Importance of ALDH1A enzymes in determining human testicular retinoic acid concentrations.

Retinoic acid (RA), the active metabolite of vitamin A, is required for spermatogenesis and many other biological processes. RA formation requires irreversible oxidation of retinal to RA by aldehyde dehydrogenase enzymes of the 1A family (ALDH1A). While ALDH1A1, ALDH1A2, and ALDH1A3 all form RA, the expression pattern and relative contribution of these enzymes to RA formation in the testis is unknown. In this study, novel methods to measure ALDH1A protein levels and intrinsic RA formation were used to accurately predict RA formation velocities in individual human testis samples and an association between RA formation and intratesticular RA concentrations was observed. The distinct localization of ALDH1A in the testis suggests a specific role for each enzyme in controlling RA formation. ALDH1A1 was found in Sertoli cells, while only ALDH1A2 was found in spermatogonia, spermatids, and spermatocytes. In the absence of cellular retinol binding protein (CRBP)1, ALDH1A1 was predicted to be the main contributor to intratesticular RA formation, but when CRBP1 was present, ALDH1A2 was predicted to be equally important in RA formation as ALDH1A1. This study provides a comprehensive novel methodology to evaluate RA homeostasis in human tissues and provides insight to how the individual ALDH1A enzymes mediate RA concentrations in specific cell types.

concentrations in paired samples from the same donors. In addition, the effect of cellular retinol binding protein (CRBP) on at RA synthesis in the testis was evaluated. These novel methods allowed the prediction of the relative contribution of each of the ALDH1A enzymes to overall intratesticular RA formation. The validity of the novel methods was confi rmed via specifi c inhibition studies. The importance of individual ALDH1A enzymes to at RA formation in specifi c cell types in the testis was evaluated via tissue localization studies of ALDH1A enzymes. The approach used can also be applied to measuring at RA formation in any human tissue biopsy by combining quantitative proteomics and sensitive kinetic analyses.

Human samples used in the study
Samples of testicular tissue used to characterize ALDH1A expression and activity were obtained from 18 anonymous donors at the Center of Reproductive Medicine and Andrology in Munster, Germany. Fifteen of the testes were from men who underwent orchiectomy as an androgen depletion therapy for the treatment of prostate carcinoma and all samples were anonymized following informed consent. All of the men had qualitatively normal spermatogenesis based on histological analysis. The men were 62-87 years old (mean, 70.7 years). In addition, three samples from individuals undergoing male to female sex reassignment surgery after at least two years of estrogen treatment were analyzed. These samples were also anonymized following informed consent and the age range of this group was 49-75 years (mean, 58.3 years). Due to the estrogen treatment, spermatogenesis was severely reduced on histological analysis of the tissue from these individuals. The tissues were collected into icecold Leibovitz medium at the surgery site, protected from light, and taken to the laboratory on ice. The testes were decapsulated and fragmented by scalpels and scissors. The fragments were snap-frozen in liquid nitrogen and stored at Ϫ 80°C until RA measurements or preparation of S10 fractions.

Quantifi cation of ALDH1A1, ALDH1A2, and ALDH1A3 in human testes using LC-MS/MS
Tissue biopsies (33-102 mg) from 18 donors were homogenized on ice using a drill-powered 2 ml Potter-Elvehjem glass homogenizer (Kimbel Glass, Vineland, NJ). Samples were homogenized in 3× volume tissue homogenization buffer (50 mM potassium phosphate and 250 mM sucrose) with an EDTA-free protease inhibitor cocktail (Roche, San Francisco, CA). The samples were transferred to a 1.5 ml microcentrifuge tube, the homogenizer washed with 2× volume of homogenization buffer, and the wash added to the homogenate. The homogenate was centrifuged at 10,000 g for 25 min at 4°C to pellet large organelles and cell membranes. The resulting supernatant (S10 fraction) that contains microsomes and cytosolic proteins was collected, aliquoted, fl ash-frozen in liquid nitrogen, and stored at Ϫ 80°C. Total protein concentration in each S10 fraction was measured using a BCA assay (Thermo Fisher, Waltham, MA).
