The External Aldimine Form of Serine Palmitoyltransferase

Sphingolipid biosynthesis begins with the condensation of l-serine and palmitoyl-CoA catalyzed by the PLP-dependent enzyme serine palmitoyltransferase (SPT). Mutations in human SPT cause hereditary sensory autonomic neuropathy type 1, a disease characterized by loss of feeling in extremities and severe pain. The human enzyme is a membrane-bound hetereodimer, and the most common mutations are located in the enzymatically incompetent monomer, suggesting a “dominant” or regulatory effect. The molecular basis of how these mutations perturb SPT activity is subtle and is not simply loss of activity. To further explore the structure and mechanism of SPT, we have studied the homodimeric bacterial enzyme from Sphingomonas paucimobilis. We have analyzed two mutants (N100Y and N100W) engineered to mimic the mutations seen in hereditary sensory autonomic neuropathy type 1 as well as a third mutant N100C designed to mimic the wild-type human SPT. The N100C mutant appears fully active, whereas both N100Y and N100W are significantly compromised. The structures of the holoenzymes reveal differences around the active site and in neighboring secondary structure that transmit across the dimeric interface in both N100Y and N100W. Comparison of the l-Ser external aldimine structures of both native and N100Y reveals significant differences that hinder the movement of a catalytically important Arg378 residue into the active site. Spectroscopic analysis confirms that both N100Y and N100W mutants subtly affect the chemistry of the PLP. Furthermore, the N100Y and R378A mutants appear less able to stabilize a quinonoid intermediate. These data provide the first experimental insight into how the most common disease-associated mutations of human SPT may lead to perturbation of enzyme activity.

Sphingolipids are ubiquitous constituents of eukaryotic cells, where they play important roles in signaling, differentiation, and apoptosis (1)(2)(3)(4). Defects in sphingolipid catabolism have been linked to several human diseases, such as hypertension, cancer, and disorders of the peripheral nervous system. The most common inherited peripheral neuropathy is hereditary sensory autonomic neuropathy type 1 (HSAN1). 6 The disease leads to progressive loss of sensation in extremities and is often associated with searing pain (5)(6)(7)(8). Genetic studies by two independent groups mapped the disease-associated mutations to the lcb1 (long chain base 1) gene on chromosome 9q22, which encodes the SPT1 subunit of serine palmitoyltransferase (SPT; EC 2.3.1.50) (9 -11). SPT catalyzes the first and rate-limiting step of the sphingolipid biosynthetic pathway in all organisms studied to date (12). The reaction is a pyridoxal 5Ј-phosphate (PLP)-dependent, decarboxylative, Claisen condensation of the amino acid L-serine and the long chain (C16) fatty acid palmitoyl-CoA, which produces the sphingolipid precursor, 3-ketodihydrosphingosine (KDS).
Analysis of a number of HSAN1 patients has revealed the four most common mutations to be C133W, C133Y, V144D, and G387A, with the cysteine mutations appearing to be the most prevalent in populations (10,25). The impact that these mutations have on SPT activity and sphingolipid metabolism has been the focus of attention by a number of groups. Most surprisingly, these disease-associated mutations occur in the SPT1 monomer, a protein that must be inactive, since it lacks the key lysine, histidine, and aspartate residues necessary to bind and stabilize the PLP cofactor. It is the SPT2 protein that contains the conserved, active site lysine residue that forms a Schiff base internal aldimine with the PLP. However, both subunits are essential to produce functionally active SPT heterodimer (26 -29), indicating that the inactive subunit has a crucial role in function. Dunn and co-workers (30) created several mutations in the yeast lcb1 and lcb2 genes, including those corresponding to the HSAN1 SPT1 mutations, and found that they dominantly inhibit SPT activity. A transgenic mouse model also revealed that the LCB1 C133W mutation led to mice with decreased tissue SPT activity and HSAN1 symptoms despite unaltered ceramide concentrations (31). A different study found that SPT activity was decreased in the tissues of HSAN1 patients, and in a Chinese hamster ovary model, hamster LCB1 C133Y and C133W mutations could not rescue cells lacking endogenous LCB1 (32). These combined studies revealed that the HSAN1 mutations act directly on the SPT enzyme but the pathological consequence on sphingolipid metabolism is unclear. Very recently, Dunn and colleagues discovered a small 80-amino acid protein (TSC3p) in yeast that stimulated yeast SPT activity, but its role is still unclear (33). Recently, a third eukaryotic subunit (SPT3) has been characterized and appears to be required for optimum SPT activity, although it is expressed most highly in placental tissue and human trophoblasts (34). It has been suggested that these three subunits could form a higher order SPT complex (35). The instability and hydrophobic nature of eukaryotic SPTs has made their isolation and therefore biophysical characterization particularly challenging (24, 36 -38). The dotted line shows that steps e and f may be bypassed by an alternative mechanism that does not proceed via a product quinonoid (see Ref. 21).
