Structural and Chemical Profiling of the Human Cytosolic Sulfotransferases

Abdellah Allali‐Hassani, Patricia W. Pan, L. Dombrovski, Rafaël Najmanovich, W. Tempel, Aiping Dong, P. Loppnau, Fernando A. Martín, Janet Thonton, A.M. Edwards, Alexey Bochkarev, A.N. Plotnikov, Masoud Vedadi, C.H. Arrowsmith•View original

OverviewBalancedjames voice

A researcher holds a vial containing one of twelve nearly identical enzymes, the same fold, the same cofactor, and the same reaction. They ask why one of them activates a carcinogen while its sibling detoxifies it. For most of this enzyme family, no one had a crystal structure, no one had run a systematic chemical screen, and the answer was simply unknown. That’s what Allali-Hassani and colleagues set out to fix. The enzymes in question are cytosolic sulfotransferases, or SULTs. They perform a single chemical transformation: transferring a sulfonate group from a cofactor called PAPS, which stands for three-prime-phosphoadenosine five-prime-phosphosulfate, the universal sulfonate donor, onto small molecules ranging from steroid hormones to drugs to environmental toxins. That transfer usually increases water solubility and flags the molecule for excretion. But sometimes, it does the opposite, converting a dietary or environmental compound into something mutagenic. Humans have twelve distinct cytosolic sulfotransferase proteins, spread across four families. They share a conserved protein fold built around a central four-stranded parallel beta sheet. They have the same architecture and the same chemistry — and yet radically different biological outcomes depending on which enzyme does the sulfonating. Before this work, five of the twelve had no published crystal structure, and no one had systematically screened the whole family against the same set of compounds. That gap is what the study closes. The structural work came first. The team solved crystal structures for five previously uncharacterized human SULTs, including SULT1C3 bound to the cofactor product PAP at a resolution of three point two angstroms, an apo SULT1C2 structure at a resolution of two angstroms, a ternary complex of SULT1C2 with PAP and the environmental toxin pentachlorophenol at a resolution of one point eight angstroms, and SULT4A1 at a resolution of two point two angstroms. Across the family, the global fold is conserved — but what varies are three flexible loop segments: a thirteen-residue loop, a shorter loop of four to ten residues, and a large loop of thirty-two to forty-six residues. In structures without any ligand, these loops are often disordered, appearing as gaps in the electron density map. In PAP-bound structures, they gain order — in some cases, folding into defined helices. One practical limit also emerged: the database sequence for SULT6B1 lacks the N-terminal region encoding a beta sheet critical to the fold, and the protein simply wouldn’t express in bacterial purification attempts. The most important thing the structures revealed, though, wasn’t a static snapshot. It was a dynamic process — and it’s the conceptual heart of the paper. When PAP binds, a helix called alpha-14 to alpha-15 and the C-terminal segment of the largest flexible loop become ordered, positioning three absolutely conserved residues — threonine two hundred twenty-eight, arginine two hundred fifty-eight, and glycine two hundred sixty — that are required for PAPS binding. The cofactor, in other words, organizes the very loops that will later contact the substrate. Allali-Hassani and colleagues call this "priming." Think of it as a lock that subtly changes shape when one key is inserted, so only certain other keys will later fit productively. The enzyme can formally accept its substrate or cofactor in either order — the structural data include binary complexes with either ligand alone, ruling out a strict ordered mechanism. But cofactor binding biases the loop ensemble toward ordered, catalytically competent states. That distinction matters. The priming idea also explains one of the stranger behaviors in this enzyme family: substrate inhibition, where too much substrate actually shuts the enzyme down. Prior explanations focused on two substrate molecules crowding the active site simultaneously. The structures suggest a third mechanism. A SULT2A1 structure containing two molecules of DHEA, which stands for dehydroepiandrosterone, a steroid, shows the two substrates oriented roughly thirty degrees apart, and part of the large flexible loop folded into a helix that swings critical PAPS-binding residues away from the cofactor pocket. That conformation is absent in the PAP-bound structure of the same enzyme. The interpretation is that at high substrate concentrations, the substrate can induce a loop configuration that’s incompatible with PAPS binding, locking the enzyme in a non-productive state before catalysis can occur. With structures in hand, the team turned to chemistry. They screened nine of the twelve human SULTs against a library of ninety compounds using a thermostability-based binding assay: proteins were heated from twenty-seven to eighty degrees Celsius while aggregation was monitored by scattered light. A ligand that binds stabilizes the protein, shifting its aggregation temperature upward. A shift greater than two degrees Celsius counted as a positive binding event. They ran this screen both with and without PAP present, because whether the cofactor is there turns out to matter enormously for what gets detected. Then, for eight enzymes, they ran a parallel high-performance liquid chromatography activity assay — tracking conversion of PAPS to PAP as a direct readout of sulfonation. The combination produced what the authors call "chemical fingerprints": each enzyme showed a characteristic pattern of what it binds and what it actually converts. Several findings stand out. Dopamine binding to SULT1A3 and one-naphthol binding to SULT1B1 were only detectable when PAP was present — consistent with the priming model, where the cofactor reorganizes the binding site into a configuration that can capture those substrates. Conversely, many alternative ligands bound only in the absence of PAP. Three compounds — AMP-PNP, pyridoxal five-prime-phosphate, and quercetin — bound nearly all human SULTs, flagging them as broad-spectrum inhibitors. SULT1B1 was potently inhibited by pentachlorophenol and the hydrazide compound DBHD, a finding confirmed in follow-up assays and structurally rationalized by the one point eight angstrom SULT1C2-PAP-pentachlorophenol complex. Two enzymes in particular — SULT1C3 and SULT4A1 — deserve their own moment, because they illustrate the limits of what sequence alone can tell you. SULT1C3 was an orphan: its substrates were unknown. The screening data gave it a function. It sulfonated alpha-zearalenol at four point one nanomoles per minute per milligram, two-ethylphenol at two point two nanomoles per minute per milligram, and also p-nitrophenol, one-naphthol, and lithocholic acid. It bound catecholamines but didn't convert them — binding without activity. That pattern points toward a role in steroid and phenolic metabolism. SULT4A1 is stranger still. It is the only human SULT that doesn’t bind PAP. Titrations of PAP up to ninety millimolar produced no detectable stabilization of SULT4A1 — while the other family members reached binding saturation around one hundred micromolar. Structural inspection explains why: the conserved tryptophan that normally stacks with the adenine ring of PAP is replaced by leucine in SULT4A1, and the PAP-binding loop is shorter and missing a conserved lysine, leaving critical residues out of register. Tests of six alternative sulfonate donors all failed to stabilize the protein. SULT4A1 likely has no significant sulfonation activity. Yet it binds epinephrine, norepinephrine, two-hydroxyestradiol, and thyroid hormone T4 — hinting at a regulatory role, particularly in the brain, that doesn't involve catalysis. The clearest demonstration that chemical behavior doesn’t map onto sequence comes from the clustering analysis. When the authors grouped the nine enzymes by their binding and activity profiles, the resulting trees contradicted the sequence-based family tree in telling ways. SULT1A1 and SULT1A3 are ninety-five percent identical at the sequence level — and yet they don’t cluster together by binding profile or by activity. Meanwhile, SULT1B1 and SULT1C2 cluster closely by both activity and binding despite being more distantly related in sequence. The local architecture of the substrate-binding pocket, reshaped by those flexible loops, is doing work that global sequence similarity can’t capture. What this all adds up to is a new toolkit for understanding how sulfonation goes wrong or right. Knowing which enzyme activates dietary carcinogens, which environmental toxins like pentachlorophenol can freeze hormone-metabolizing enzymes in non-productive conformations, and why individuals with different SULT variants respond differently to the same drug — these questions all hinge on the kind of family-wide, structure-plus-chemistry data that Allali-Hassani and colleagues assembled. Binding-site plasticity remains the central challenge going forward: the loops that make these enzymes interesting are also the loops that are hardest to capture in a single structure. Multiple ligand-bound snapshots per enzyme, alongside cellular assays, will be needed to fully translate these fingerprints into predictive biology. But the template is here — and it applies wherever sequence comparison alone has been leading researchers to the wrong conclusions. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.