ALDH1A expression in human testicular tissue was measured by peptide quantifi cation using LC-MS/MS. Peptides were chosen based on in silico analysis of their selectivity and analytical performance ( 34 ). The FASTA fi les for each ALDH1A sequence were downloaded from UniProt with the accession numbers P00352, O94788-1, and P47895 . The human ALDH1A enzymes In humans, ALDH1A1 mRNA has been detected in the liver, kidney, testis, brain, lung, red blood cells, and lens of the eye (17)(18)(19). In contrast, ALDH1A2 mRNA is found predominantly in the testis, uterus, and skeletal muscle, while ALDH1A3 mRNA is localized in the prostate, trachea, intestine, and testis ( 18 ). These mRNA expression patterns strongly suggest that all three ALDH1A enzymes contribute to at RA synthesis during postnatal life. In addition, ALDH1A1 expression has been proposed as a cancer stem cell marker and associated with poor prognosis in certain types of cancers, including clear cell renal, esophageal squamous cell, and papillary thyroid carcinomas (20)(21)(22). Conversely, ALDH1A2 has been identifi ed as a candidate tumor suppressor in a transgenic animal model of prostate cancer, and gene variants of ALDH1A2 have been found to be associated with osteoarthritis and increased newborn kidney size (23)(24)(25)(26). Taken together these studies point to tissue-specifi c central roles that ALDH1A enzymes play in human physiology, and the critical importance for obtaining information concerning the expression and activity of these enzymes in human tissues.
One of the major challenges in evaluating vitamin A and at RA homeostasis and the importance of ALDH1A enzymes in these processes has been the fact that serum concentrations do not readily refl ect tissue retinoid concentrations, as at RA is believed to be synthesized in the target tissues ( 27 ). Because of this, measurements of tissue at RA concentrations, and expression and activity of at RAsynthesizing enzymes in small tissue fractions, such as biopsies from humans, is critically important. To date, methods have not been readily available to determine the specifi c activity and expression of ALDH1A enzymes in human tissues or for demonstrating that ALDH1A activity infl uences tissue at RA concentrations and net at RA formation velocity. As such, it has not been possible to predict the pharmacological outcomes within a specifi c target tissue or globally in the whole body in response to specifi c ALDH1A enzyme inhibition or induction.
An excellent model organ for the tissue-specifi c synthesis of at RA is the testis ( 28 ), as at RA synthesis within the testis is required for male fertility ( 29,30 ). For example, there is no initiation of spermatogenesis in mice which lack ALDH1A1, ALDH1A2, and ALDH1A3 in the Sertoli cells ( 30 ). Similarly, an inhibitor of the ALDH1A enzymes, WIN 18,446, has been shown to affect the ability of multiple mammalian species, including humans, to produce sperm (31)(32)(33), likely due to decreased at RA concentrations in the testes ( 33 ). However, the specifi c role of the individual ALDH1A enzymes in at RA formation in the testis has not been evaluated. Hence, the human testis was chosen as a model tissue to establish how at RA synthesis is orchestrated by the three ALDH1A enzymes in a tissue-and cell type-specifi c manner to regulate at RA concentrations in the organ. A novel LC-MS/MS peptide quantifi cation method was developed that allows accurate quantifi cation of the expression levels of ALDH1As in small biopsy scale samples. This method was coupled with measurements of ALDH1A activity, the formation of at RA in tissue samples from 18 human donors, and measurements of at RA room temperature, 4 l of iodoacetamide (200 mM) was added and the sample was incubated at room temperature in the dark for 20 min. Trypsin was added at a 1:25 trypsin:protein ratio, and the sample was digested for 24 h at 37°C. The incubation was quenched by addition of 20 l chilled acetonitrile with 8% trifl uoroacetic acid containing the non-lagging end heavy labeled ALDH1A1 and ALDH1A2 peptide internal standards. Samples were centrifuged at 3,000 g for 25 min at 4°C and analyzed by LC-MS/MS.