We have targeted homodimeric SPT from Sphingomonas paucimobilis EY2395 to provide molecular insight into the enzymatic properties of SPT (39,40). Recently, we reported the first high resolution x-ray crystal structure of the S. paucimobilis holo-SPT (41) and showed that the active site containing the PLP cofactor is at the dimer interface. We used this structure to model the human enzyme and map the human cysteine residue (Cys 133 ) of SPT1 onto Asn 100 of the bacterial SPT (41). We found that Asn 100 is proximal to the PLP binding site and lies at the dimer interface. To explore the effects of the HSAN1 mutations on human SPT activity and structure, we have studied mutants of the bacterial enzyme. We made the N100C mutant to better mimic native human SPT and N100W and N100Y to mimic the most common mutations in human SPT. We have characterized the mutations using kinetics, spectroscopy, and structural biology, including structures of the external L-Ser aldimine. These data provide molecular insights into the effects of HSAN1 mutations. Also, as part of these studies, we identify residues that play a role in the stabilization of intermediates in the SPT reaction and have been able to generate new insights into the substrate specificity and mechanism of the ␣-oxamine synthase family.
The recombinant plasmids were used to transform E. coli BL21 (DE3) competent cells, and selection was carried out on LB agar containing 30 g/ml kanamycin. For each mutant (and native), a single colony was used to inoculate 500 ml of 2YT broth (16 g/liter Bacto-tryptone, 10 g/liter Bacto-yeast extract, 5 g/liter sodium chloride (pH 7.5)), which was shaken at 250 rpm overnight at 37°C. The overnight culture was added to 4 liters of 2YT broth supplemented with kanamycin and grown at 37°C to an A 600 of 0.6. Protein expression was induced by the addition of isopropyl 1-thio-␤-D-galactopyranoside to give a final concentration of 0.1 mM, and growth was continued for 5 h at 30°C. Cells were harvested (Sorvall RC5B centrifuge) by centrifugation at 3500 rpm for 20 min at 4°C.
Purification of SPT WT, N100W, N100Y, N100C, R378A, and R378N-The proteins were purified essentially according to the previously described method, which is based polyhistidinetagged protein binding to Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) (41). After elution, fractions were pooled and desalted by dialysis against 20 mM potassium phosphate (pH 7.5), 150 mM NaCl, and 25 M PLP. A gel filtration step (Sephacryl S200 26/60 HR; GE Healthcare) was carried out in the same buffer. For enzymatic assays and UV-visible spectroscopy, this buffer is also used. For x-ray crystal trials, the protein was exchanged into 10 mM Tris (pH 7.5), 150 mM NaCl, and 25 M or 250 M PLP. For storage, the enzymes were transferred to the same buffer containing 20% glycerol (v/v) and stored at Ϫ80°C until use. Protein identity and integrity were confirmed by high pressure liquid chromatography electrospray mass spectrometry on a MicroMass Platform II quadrupole mass spectrometer equipped with an electrospray ion source. The experimentally determined masses of recombinant S. paucimobilis SPT WT and each of the mutants were within 0.1% of the theoretical mass, which includes the C-terminal LEHHHHHH fusion affinity tag with the enzyme lacking the N-terminal methionine (41).
Spectroscopic Measurements-All UV-visible spectra were recorded on a Cary 50 UV-visible spectrophotometer (Varian) and analyzed using Cary WinUV software (Varian). Prior to enzymatic assays, SPT was converted to the holo-form by dialysis against freshly prepared 20 mM potassium phosphate (pH 7.5) containing 150 mM NaCl and 25 M PLP for 1 h at 4°C. Excess PLP was removed by passing the protein through a PD-10 (Sephadex G-25M) desalting column (GE Healthcare) before concentration to 10 -20 mg/ml using a VivaSpin 30 kDa cut-off concentration filter. For UV-visible assays, the concentration of recombinant SPT was 10 M, and the spectrophotometer was blanked with 20 mM potassium phosphate (pH 7.5) containing 150 mM NaCl.
