided into two equal pieces. Half of all tissues were snap frozen in liquid nitrogen and stored Western blotting. Tissues were homogenized 60 s on ice in 10 mM Tris buf- at −80 °C for western blot analysis; the other half was weighed and counted fer (pH 7.4) with 1 mM EDTA and Roche complete protease inhibitor cocktail for radioactivity prior to ground tissue metabolite analysis. (1 mL∕g), centrifuged at 100; 000 × g for 1 h, and then concentration was de- termined by Bradford assay (BioRad). Protein was resolved on precast Tris-HCl Determination of radioactive probe sulfate content in tissue and plas- 12% gels (BioRad), transferred to PVDF membranes, and probed with anti- SULT1E1 mouse monoclonal antibodies (1∶100, Biovision) visualized by che- ma. Harvested tissues were homogenized on ice in cold acetonitrile (1 mL∕g) to precipitate proteins by using an automated homogenizer (Biospec Pro- miluminesence with HRP-conjugated ani-mouse antibodies (Vector). Testis ducts inc.; Bartlesville, OK) for 60 sec. Samples were transferred to microcen- tissue was used as a positive control because of its high expression of SULT1E1 trifuge tubes and centrifuged for 2 min at 10,000 g. Ten μL aliquots of the (3, 8). supernatant were carefully spotted onto reverse phase (C-18) TLC plates and developed with MeOH:H2O:HOAc (80∶15∶5) allowing separation of the : 11 Statistical methods. All results are shown as mean s d. primary metabolite [ C-ð2bÞ-6-O-sulfate; Rf ¼ 0.71 :05) and secondary 11 metabolite Rf ¼ 0.78 0.03] from the precursor, C-ð2bÞ (Rf ¼ 0.47 :06) (Figs. S1A, S2A, and S3). TLC plates were exposed on phosphor imaging plates ACKNOWLEDGMENTS. We thank L. Petersen (National Institute of Health) for a previously published figure and G. Timbol, A. Smid, and E. Basarah for and read as above. Spots were localized and the percentage of metabolized technical assistance. We acknowledge financial support from the National probe was quantified by using Fuji Multi-Gauge 3.0 software. Institutes of Health (Grant P01AG025831) and the Korean Research Founda- tion (G. Keum, MOEHRDl KRF-2006-611-C00004). J.R.B. gratefully acknowl- Polar metabolite identification. Harvested tissues were homogenized on edges the support of the Elizabeth and Thomas Plott Chair Endowment in ice in cold PBS (1 mL∕g) by using an automated homogenizer then centri- Gerontology.

1. Falany C (1997) Enzymology of human cytosolic sulfotransferases. FASEB J 24. Falany CN, Ström P, Swedmark S (2005) Sulphation of o-desmethylnaproxen and 11(4):206–216. related compounds by human cytosolic sulfotransferases. Br J Clin Pharmacol 2. Blanchard RL (2005) Human cytosolic sulfotransferases. (Taylor and Francis Group, LLC, 60:632–640. Boca Raton, FL), pp 1–8. 25. Shevtsov S, Petrotchenko EV, Pedersen LC, Negishi M (2003) Crystallographic analysis 3. Miki Y, et al. (2002) Systemic distribution of steroid sulfatase and estrogen sulfotrans- of a hydroxylated polychlorinated biphenyl (OH-PCB) bound to the catalytic estrogen ferase in human adult and fetal tissues. J Clin Endocrinol Metab 87(12):5760–5768. binding site of human estrogen sulfotransferase. Environ Health Perspect 4. Zhang H, Varmalova O, Vargas FM, Falany CN, Leyh TS (1998) Sulfuryl transfer: 111:884–888. The catalytic mechanism of human estrogen sulfotransferase. J Biol Chem 26. Marshall AD, McPhie P, Jakoby WB (2000) Redox control of – 273(18):10888 10892. specificity. Arch Biochem Biophys 382:95–104. 5. Syed F, Khosla S (2005) Mechanisms of sex steroid effects on bone. Biochem Biophys 27. Petrotchenko EV, Doerflein ME, Kakuta Y, Pedersen LC, Negishi M (1999) Substrate Res Commun 328:688–696. gating confers steroid specificity to estrogen sulfotransferase. J Biol Chem 6. Xing D, Nozell S, Chen YF, Hage F, Oparil S (2009) Estrogen and mechanisms of vascular 274:30019–30022. protection. Arterioscler Thromb Vasc Biol 29:289–95. 28. Yoshinari K, Petrotchenko EV, Pedersen LC, Negishi M (2001) Crystal structure based 7. Straub RH (2007) The complex role of estrogens in inflammation. Endocr Rev Studies of cytosolic sulfotransferases. J Biochem Mol Toxic 15:67–75. 28:521–574. 29. Gamage NU, et al. (2003) The structure of a human carcinogen-converting 8. Song WC (2001) Biochemistry and reproductive biology of estrogen sulfotransferase. enzyme, SULT1A1 structural and kinetic implications of substrate inhibition. J Biol Ann NY Acad Sci 948:43–50. – 9. Hobkirk R (1985) and : Characteristics Chem 278:7655 7662. and biological roles. Can J Biochem Cell Biol 63:1127–1144. 30. Gamage NU, et al. (2005) The structure of human SULT1A1 crystallyzed with estradiol. 10. Pasqualini JR, Cortes-Prieto J, Chetrite G, Talbi M, Ruiz A (1997) Concentrations of An insight into active site plasticity and substrate inhibition with multi-ring substrates. – , estradiol and their , and evaluation of sulfatase and J Biol Chem 280:41482 41486. activities in patients with breast fibroadenoma. Int J Cancer 70(6):639–643. 31. Barrio JR () PET: Molecular Imaging and Its Biological Applications. (Springer-Verlag