The peptides were quantifi ed by MS using an AB Sciex 5500 qTrap Q-LIT mass spectrometer (AB Sciex, Foster City, CA) equipped with an Agilent 1290 ultra (U)HPLC (Agilent, Santa Clara, CA). The tryptic peptides were separated using an Aeris peptide XB-C18 column (50 × 2.1 mm) with 1.7 m particle size at 40°C and a SecurityGuard Ultra UHPLC C18-peptide cartridge (Phenomenex, Torrance, CA). The eluting solvents were: A, water + 0.1% formic acid, and B, acetonitrile (ACN) + 0.1% formic acid. For chromatographic separation, the following 18 min linear gradient with a 400 l/min fl ow rate was used: 0 → 3.5 min 3% B, then increased by 12 For quantifi cation of ALDH1As, purifi ed ALDH1A1, ALDH1A2, and ALDH1A3 were used as standards. A nine point standard curve was generated with values of 0.018-7.2 pmol for ALDH1A1, 0.005-1.35 pmol for ALDH1A2, and 0.002-0.72 pmol ALDH1A3. The amount of enzyme detected in each sample was normalized to total S10 protein (0.08 mg).

Mass spectrometric quantifi cation of at RA, 13-cis RA, and retinal
The concentrations of at RA, 13-cis RA, and retinal were measured using an AB Sciex 5500 qTrap Q-LIT mass spectrometer equipped with an Agilent 1290 UHPLC as previously described ( 35 ) with several minor modifi cations. Briefl y, at RA and 13-cis RA were monitored using positive ion APCI and MS transitions of m/z 301 → 205 and m/z 301 → 123 with the 205 fragment used for quantifi cation. For quantifi cation, at RA and 13-cis RA peak areas were normalized to the at RA-d 5 internal standard area. The concentration of retinal was measured only in the in vitro samples using positive ion APCI and the MS transition m/z 285 → 161 with declustering potential of 66.0, collision energy of 13.0, and collision cell exit potential of 4.0. The at -retinal-d 5 was used as an were fi rst trypsin digested in silico, and peptides with a predicted m/z greater than 1,100 were excluded due to the m/z range limitations of the mass spectrometer. Next, the predicted peptides were screened against a trypsin-digested human proteome to ensure selectivity. Additionally, sites predicted or reported to have posttranslational modifi cations or mutations were avoided. Purifi ed recombinant ALDH1A enzymes were trypsin digested in vitro to determine the coverage of ALDH1A digestion. Of the predicted peptides that were within the detectable mass range of the mass spectrometer, 100% were detected for ALDH1A1, 100% for ALDH1A2, and 93% for ALDH1A3.
For the peptides that passed the initial criteria described above, the predicted precursor ion and fragments were incorporated into an LC-MS/MS method. Human testicular S10 protein and purifi ed recombinant ALDH1A were trypsin digested to determine which remaining peptides were detected in human testicular tissue and were specifi c for ALDH1A. For each ALDH1A, one peptide was selected as a quantifi cation peptide and a second for verifi cation ( Table 1 ). For the quantifi cation peptide, a matching heavy labeled peptide was synthesized as an internal standard (Pierce, Rockford, IL). In order to account for the efficiency of the trypsin digestion, a heavy labeled peptide with extended sequence over the trypsin cleavage site (lagging end) was synthesized for ALDH1A3. This peptide was labeled at the N terminus with a [ 13  N 2 ]arginine. The lagging end peptide requires two cleavages by trypsin to produce the target peptide. In addition, a separate labeled internal standard for ALDH1A1 and ALDH1A2 quantifi cation peptides was incorporated that did not require trypsin cleavage ( Table 1 ). For each peptide, two fragments from each of the two peptides were used for detection and to confi rm the presence of the protein. The stability of the peptides was validated by confi rming that there was no signifi cant sample degradation after three freeze-thaw cycles, 24 h at 37°C and 24 h at room temperature.