Determination of Dissociation Constants-Assays were carried out in 0.5-ml quartz cuvettes (1-cm path length) and typically contained 10 M SPT in 20 mM potassium phosphate (pH 7.5). Varying amounts of L-serine (0 -80 mM) were used in each assay. After the addition of substrate, the reactants were mixed and allowed to equilibrate for 15 min at 25°C. Small base-line changes were corrected using Sigma Plot software. Changes in absorbance at 425 nm (SPT WT, N100C, R378A, and R378N) and 420 nm (SPT N100W and N100Y) were plotted against L-serine concentrations, and data points were fitted to a hyperbolic saturation curve (Equation 1) using Sigma Plot software, where ⌬A obs represents the observed change in absorbance at 422 nm, and ⌬A max is the maximal absorbance change, [serine] is the L-serine concentration, and K d is the dissociation constant ( Table 1). The errors are asymptotic S.E. values of each parameter derived from the curve fitting procedure within Sigma Plot, and values were reproducible across a number of enzyme preparations. SPT WT, N100Y, and R378A Quinonoid Formation-Assays were carried out in 0.5-ml quartz cuvettes and typically con- This was then extracted with an equal volume of CHCl 3 /CH 3 OH (2:1, v/v). The sample was centrifuged at 13,000 rpm for 5 min, and the aqueous phase was discarded before the organic phase was allowed to evaporate overnight. The resulting lipid residue was resuspended in 15 l of CHCl 3 / CH 3 OH (2:1, v/v) and spotted onto a silica gel 60 TLC plate. Separation was carried out with a mobile phase of CHCl 3 / CH 3 OH/NH 4 OH (40:10:1, v/v/v), using a KDS reference standard (Matreya). The TLC was developed with a Storage Phos-phorImager (GE Healthcare) for 4 days at room temperature, and the phosphor screen was visualized using ImageJ software.
SPT activity was also measured using a continuous spectrophotometric assay by monitoring the release of CoASH from acyl-CoA substrates and reaction with 5,5Ј-dithiobis-2-nitrobenzoic acid (DTNB) (42). Assays were carried out in 0.5-ml cuvettes in a Varian Cary-50 UV-visible spectrophotometer. The enzyme was incubated with L-serine in a buffered solution containing DTNB, and the assay was started by the addition of the second substrate, palmitoyl-CoA. The CoASH thiol product was monitored by observation of the TNB Ϫ anion at 412 nm ⑀ max ϭ 14,150 M Ϫ1 cm Ϫ1 (43). Initial rates were measured at increasing concentrations of L-serine (0.12-50 mM) while maintaining palmitoyl-CoA in excess. A typical experiment to determine the K m value for L-serine contained 0.16 M SPT, 0.1-50 mM L-serine, 250 M palmitoyl-CoA, and 0.2 mM DTNB in 50 mM potassium phosphate buffer, pH 7.5. Kinetic constants (k cat and K m values; Table 1) were calculated from Michaelis-Menten plots using Sigma Plot. Kinetic constant values were calculated in a similar way for CoA substrates (decanoyl, lauroyl, myristoyl, palmitoyl, stearoyl, and arachidoyl) but by maintaining L-serine (100 mM) in excess ( Table 2). As a check, the k cat /K m parameters for CoA substrates were also determined by the method of complete condensation when [CoA] was Ͻ ϽK m . In this case, k cat /K m is calculated at a single substrate concentration as an observed rate constant (k obs ) by monitoring disappearance of the substrate. The data were fitted to the equation, A ϭ A lim ϫ (1 Ϫ exp(Ϫk obs ϫ t)), where A is the absorbance at 412 nm, A lim is the limiting absorbance value when the substrate has completely disappeared, k obs is the observed rate constant, and t is the time in seconds. The errors are asymptotic S.E. values of each parameter derived from the curve fitting procedure within Sigma Plot. The values obtained were reproducible over a number of enzyme preparations, and the errors can be used as a gauge of the fitted curve's accuracy and were generally less than 5% of the value of the parameter.
Structural Biology-The three mutant proteins were screened for suitable crystallization conditions at the Scottish Structural Proteomics Facility. The proteins were dialyzed in the presence of excess PLP to ensure complete reloading prior to crystallization. Conditions used to obtain crystals of the wildtype/PLP protein (41) were used as a starting point to formulate stochastic optimization screens using software developed in house. The screens were built on a Hamilton Microstar liquidhandling robot controlled by Rhombix system software (Thermo). Crystallization trials were set up in hanging drop plates (EasyXtal DG-CrystalSupport; Qiagen) using 2-3 l of the protein solution and 1 l of well solution in the hanging drop.
Crystals of the mutant SPTs were obtained in the form of yellow plates corresponding to the high resolution form of the holo wild-type SPT. The SPT N100W mutant was crystallized at 20 mg/ml in 10 mM Tris (pH 7.5), 150 mM NaCl, 25 M PLP using a well solution of 0.1 M MgCl 2 , 22% (w/v) PEG 3350, and 0.10 M Hepes (pH 6.5). The N100Y mutant was crystallized at 20 mg/ml in 10 mM Tris (pH 7.5), 150 mM NaCl, 250 M PLP, using a well solution of 0.15 M MgCl 2 , 22% (w/v) PEG 3350, and 0.10 M Hepes (pH 6.5). Finally, the N100C mutant was crystallized at 20 mg/ml in 10 mM Tris (pH 7.5), 150 mM NaCl, 250 M PLP, using a well solution of 0.1 M MgCl 2 , 27% (w/v) PEG 3350, and 0.10 M Hepes (pH 6.5).