11. Suzuki T, et al. (2003) Estrogen sulfotransferase and steroid sulfatase in human breast New York, Inc., New York, NY) p 279. BIOCHEMISTRY carcinoma. Cancer Res 63(11):2762–2770. 32. Mathis CA, et al. (2004) Species-dependent of the amyloid imaging agent 12. Utsunomiya H, et al. (2004) Steroid sulfatase and estrogen sulfotransferase in human [C-11]PIB. J Nucl Med 45(Suppl):114P. endometrial carcinoma. Clin Cancer Res 10:5850–5856. 33. Raftoganis R, Creveling C, Weinshilboum R, Weisz J (2000) Estrogen metabolism by 13. Sato R, et al. (2009) Steroid sulfatase and estrogen sulfotransferase in colon carcinoma: conjugation. J Natl Cancer Inst Monogr 27:113–124. Regulators of intratumoral estrogen concentrations and potent prognostic sactors. 34. Young WF, Jr, Okazaki H, Laws ER, Jr, Weinshilboum RM (1984) Human brain phenol Cancer Res 69:914–922. sulfotransferase: Biochemical properties and regional localization. J Neurochem 14. Sicotte NL, et al. (2002) Treatment of multiple sclerosis with the pregnancy hormone 43:706–715. . Ann Neurol 52:421–428. 35. Mathis CA, et al. (2003) Synthesis and evaluation of 11C-labeled 6-substituted ’ 15. Baum LW (2005) Sex, hormones, and Alzheimer s disease. J Gerontol A Biol Sci Med Sci 2-arylbenzothiazoles as amyloid imaging agents. J Med Chem 46:2740–2754. – 60:736 743. 36. Solbach C, Uebele M, Reischl G, Machulla HJ (2005) Efficient radiosynthesis of carbon- 16. Nishiyama T, et al. (2002) Sulfation of environmental estrogens by cytosolic human 11 labeled uncharged thioflavin T derivatives using [11C] methyl triflate for beta-amy- – sulfotransferases. Drug Metab Pharmacok 17(3):221 228. loid imaging in Alzheimer’s disease with PET. Appl Radiat Isotopes 62:591–595. 17. Gordon AJ, Ford RA The chemist’s companion: A handbook of practical data, 37. Johnson AE, et al. (2009) AZD2184: A radioligand for sensitive detection of beta- techniques, and references. (John Wiley & Sons, New York, NY), pp 45–47. amyloid deposits. J Neurochem 108:1177–1186. 18. Duffel MW, Jakoby WB (1981) On the mechanism of aryl sulfotransferase. J Biol Chem 38. Malmstroem J, et al. (2007) Preparation of benzoxazole derivatives for imaging 256:11123–11127. amyloid deposits in patients. PCT Int Appl WO2007149030A1. 19. Falany JL, Pilloff DE, Leyh TS, Falany CN (2006) Sulfation of raloxifene and 4-Hydro- 39. Otake Y, Nolan AL, Walle UK, Walle T (2000) Quercetin and resveratrol potently reduce xytamoxifen by human cytosolic sulfotransferases. Drug Metab Dispos 34:361–368. estrogen sulfotransferase activity in normal human mammary epithelial cells. J Steroid 20. Ung D, Nagar S (2007) Variable sulfation of dietary polyphenols by recombinant – human sulfotransferase (SULT) 1A1 genetic variants and SULT1E1. Drug Metab Dispos Biochem Mol Biol 73:265 270. 35:740–746. 40. Tremaine LM, Diamond GL, Quebbemann AJ (1984) In Vivo Quantification of renal p 21. Klunk WE, et al. (2004) Imaging brain amyloid in Alzheimer’s disease with Pittsburgh glucuronide and sulfate conjugation of 1-naphthol and -nitrophenol in the rat. – compound-B. Ann Neurol 55:306–319. Biochem Pharmacol 33:419 427. 18 22. Nelissen N, et al. (2009) Phase 1 study of the Pittsburgh compound B derivative 41. Fueger BJ, et al. (2006) Impact of animal handling on the results of F-FDG PET studies 18F-flutemetamol in healthy volunteers and patients with probable Alzheimer in mice. J Nucl Med 47:999–1006. disease. J Nucl Med 50:1251–1259. 42. Chow PL, Stout DB, Komisopoulou E, Chatziioannou AF (2006) A method of image 23. Shi DF, et al. (1996) Antitumor benzothiazoles. 3. synthesis of 2-(4-aminophenyl) registration for small animal, multi-modality imaging. Phys Med Biol 51:379–390. benzothiazoles and evaluation of their activities against breast cancer cell lines in 43. Lopresti BJ, et al. (2005) Simplified quantification of Pittsburgh compound β-amyloid Vitro and in Vivo. J Med Chem 39:3375–3384. imaging PET studies: A comparative analysis. J Nucl Med 46:1959–1972.

Cole et al. PNAS ∣ April 6, 2010 ∣ vol. 107 ∣ no. 14 ∣ 6227 Downloaded by guest on September 25, 2021