Trypsin digestion of purifi ed ALDH1As and testicular S10 fractions was done in 96-well plates according to the following protocol: First, 15 l of sample (0-400 nM of ALDH1A or 5.33 mg/ml S10 fraction) were added to each well. Next, 5 l of 300 nM ALDH1A3 lagging end peptide was added together with 4 l dithiothreitol (100 mM) and 10 l ammonium bicarbonate buffer (100 mM, pH 7.8). Next, 5 l of 10% sodium deoxycholate (Sigma, St. Louis, MO) was added in the well, and the sample was mixed before incubation at 95°C for 5 min. After cooling to Two signature peptides were chosen for quantifi cation and confi rmation of detection of ALDH1A proteins. The peptides used as an internal standard for quantifi cation are marked with *. All peptides were synthesized with a heavy labeled terminal lysine or arginine. The peptide marked with # was synthesized as EVTDNMRIAKEEIFGPVQPILK and required cleavage by trypsin to be detected. This peptide was used to control for the variability in trypsin digestion across the sample set. For each peptide, values of 10 and 13 were used for the entrance and collision exit potential. monitored as a function of time to generate a chromatogram of the m/z 301.2 + > 205.3 + > 159.1 + transition.

Formation of at RA and 13-cis RA by recombinant ALDH1A enzymes and human testis S10 fractions
Human ALDH1A1 cDNA was purchased from OriGene (SC321535). Full-length human ALDH1A2 cDNA was cloned from human testis RNA using RT-PCR. An open reading frame of ALDH1A1 was subcloned into a pETite expression vector (Lucigen, Middleton, WI) with a 5 ′ hexahistidine tag and that of ALDH1A2 into a pET28 bacterial expression vector (Novagen, Billerica, MA) containing a 3 ′ hexahistidine tag. The resulting plasmids were verifi ed through DNA sequencing for correct frame and sequences. Enzymes were expressed in Escherichia coli [BL21(DE3)] grown to an optical density at 600 nm of 0.6-1.0. Expression was induced with 0.5 mM isopropyl-beta-D-l-thio-galactopyranoside and continued for 4 h at 37°C before harvest . ALDH1A1 and ALDH1A2 were purifi ed using the His.Bind resin and buffer kit (catalog # 69755-3, Novagen) as previously described ( 36 ). Purifi ed enzyme was dialyzed against 20 mM HEPES buffer (pH 8.5) containing 150 mM KCl and 1 mM EDTA. Protein purity was confi rmed by SDS-PAGE followed by Coomassie staining and concentrations were determined using BCA protein assay kit (Thermo Fisher) or protein assay kit II (BioRad, Hercules, CA). Enzyme was stored at 4°C in the dialysis buffer with 1 mM TCEP (Thermo Fisher), a reducing agent. Purifi ed recombinant human ALDH1A3 was purchased from Life Technologies (Grand Island, NY). The lyophilized protein was reconstituted to a fi nal concentration of 0.2 mg/ml with fi ltered water according to the manufacturer's directions and stored at Ϫ 80°C.
To characterize the kinetics of at RA formation by ALDH1A1 and ALDH1A3 enzymes, at-retinal (0-2,000 nM) was incubated with 18 nM ALDH1A in 100 l of buffer consisting of 750 mM KCl and 50 mM HEPES at a pH of 8.0 with 2 mM NAD + . Following a preincubation of 5 min at 37°C, the incubations were initiated with addition of substrate. The incubations were terminated at 5 min by transferring 75 l of the incubation into an equal volume of chilled acetonitrile with 100 nM at RA-d 5 (internal standard). In order to characterize the at RA formation kinetics by ALDH1A2, the protein concentration was decreased to 2 nM, and the incubations were terminated after 2 min. The formation kinetics of 13-cis RA by ALDH1A1 were characterized by incubating 13-cis-retinal (0-15,000 nM) with 15 nM ALDH1A1 and quenching after 5 min. After termination, all plates were spun at 3,000 g for 25 min, the samples were transferred to a 96-well plate and analyzed by LC-MS/MS for at RA and 13-cis RA concentrations. In order to determine whether 