The solutions used for equilibrating harvested crystals varied slightly according to the crystallization conditions and consisted of 22-29% PEG 3350 (w/v), 0.1 M HEPES (pH 6.5), 0.08 -0.12 M MgCl 2 . All crystals used for data collection were soaked for a few minutes in the respective equilibrating solution plus 1-2 mM PLP. Additionally, the external aldimine form of the SPT wild-type was produced by soaking the crystal in the equilibrating solution plus 5 mM L-serine for 10 min, and that of the N100Y mutant was produced by soaking the crystal in the equilibrating solution plus 50 mM L-serine for 15 min at room temperature. The crystals were mounted in a cryo-loop (Molecular Dimensions) and cryo-protected in solutions varying slightly for each crystal and containing the equilibrating solutions plus 15-20% PEG 400 (v/v). The crystals were then frozen by plunging them into liquid nitro-gen and carried in a dry cryogenic Dewar to the European Synchrotron Radiation Facility (Grenoble, France) for data collection. The data sets were collected at 100 K to varying resolutions using three different beam lines (supplemental Table 1). The data for the wild-type serine external aldimine were processed with Mosflm and scaled with Scala from the CCP4 suite of programs (44). The data for the N100C, N100Y, N100W, and N100Y serine external aldimine structures were processed using XDS and scaled using XSCALE (available on the World Wide Web) (45).
Analysis of the density revealed that the PLP was covalently bound to Lys 265 in every mutant, confirming that we had obtained the holo-form of the proteins. All mutant models were refined using Refmac5 (46) and manually adjusted, including the addition of water molecules with WinCoot (47). Data and structures have been deposited in the Protein Data Bank. Density was clearly visible to the N100W mutation, but we could not satisfactorily locate the tryptophan side chain; there is disordered electron density, which indicates multiple conformations.

RESULTS
Spectroscopic Properties of SPT N100 Mutants-The three mutants were obtained in milligram quantities as the internal aldmine, holo-form, confirming their capacity to bind the PLP cofactor (Fig. 2). The SPT N100C mutant has absorbance maxima at 335 and 425 nm, akin to the wild-type SPT values of 340 and 425 nm (Fig. 2, A and B). In both native and N100C, the enolimine peak at 335 nm is dominant, indicating that the PLP cofactor is in a very similar chemical environment in both proteins. In contrast, the UV-visible spectra of the SPT N100W and N100Y mutants have maximum absorbance values at 340 and 415 nm (Fig. 2, C and D). We noted that the ketoenamine peak (415 nm) had not only blue-shifted; it had also become the dominant form in the spectrum in these enzymes. These differences suggest a common perturbation in the PLP binding site of the mutants compared with the wild type (Fig. 2, C and D, solid  lines). The ability of the mutants to form the external aldimine form by binding L-serine was measured by the addition of increasing concentrations of the amino acid (ranging between 0 and 100 mM). This addition resulted in an increase of the absorbance at 425 nm for the wild-type and the N100C enzyme. Titration of L-serine into the SPT N100W and N100Y mutants led to a small shift of a few nm to a longer wavelength. Although all mutants form the external aldimine, the N100W and N100Y mutants have clear differences (Fig. 2, A-D). By measuring the change in absorbance at ketoenamine-specific wavelengths that are specific to the external aldimine, the apparent serine dissociation constants (K d Ser ) have been determined for the wild type SPT and each of the mutants (  (Fig. 3A), using the method previously described (48). The N100C mutant showed an activity similar to the wild-type SPT; however, both the N100W and N100Y mutants showed a large decrease in activity (ϳ84%) compared with the wild-type enzyme (with saturating substrate concentrations) (41). The R378N and R378A produced 50 and 40% KDS, respectively, in comparison with the wild-type enzyme (Fig. 3B).
Kinetic Analyses of Wild-type, HSAN1 Mimics, and Arg 378 Mutants-The raw data for wild-type SPT are shown in Fig. 3C (12,49). Similar analysis was also carried out on the SPT R378A and R378N mutants. These bound PLP in a manner similar to wild-type with absorbance maxima at 335 and 425 nm (data not shown). The values obtained were similar to that of the   R378N, respectively). Using the coupled spectrophotometric assay, we measured 60-and 40-fold decreases in k cat /K m Ser for each mutant (R378A and R378N, respectively; Table 1).