They have the same architecture and the same chemistry — and yet radically different biological outcomes depending on which enzyme does the sulfonating. Before this work, five of the twelve had no published crystal structure, and no one had systematically screened the whole family against the same set of compounds. That gap is what the study closes. The structural work came first. The team solved crystal structures for five previously uncharacterized human SULTs, including SULT1C3 bound to the cofactor product PAP at a resolution of three point two angstroms, an apo SULT1C2 structure at a resolution of two angstroms, a ternary complex of SULT1C2 with PAP and the environmental toxin pentachlorophenol at a resolution of one point eight angstroms, and SULT4A1 at a resolution of two point two angstroms. Across the family, the global fold is conserved — but what varies are three flexible loop segments: a thirteen-residue loop, a shorter loop of four to ten residues, and a large loop of thirty-two to forty-six residues. In structures without any ligand, these loops are often disordered, appearing as gaps in the electron density map. In PAP-bound structures, they gain order — in some cases, folding into defined helices. One practical limit also emerged: the database sequence for SULT6B1 lacks the N-terminal region encoding a beta sheet critical to the fold, and the protein simply wouldn’t express in bacterial purification attempts.

The most important thing the structures revealed, though, wasn’t a static snapshot. It was a dynamic process — and it’s the conceptual heart of the paper. When PAP binds, a helix called alpha-14 to alpha-15 and the C-terminal segment of the largest flexible loop become ordered, positioning three absolutely conserved residues — threonine two hundred twenty-eight, arginine two hundred fifty-eight, and glycine two hundred sixty — that are required for PAPS binding. The cofactor, in other words, organizes the very loops that will later contact the substrate. Allali-Hassani and colleagues call this "priming." Think of it as a lock that subtly changes shape when one key is inserted, so only certain other keys will later fit productively. The enzyme can formally accept its substrate or cofactor in either order — the structural data include binary complexes with either ligand alone, ruling out a strict ordered mechanism. But cofactor binding biases the loop ensemble toward ordered, catalytically competent states. That distinction matters. The priming idea also explains one of the stranger behaviors in this enzyme family: substrate inhibition, where too much substrate actually shuts the enzyme down. Prior explanations focused on two substrate molecules crowding the active site simultaneously. The structures suggest a third mechanism.