13-cis -retinal was a substrate of ALDH1A2 or ALDH1A3, 5,000 nM 13-cis -retinal was incubated with 50 nM ALDH1A2 and ALDH1A3 for up to 45 min. Over the course of 45 min, no enzyme-dependent 13-cis RA formation greater than 0.0007 pmol/min/pmol (based on assay lower limit of detection) was detected. All incubations were conducted under protein and time linearity and as triplicate measurements. The Michaelis-Menten equation was fi tted to the velocity of at RA formation as a function of substrate concentration data to obtain kinetic parameters for each enzyme. When ALDH1A1 was incubated with 13-cis -retinal, 13-cis RA formation demonstrated substrate inhibition kinetics, and the substrate inhibition equation was used to fi t the data according to equation 1 : where v is the predicted velocity, V max is the maximum observed velocity, [S] is the substrate concentration, K m is the affi nity internal standard and was monitored using the m/z 290 → 161 transition. For in vitro incubation samples, the analytes were separated using a Kinetex C18 column (100 × 2.1 mm, 1.7 m particle size) heated to 40°C with a SecurityGuard Ultra UHPLC C18 cartridge (Phenomenex) and a linear gradient with A, water + 0.1% formic acid, and B, ACN + 0.1% formic acid. The mobile phase fl ow was 600 l/min with the following linear gradient: 0 → 0.25 min 40% B, then increased to 95.0% at 4 min, 4 → 5 min 95.0% B, 5 → 6 min 40% B. A standard curve consisting of RA spiked into buffer containing 0.05 mg/ml S10 protein at concentrations of 0, 1, 5, 10, 25, 50, and 100 nM was prepared simultaneously with all incubations. Analyst software (AB Sciex) was used to quantify peak areas for each analyte.
For testicular tissue analysis, the extraction of RA from tissue was optimized to increase the recovery of RA without causing any isomerization, and chromatographic separation was optimized to avoid endogenous interference from tissue. Briefl y, tissue was homogenized with a 2 ml Potter-Elvehjem glass homogenizer (Kimbel Glass, Vineland, NJ) in a 1:1 volume of 0.9% NaCl. After transferring the sample to a 15 ml glass culture tube, a 2:1 volume of ACN with 1% formic acid was added followed by 10 l of 1 M at RA-d 5 , which served as an internal standard. Hexanes (10 ml) were used to extract RA, the organic layer was transferred to a glass tube and dried under nitrogen at 37°C. The sample was reconstituted in 60:40 ACN/water. A standard curve and quality control samples were generated with RA standards spiked into homogenized dog testicular tissue (a kind gift from Dr. Mary Zoulas of the Seattle Animal Shelter) exposed to UV light to destroy endogenous RA. Dog testicular tissue was homogenized in a 1:1 volume with 0.9% NaCl and 198 l were placed in a 15 ml glass culture tube. at RA and 13-cis RA (2 l) standards were added to each tube to fi nal concentrations of 0, 1, 2.5, 5, 7.5, and 10 nM for each standard. The quality control samples were prepared at 1.5, 3.75, and 8.75 nM, and duplicate quality control samples were included together with the standard curve in all analytical assays. at RA and 13-cis RA were separated from endogenous interferences using a 150 × 2.1 mm Supelco Ascentis Express reverse phase amide column (Sigma) with 2.7 m particle size and an Ascentis Express reverse phase amide 2.7 m guard cartridge. The solvents for the UHPLC analysis were A, water + 0.1% formic acid, and B, ACN/methanol (60/40) + 0.1% formic acid. The solvent fl ow was 500 l/min and the following linear gradient was used: 0 → 2 min 40% B, then increased to 95% B at 10 min, 10 → 15 min 95% B, 15 → 17 min 40% B. at RA and 13-cis RA peak areas were normalized to the at RA-d 5 internal standard and peak area ratios were used for quantifi cation using Analyst software (AB Sciex). A signal:noise of nine was set as the minimum threshold for quantitation.