SPT WT, N100Y, and R378A Quinonoid Formation-The ability of the wild-type enzyme and both the N100Y and R378A mutants to form a quinonoid intermediate was tested by the addition of a thioether analogue of palmitoyl-CoA reported by Ikushiro et al. (50). The addition of the analogue to the L-serine external aldimine form of the wild-type enzyme led to the appearance of a clear peak at 495 nm due to the quinonoid (increase in absorbance ϳ0.06), in agreement with that found previously (Fig. 4A). In contrast, the SPT N100Y mutant, under the same conditions, produced a small, broad shoulder 490 -500 nm (increase in absorbance ϳ0.01 absorbance units; Fig.  4B). We could not observe the formation of a quinonoid species at all in the R378A mutant (Fig. 4C).
The Structure of the Wild-type SPT⅐L-Ser External Aldimine Complex-The dimeric form of the L-Ser external aldimine SPT complex is shown in Fig. 5A. Of course, the major difference between the internal (holo-) and external aldimine forms is that in the internal adimine, the Schiff base is between Lys 265 of the protein and the PLP co-factor, whereas in the L-Ser external aldimine, this C-N bond has broken. The PLP⅐L-ser external aldmine is clearly observed in the electron density (Fig. 5B). An overlay of the structures of the internal (41) and external aldimine shows an r.m.s. deviation of 0.3 Å and reveals that most of the interactions between the enzyme and the PLP cofactor are unchanged. In both structures, the PLP hydrogen-bonds to the side chains of Asp 231 , His 234 , Thr 262 , and Ser 264 , the main chain of Gly 134 and Thr 135 , and -stacks with His 159 . His 159 is con-  (50), is clearly observed in the wild-type enzyme, but only a broad shoulder is produced in the N100Y mutant, and no peak is observed for the R378A mutant. AU, absorbance units. served throughout all of the members of the ␣-oxoamine synthase family (41). As a result of external aldimine formation, the PLP has shifted, essentially rotating around the phosphate, with the Schiff base nitrogen moving over 1.5 Å. Lys 265 adopts a different conformation as a result of breaking its link to PLP and now makes a hydrogen bond with the hydroxyl of the L-ser component of the external aldimine (Fig. 6A). Accompanying the change in Lys 265 conformation, Tyr 73 substantially alters its position. The carboxylate group of L-Ser hydrogen-bonds to the conserved His 159 and makes a salt link with Arg 378 , which has "swung" into the active site. In the native structure, this residue is at the end of a ␤-strand and hydrogen-bonded to the side chain of Gln 357 . In order to make the salt contact, the ␤-strand has been disrupted, and the C␣ of Arg 378 has been moved over 5 Å. The loop following Arg 378 (the "PPATP" loop) has significantly altered its conformation and position (e.g. the C␥ atom of Pro 379 has moved over 10 Å). The reordering of the structure around Arg 378 seems to be linked to other movements at Phe 47 (1.2-Å C␣ shift). These residues are contained within the one monomer. The phosphate of PLP binds to the side chain of Thr 294 from one subunit and to the main chain of Ala 294 from the other subunit. The hydroxyl of L-Ser makes a polar contact to the main chain of Ala 294 . In the other monomer, the side chain Met 104 adjusts its conformation, and there is a slight rigid body shift of Leu 105 to Asn 106 . The side chain of Asn 100 from the other subunit has altered its position, but the hydrogen bond to Lys 265 is maintained. Structural Biology of the Holo-forms of the HSAN1 Mutant Mimics-The wild-type SPT and N100C structures superimpose with an r.m.s. deviation of 0.2 Å along the 396 C␣ residues and are largely identical. Similarly, the N100W and N100Y mutant structures superimpose with 0.1 Å r.m.s. deviation for their C␣ positions, mirroring their similar spectroscopic properties. However, both N100W and N100Y superimpose with the native structure (and N100C) with an r.m.s. deviation of 0.6 Å. For ease of discussion, only the differences between N100Y and the native enzyme are reported. The N100W structure is essentially identical to N100Y, whereas N100C is essentially identical to native. Reflecting the large deviation in C␣ positions, there are a number of small shifts in secondary structure elements throughout the structure when comparing the N100Y mutant with the native SPT. These shifts are largely conserved, regardless of which structures one compares (N100W versus native, N100W versus N100C, N100Y versus N100C, and N100Y versus native), suggesting that they are real and not simply a crystallographic artifact. The most obviously visible change is the N-terminal helix (Asp 23 -Gly 42 ), which has undergone an approximately 1.7-Å shift. This helix is an important part of the dimeric interface of SPT (41). At the site of mutation itself, the tyrosine side chain has "flipped" out the pocket occupied by Asn 100 (Fig. 6B). At the C␣ level, this is a displacement of 3.4 Å and is accomplished by an almost 180°rotation of the Ramachandran angle at Thr 99 . As a result, the hydrogen bonds between the side chain of Asn 100 with the key Lys 265 from the other monomer have been broken. Instead, the Tyr 100 now occupies a new pocket at the dimer interface formed by hydrophobic residues from both monomers (Phe 109 from the same monomer; Met 51 , Thr 72 , and Ile 69 from the other monomer) (Fig. 6B). These hydrophobic residues have adjusted their position to accommodate the bulky tyrosine side chain. The most pronounced of these changes appear to accompany the movement of Phe 109 , which connects to an ␣-helix. Residues of the first turn of the helix, His 110 , Asp 111 , His 112 , Met 113 , and Glu 114 , all show disturbed side chain and main chain positions relative to the native structure. The main chain at Leu 105 , Asn 106 , and Gly 107 sits opposite and close to Asn 100 in the native structure. The movement of the C␣ atom at position 100 by 3.4 Å causes Leu 105 -Gly 107 to undergo large positional movements (over 4 Å for the C␣ positions of Gly 107 ) and changes in their Ramachandran angles.