A SULT2A1 structure containing two molecules of DHEA, which stands for dehydroepiandrosterone, a steroid, shows the two substrates oriented roughly thirty degrees apart, and part of the large flexible loop folded into a helix that swings critical PAPS-binding residues away from the cofactor pocket. That conformation is absent in the PAP-bound structure of the same enzyme. The interpretation is that at high substrate concentrations, the substrate can induce a loop configuration that’s incompatible with PAPS binding, locking the enzyme in a non-productive state before catalysis can occur. With structures in hand, the team turned to chemistry. They screened nine of the twelve human SULTs against a library of ninety compounds using a thermostability-based binding assay: proteins were heated from twenty-seven to eighty degrees Celsius while aggregation was monitored by scattered light. A ligand that binds stabilizes the protein, shifting its aggregation temperature upward. A shift greater than two degrees Celsius counted as a positive binding event. They ran this screen both with and without PAP present, because whether the cofactor is there turns out to matter enormously for what gets detected. Then, for eight enzymes, they ran a parallel high-performance liquid chromatography activity assay — tracking conversion of PAPS to PAP as a direct readout of sulfonation.

The combination produced what the authors call "chemical fingerprints": each enzyme showed a characteristic pattern of what it binds and what it actually converts. Several findings stand out. Dopamine binding to SULT1A3 and one-naphthol binding to SULT1B1 were only detectable when PAP was present — consistent with the priming model, where the cofactor reorganizes the binding site into a configuration that can capture those substrates. Conversely, many alternative ligands bound only in the absence of PAP. Three compounds — AMP-PNP, pyridoxal five-prime-phosphate, and quercetin — bound nearly all human SULTs, flagging them as broad-spectrum inhibitors. SULT1B1 was potently inhibited by pentachlorophenol and the hydrazide compound DBHD, a finding confirmed in follow-up assays and structurally rationalized by the one point eight angstrom SULT1C2-PAP-pentachlorophenol complex. Two enzymes in particular — SULT1C3 and SULT4A1 — deserve their own moment, because they illustrate the limits of what sequence alone can tell you. SULT1C3 was an orphan: its substrates were unknown. The screening data gave it a function. It sulfonated alpha-zearalenol at four point one nanomoles per minute per milligram, two-ethylphenol at two point two nanomoles per minute per milligram, and also p-nitrophenol, one-naphthol, and lithocholic acid. It bound catecholamines but didn't convert them — binding without activity. That pattern points toward a role in steroid and phenolic metabolism.

SULT4A1 is stranger still. It is the only human SULT that doesn’t bind PAP. Titrations of PAP up to ninety millimolar produced no detectable stabilization of SULT4A1 — while the other family members reached binding saturation around one hundred micromolar. Structural inspection explains why: the conserved tryptophan that normally stacks with the adenine ring of PAP is replaced by leucine in SULT4A1, and the PAP-binding loop is shorter and missing a conserved lysine, leaving critical residues out of register. Tests of six alternative sulfonate donors all failed to stabilize the protein. SULT4A1 likely has no significant sulfonation activity. Yet it binds epinephrine, norepinephrine, two-hydroxyestradiol, and thyroid hormone T4 — hinting at a regulatory role, particularly in the brain, that doesn't involve catalysis. The clearest demonstration that chemical behavior doesn’t map onto sequence comes from the clustering analysis. When the authors grouped the nine enzymes by their binding and activity profiles, the resulting trees contradicted the sequence-based family tree in telling ways. SULT1A1 and SULT1A3 are ninety-five percent identical at the sequence level — and yet they don’t cluster together by binding profile or by activity.

Meanwhile, SULT1B1 and SULT1C2 cluster closely by both activity and binding despite being more distantly related in sequence. The local architecture of the substrate-binding pocket, reshaped by those flexible loops, is doing work that global sequence similarity can’t capture. What this all adds up to is a new toolkit for understanding how sulfonation goes wrong or right. Knowing which enzyme activates dietary carcinogens, which environmental toxins like pentachlorophenol can freeze hormone-metabolizing enzymes in non-productive conformations, and why individuals with different SULT variants respond differently to the same drug — these questions all hinge on the kind of family-wide, structure-plus-chemistry data that Allali-Hassani and colleagues assembled. Binding-site plasticity remains the central challenge going forward: the loops that make these enzymes interesting are also the loops that are hardest to capture in a single structure. Multiple ligand-bound snapshots per enzyme, alongside cellular assays, will be needed to fully translate these fingerprints into predictive biology. But the template is here — and it applies wherever sequence comparison alone has been leading researchers to the wrong conclusions. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.

Related lectures