To further confi rm the identity of retinoids detected in the testes, MS/MS/MS was used. To collect reference spectra, at RA and 13-cis RA standards were spiked into 60:40 ACN:water + 0.1% formic acid. The standards and two separate 75 mg testicular tissue samples were extracted and analyzed as described above with the following modifi cations on the MS settings. The m/z 205.3 + fragment generated from m/z 301.2 + (RA) was fragmented using the linear ion trap of the API 5500 with a scan rate of 10,000 daltons/second, Q3 entry barrier of 8.0 volts, and a dynamic fi ll time set for the ion trap. All the fragments generated from m/z 205.3 + between m/z 50 and 200 were collected in the trap and detected. The MS/MS/MS product ion spectrum was collected for each chromatographic peak and the spectra of compounds eluting from the tissue sample were compared with the RA standards. Based on the fact the m/z 159.1 + fragment was the main fragment generated from m/z 205.3 + , from RA, the m/z 159.1 + fragment was extracted from the MS/MS/MS total ion chromatograms and and K I is the concentration of inhibitor when the inactivation rate is half of the k inact .

Quantifi cation of CRBP1 expression in human testes and determination of the effect of CRBP1 to RA formation in human testes
Testicular CRBP1 protein expression was quantifi ed in all tissue donors using an ELISA kit obtained from Cloud Clone Corp. (Houston, TX). Standards and samples were prepared according to the manufacturer's instructions. Testicular S10 protein fractions were diluted 1:20-fold into phosphate buffer (pH 7.4) and CRBP1 quantifi ed based on the standard provided. The amount of CRBP1 in each sample was normalized to milligrams of S10 protein, and the expression was scaled to the concentration of CRBP1 as picomoles per gram testes using the S10 recovery (milligrams S10 protein per gram testes). In addition, the concentration of CRBP1 in the incubations with testicular S10 protein was calculated using the measured concentration in each donor.
The effect of CRBP1 (purchased as purifi ed protein from Origene, Rockville, MD) on at RA formation by ALDH1A was determined using recombinant ALDH1A proteins and testicular S10 protein. CRBP bound with an equal concentration of retinal (holo-CRBP)1 was generated by preincubating 3,000 nM at -retinal for 5 min at room temperature with 3,000 nM CRBP1 in buffer [10% glycerol, 100 mM glycine, 25 mM Tris-HCl (pH 7.3)]. The holo-CRBP1 (300 nM fi nal concentration) or at -retinal (300 nM fi nal concentration) in buffer was used to initiate incubations with 0.05 mg/ml pooled S10 protein from fi ve donors, 18 nM ALDH1A1, 18 nM ALDH1A3, or 2 nM ALDH1A2 in a 100 l volume. The concentration of 300 nM was chosen for holo-CRBP1 and at -retinal concentrations to mimic expected testicular concentrations as closely as possible. All of the incubations were performed in quadruplicate and contained 2 mM NAD + . Incubations were quenched after 5 min (ALDH1A enzymes) or 10 min (S10 protein) by transferring 75 l of the incubation into an equal volume of chilled acetonitrile with 100 nM at RA-d 5 (internal standard). After termination, the samples were analyzed by LC-MS as described above.

Contribution of individual ALDH1A enzymes to RA formation in human testes
To establish the relative importance of each ALDH1A enzyme in RA formation in human testes, the ALDH1A concentrations for each subject and the in vitro kinetics of RA formation from retinal by the recombinant ALDH1A enzymes were used to predict the overall at RA and 13-cis RA formation velocity in each donor. Due to the different K m and k cat values for ALDH1A enzymes with at -retinal as a substrate, their contribution to at RA formation was predicted at two different substrate concentrations, while 13-cis RA formation was predicted only at one concentration. at RA formation in each