Structural Biology of Wild Type and N100Y L-Ser External Aldimines-The r.m.s. deviation between the two external aldimines, native SPT, and N100Y mutant at 0.4 Å is less than the value for the internal aldimine structures (0.6 Å). Thus, it appears that formation of the external aldimine drives N100Y toward a more "native" conformation. This can be seen most clearly in the loop between Tyr 100 and Gly 107 , which is more similar to the native structure in the external aldimine complex. The most striking difference between the two external aldimine structures is that in the N100Y structure, the "PPATP loop" remains in its internal aldimine conformation, and consequently Arg 378 does not enter the active site and does not make salt links with the carboxylate of L-ser (Fig. 6C). Rather, in the external aldimine of N100Y, Arg 378 is hydrogen-bonded to Gln 357 , an interaction that is absent in the N100Y internal aldimine due to a small shift in structure.

DISCUSSION
Mechanistic Implications-The seminal hypothesis, forwarded by Dunathan (51) over 40 years ago, to explain how PLP-dependent enzymes can catalyze numerous chemical reactions is still valid. The key steps involve the deprotonation of the amino acid⅐PLP external aldmine to form the quinonoid species, which condenses with the incoming acyl-CoA thioester (Fig. 1). This is thought to generate a ␤-keto acid intermediate that then decarboxylates to give a product aldimine (23). The first intermediate in the catalytic cycle of SPT is the external aldimine formed between L-ser and PLP. Deprotonation of the C␣ of this external aldimine intermediate is the crucial next step. In studies with ALAS, this deprotonation has been observed to be accelerated by over 250,000-fold by the binding of the incoming thioester substrate (14,52). Studies on AONS, which uses L-alanine and pimeloyl-CoA, suggest that this enzyme undergoes conformational change during intermediate formation. Incubation of the enzyme with palmitoyl-CoA preequilibrated with L-alanine gave a 30-ms lag before quinonoid formation at 486 nm was observed with a rate of 45 s Ϫ1 (14). Recently, a thioether CoA substrate analogue (S-(2-oxoheptadecyl)-CoA) was shown to increase the rate of proton abstraction from the serine external aldimine of SPT by around 100fold (50).
Our new structure provides insight into the residues that control the L-Ser external aldimine conformation (Fig. 7, A-C). How was this intermediate trapped in the crystal? The high resolution data confirm that the L-serine⅐PLP external aldimine is sp 3 -hybridized at the C␣ (the angles are 111.5, 106.4, and 109.6°, which are close to the theoretical values for a tetrahedral carbon). In the external aldimine structure, the Nz atom of Lys 265 contacts (3.4 Å) the C␣ proton of the L-Ser. This double role for the Schiff base forming lysine was suggested by studies of members of the ␣-oxoamine synthase family (14,17,18,20), and SPT has now been shown to use the same feature. The angles between each functional group and the N-imine (which is co-planar with the PLP ring) were measured as follows: H-C␣-N, 107.7°; CO 2 -C␣-N, 108.4°; CH 2 -C␣-N, 113.3°. None of these have the desired perpendicular orientation to the plane, indicating that in the crystal structure, the external aldimine is not in the optimal "Dunathan conformation" suitable for deprotonation.
If, as seems likely, palmitoyl-CoA binding leads to rapid deprotonation, we suggest that it does so by triggering the rotation of the external aldimine about the imine N-C␣ bond. A small turn of ϳ15-20°would bring the C␣ hydrogen into the required perpendicular geometry (Fig. 7B). Of course, this would cause the simultaneous movement of the -CO 2 Ϫ group and the -CH 2 OH side chain. Ikushiro proposed a role for the conserved residue Arg 390 in the SPT mechanism (50). The structural data show that Arg 390 is over 6.5 Å distant from the carboxylate; rotating around the C␣-N bond to reach the likely quinonoid conformation shortens this distance to 4.6 Å (Fig. 7, C and D). An additional rearrangement, promoted by palmitoyl-CoA binding, could allow Arg 390 to interact with the carboxylate, as seen in the KBL external aldimine. We note that Arg 390 plays an important role in stabilizing the conformation of Tyr 73 , which forms the active site. The rotation to form the quinonoid would disrupt the salt link to Arg 378 in the absence of any additional structural adjustment. The interplay of Arg 390 and Arg 378 in interacting with the carboxylate group may be an important component in controlling the formation and deprotonation of the external aldimine. Together, they could act to pick up, ferry, then anchor the carboxylic acid group. This functional group is ultimately lost from the ␤-keto acid as CO 2 , and future work will be required to understand how this important B, a stereo image of an overlay of the external aldimine forms of the wild-type and N100Y. The wild-type structure has been rendered with monomer A colored with a blue backbone and yellow side chains, and monomer B is shown with a purple backbone and green side chains. In the N100Y mutant, the monomer A backbone and side chains are colored gold, and monomer B is colored with a red backbone and salmon pink side chains. The N100 side chain from monomer B of the wildtype SPT (green stick, asterisk) interacts with residues from monomer A. However, in the SPT N100Y mutant, the Tyr 100 side chain has flipped into a pocket formed by displacement of Phe 109 of the same monomer B, residue Asn 106 from monomer B has also undergone a large shift, and residue Asn 375 from monomer A has also changed. The backbone containing residue Tyr 100 has also undergone a large conformational change. C, a stereo image of an overlay of the external aldimine forms of the wild-type and N100Y showing the impact of mutation on Arg 378 and the PPATP loop. The coloring is the same as in B. In wild-type SPT, residue Arg 378 (side chain yellow on the blue backbone) is in the swung in position and interacts with the L-Ser carboxylic acid. In the SPT N100Y mutant, the "PPATP loop" and the Arg 378 residue (backbone and side chain colored gold) are in the "swung out" conformation observed in the wild-type internal aldimine form. The interaction between Arg 378 and Gln 357 (both in gold) and the different positions of the Asn 100 and Tyr 100 residues are shown. step is controlled. Ikushiro et al. (50) also proposed a role for His 159 acting as a hydrogen donor to accommodate the developing negative charge on the CϭO of the thioester. The structure shows that the N⑀ atom of His 159 hydrogen-bonds to the L-serine CO 2 Ϫ group as well as maintaining a -stacking interaction with the PLP ring. This is consistent with the reported SPT H159F mutant, which bound L-serine weakly (K d ϭ 11.1 mM) and was incapable of KDS formation. Whether the residue also plays a role in the chemistry of the thioester activation remains an open question.
That Arg 378 binds to the external aldimine was unsuspected, since it required a very large conformational change from the holoenzyme structure. Mutation of this residue (R378A and R378N) reduces enzyme efficiency by over a factor of 10 (Table  1). Thus, we suggest that this residue plays an important, but not essential, role possibly influencing the stabilization of an intermediate after both substrates have bound. Weight is given to this hypothesis by the fact that R378A cannot form a quinonoid in the presence of the thioether analogue (Fig. 4C). The nonessential nature of the residue is consistent with its lack of sequence conservation in all SPT isoforms, and an equivalent residue in the human SPT1/SPT2 heterodimer awaits identification. That there may be such a residue comes from analysis of enzymes in the ␣-oxoamine synthase family. In the two glycine-specific members, KBL and ALAS, residues Asn 50 and Asn 54 , respectively, make direct contact with the -CO 2 Ϫ group (17, 18) of the amino acid. In AONS, Webster et al. (14) used the AONS-AON product external aldimine structure to predict that the L-alanine carboxylate would coordinate with residue Asn 47 . Sequence homology lines up residue Asn 74 from S. paucimobilis SPT with the three Asn residues from AONS, ALAS, and KBL; however, analysis of the structure reveals that this residue in not involved in the active site. Thus, there is reason to suspect that some residue is required to coordinate to the carboxylic group of the external aldimine; however, given that the function is not essential, the conservation of the residue is not absolute.
The organization of the active sites of AONS, KBL, and ALAS is different compared with the bacterial SPT; the constellation of residues around the PLP cofactor is conserved, but there are subtle differences. SPT favors L-Ser as the substrate, and in the SPT external aldimine structure, the side chain hydroxyl makes a number of hydrogen bonds, which could explain this preference. Interestingly, although SPT can form an external aldimine with L-Thr (K d ϭ 3.8 mM) (40), it cannot be deprotonated to the quinonoid in our hands (data not shown). Looking at the structure of the aldimine, the additional methyl group of L-Thr would appear to clash with Ser 102 and Arg 378 in the proposed quinonoid model. In the ALAS structure, Thr 83 would seem to select against any substituent at the C␣ position, explaining the preference for Gly of this enzyme.