testis S10 sample was predicted using equation 3: in which v is the predicted product formation velocity, [ALDH1A] is the measured expression level of each individual ALDH1A enzyme, k cat is the maximum product formation velocity measured with recombinant enzymes, f u is the experimentally determined free (unbound) fraction in S10 preparation, [S] is the constant of 13-cis -retinal with the ALDH1A1, and K i is the disassociation constant for 13-cis -retinal binding in a way that allows for the enzyme to bind two substrates. All kinetic analyses were done using GraphPad Prism (La Jolla, CA) at RA and 13-cis RA formation from the corresponding retinals was measured in the individual human testis S10 preparations using a similar method as that described for recombinant ALDH1As. Briefl y, at -retinal at nominal concentrations of 100 nM and 1,000 nM or 13-cis-retinal (1,000 nM) were incubated with S10 fractions (5 g S10 protein/0.1 ml) from each donor. The incubations were initiated with substrate and terminated after 10 min and analyzed as described above. The protein binding of at -retinal and 13-cis -retinal to the nonspecifi c proteins and membranes in the activity assay was determined by ultracentrifugation, as described before ( 37 ). Retinal was added to 0.05 mg/ml pooled S10 protein to fi nal concentrations of 100 and 1,000 nM at -retinal and 1,000 nM 13-cis -retinal. The samples were aliquoted into ultracentrifuge tubes (Beckman 343775) and spun at 435,000 g at 37°C for 90 min using a Sorval Discovery M150 SE ultracentrifuge with a Thermo Scientifi c S100-AT3 rotor or incubated at 37°C for 90 min in a water bath. The incubated sample or supernatant was added to 100 l ice-cold acetonitrile containing 100 nM at -retinal-d 5 . The samples were centrifuged at 3,000 g for 20 min at 4°C. Samples were transferred to a 96-well plate and analyzed using the LC-MS/MS method described. The unbound (free) fraction of retinal in the incubations was calculated as the ratio of retinal concentration in solution with or without ultracentrifugation. This calculation assumes minimal CRBP concentration and binding in the incubations. The insignifi cant concentration of CRBP was confi rmed via direct measurement of CRBP concentrations in the S10 fractions (see below).

ALDH1A inhibition by WIN 18,446
In order to determine reversible inhibition of ALDH1A1, ALDH1A2, and ALDH1A3 by WIN 18,446, recombinant purifi ed enzymes were incubated with 100 nM of at -retinal and increasing concentrations (1-5,000 nM) of WIN 18,446. All activity assays with recombinant enzymes included 18 nM ALDH1A and 2 mM NAD + , and were terminated after 5 min by transferring 75 l of the incubation into an equal volume of chilled acetonitrile with 100 nM at RA-d 5 . The samples were prepared and analyzed by LC-MS/MS as described for incubations with recombinant purifi ed ALDH1A. A screen for time-dependent inhibition (TDI) used the dilution method and at -retinal as a substrate ( 38 ). Briefl y, TDI was determined by incubating 180 nM ALDH1A enzyme with and without 1 M WIN 18,446 in the presence of 2 mM NAD + in a fi nal volume of 100 l of buffer (750 mM KCl and 50 mM HEPES at a pH of 8.0). At time points of 0.25, 3, and 5 min, 10 l aliquots were diluted 10-fold into 100 l of buffer (750 mM KCl and 50 mM HEPES at a pH of 8.0) containing NAD + (2 mM) and a saturating concentration of at -retinal (5,000 nM). All incubations were performed in triplicate with a no NAD + control for the WIN 18,446 and at -retinal incubations. Because WIN 18,446 was a TDI for ALDH1A2, but not ALDH1A1 or ALDH1A3, the kinetics of ALDH1A2 TDI with WIN 18,446 were characterized using eight concentrations of WIN 18,446 (0-5,000 nM). The dilution method was used with time points of 0.25, 1, 3, and 5 min. The k inact and K I were determined by fi tting equation 2 to the data in Graphpad Prism.