Mimics of the HSAN1 Disease Causing Mutations-The two most common mutations (C133W and C133Y) associated with the human disease HSAN1 have intrigued a number of researchers as to how they impact on sphingolipid metabolism and regulation. The fact that these mutations are in the inactive subunit, LCB1 (SPT1), has added to their scientific interest. Tissue from HSAN1 patients contains relatively normal levels of total sphingolipids (10,31,32). However, there are differing reports as to the intrinsic SPT activity of cells from HSAN1 individuals, and questions remain as to the precise consequence of the HSAN1 mutants on SPT turnover, stability, and regulation. Gable et al. (30) produced mutants of the yeast lcb1 gene (C180Y and C180W) that mimic the HSAN1 substitutions (C133W and C133Y). Interestingly, these mutations dominantly inactivated the SPT activity of the yeast LCB1⅐LCB2 heterodimeric complex. This study proposed that the LCB1 HSAN1 mutations perturbed the active site of the LCB1⅐LCB2 dimer. McCampbell et al. (31) expressed the human LCB1 C133W mutant in a transgenic mouse and found that the SPT activity was decreased in various tissues and mice developed a number HSAN1 symptoms. However, the levels of total ceramide in these mice were unchanged, as was also observed in human HSAN1 individuals. This indicates that to some extent, SPT activity is preserved, and the mutants have a more subtle effect than simple loss of activity. It would also indicate that there is some regulation of sphingolipid biosynthesis after KDS formation by SPT.
We had previously mapped the human Cys 133 residue to Asn 100 on the bacterial SPT (41). We first engineered an N100C mutation as a better mimic for the human enzyme. This mutant enzyme behaves within the sensitivity of our UV-visible spectroscopic, kinetic, and high resolution structural analyses as native. This gave us confidence that the other mutations at Asn 100 in the bacterial enzyme should provide a meaningful insight into human SPT.
N100W and N100Y mutants displayed marked spectroscopic, enzymatic, and structural differences when compared with wild-type SPT. Although both mutants are severely compromised in their activity, importantly, they do retain the ability to make the desired KDS product. It was obvious from the distinctive UV-visible properties of PLP-containing holo-forms of the enzymes that the mutants had altered the cofactor's chemical environment (Fig. 2, A-D). This change in the chemical environment apparently disturbs catalysis. After formation of an external aldimine, deprotonation gives a quinonoid intermediate (Fig. 1), which has been detected for both AONS and ALAS enzymes (14,52). In the native SPT enzyme, the quinonoid form can be detected by adding the palmitoyl-CoA thioether analog. However, in the N100Y mutant, the signal for the quinonoid is essentially absent, indicating that the stabilization of this intermediate is significantly decreased (Fig. 4, A  and B). In the wild-type holo-SPT (and N100C), the side chain of one monomer makes a direct contact with the amide backbone of the conserved PLP Schiff base Lys 265 of the other monomer. This hydrogen bond is lost by the disruption introduced by the mutation. We suggest that this would increase the mobility of the internal aldimine. The most striking observation is that the N100Y and N100W mutations cause, because of the key location of Asn 100 at the dimer interface, complex structural changes to ripple across the dimer interface (Fig. 6B). A striking illustration of this can be seen in the external aldimine of N100Y, particularly the role of Arg 378 . In the native enzyme, this residue plays an important role in binding to the carboxylate of the external aldimine, but in the N100Y mutant, Arg 378 remains uninvolved with the intermediate (Fig. 6C). Although it is difficult to identify a clear cut interaction that disfavors Arg 378 from entering the active site, this residue is in contact with Leu 105 from the other subunit. Leu 105 is one of the residues directly affected by the mutation at Asn 100 . The change in Leu 105 position introduced by the mutation would increase the van der Waals contact between Arg 378 (in the "swung in" state) and Leu 105 in the mutant. It is this ability for structural changes to "reach across the interface" that we propose underlies the C133W and C133Y mutation effects in HSAN1. The perturbations in the neighboring structure give rise to the changed chemical environment of both the internal and external aldimine, altering the enzyme function.
Combined with previous studies on AONS, ALAS, and KBL, the SPT data presented here shed more light on the catalytic mechanism employed by members of the ␣-oxoamine synthase family. The eukaryotic SPT heterodimeric isozymes appear to be more complex than their bacterial homologs, since a third, tissue-specific SPT subunit has been identified recently (34). Furthermore, an abnormal, "dead end" sphingolipid product, 1-deoxysphinganine (also known as the natural product "spisulosine"), formed by SPT using L-Ala as a substrate, has been identified in mammalian cells (53,54). Whether the SPT HSAN1 mutations impact directly on substrate and product specificity requires further detailed investigation. In the absence of large amounts of homogenous eukaryotic complexes, the bacterial enzymes provide some insight into how the wild-type and mutant forms behave and aid in the design of isolation strategies for the more complex SPTs (55). Studies aimed at the capture of complexes of various SPTs with a range of substrates, products, and inhibitors are under way.