[I] is the inhibitor concentration, is the observed apparent inactivation rate (min Ϫ 1 ), k inact is the maximum inactivation rate,

Immunohistochemistry
Testicular tissue from three healthy fertile subjects was used to determine ALDH1A localization. These men volunteered to be part of this research study that was approved in advance by the University of Washington Institutional Review Board. All subjects provided signed written consent prior to any study procedures. Three healthy men aged 28, 52, and 43 years consented to donating testicular biopsies for the study. The tissue slices shown are from a 43 year old subject with left spermatocele, normal sperm counts, and normal hormone levels. Immunohistochemistry was performed using human testis tissue fi xed in Bouin's fi xative using commercial rabbit polyclonal antibodies raised against ALDH1A1 (Abcam, ab24343), ALDH1A2 (Proteintech Group, 13951-1-AP), and ALDH1A3 (Abgent, AP7847a) as previously described ( 40 ). Antigen retrieval was achieved using citrate buffer (pH 6) at a rolling boil for 5 min. Sections were incubated in primary antibody at a concentration of 0.5 g/ml (ALDH1A1), 4 g/ml (ALDH1A2), 2.5 g/ml (ALDH1A3), or 0.1 g/ml in 5% normal goat serum/0.1% BSA in phosphate buffered saline at room temperature overnight ( ‫ف‬ 16 h). Control sections were incubated without primary antibody. Biotinylated goat-anti-rabbit secondary antibody (Invitrogen, 956143b) was applied for 1 h at room temperature, following the manufacturer's instructions. Streptavidin conjugated horseradish peroxidase (Invitrogen, 956143b) was subsequently applied for 1 h at room temperature. Localization was determined by a brown precipitate formed by horseradish peroxidase activity in the presence of 3,3 ′ -diaminobenzidine tetrahydrochloride (Invitrogen, 002020). Sections were counterstained with a 1:3 dilution Harris hematoxylin (Sigma-Aldrich, HHS32-1L), dehydrated, and mounted under glass coverslips using DPX mounting media (VWR International, 360294H). Cell types were determined using nuclear morphology and location within the testis. Immunohistochemistry was performed on three samples for three individuals to ensure consistent results.

Statistical analysis
Linear regression analysis was used and measured at RA and 13-cis RA formation velocities. Linear regression was also used to test whether the measured and predicted 13-cis RA and at RA formation correlated. The correlation between measured at RA concentrations and at RA and 13-cis RA formation velocities was used to test whether in vitro data could predict in vivo at RA concentrations. Unpaired two-tailed Student's t -test was used to determine any signifi cant differences between transgender individuals and the rest of the sample set. P values of less than 0.05 were considered signifi cant for all analyses. All statistical tests and kinetic analyses were performed using GraphPad Prism.

ALDH1A protein quantifi cation
In order to simultaneously quantify all three ALDH1A enzymes in human tissue, a novel LC-MS/MS-based peptide quantifi cation method was developed. Recombinant purifi ed ALDH1A enzymes were used as standards and  Table 1 . The peptides were chosen based on their predicted specifi city and the total substrate concentration, and K m is the affi nity constant of at -retinal with the ALDH1A. Due to the fact 13-ci sRA formation by ALDH1A1 from 13-cis -retinal demonstrated substrate inhibition, an equation accounting for substrate inhibition was used (equation 4). K i is the disassociation constant for 13-cisretinal binding in a way that allows for the enzyme to bind two substrates.
In order to determine the contribution of each ALDH1A enzyme to the total formation of at RA, the predicted velocity of at RA formation by the given ALDH1A was divided by the predicted total velocity at multiple substrate concentrations. The accuracy of the ALDH1A activity predictions was evaluated by comparing the measured velocity to the predicted velocity of at RA and 13-cis RA formation in each subject (S10 fraction) at the different concentrations. The average fold error (afe) of the predictions was calculated for each prediction using equation 5: An afe between 0.5 and 2.0, which indicates a 2-fold prediction error cutoff, was set as the limit for acceptable predictions according to previous studies on clearance predictions ( 39 ).
The relative importance of ALDH1A2 and ALDH1As to at RA formation in the testes was confi rmed using WIN 18,446. Due to the fact WIN 18,446 inactivates ALDH1A2, but not ALDH1A1 or ALDH1A3, the contribution of ALDH1A2 to at RA formation can be determined using the TDI characteristics of WIN 18,446 and the dilution method described above. Briefl y, 625 nM WIN 18,446 and pooled human testes S10 (250 g S10 protein/0. The effect of CRBP1 on ALDH1A activity in the testes was predicted by incorporating the measured in vitro effect of CRBP1 on at RA formation by ALDH1A1 and ALDH1A2 at the physiologically relevant at -retinal concentration of 100 nM. In addition to ALDH1A expression levels and at RA formation kinetics, the observed 52% decrease in at RA formation by ALDH1A1 and 2.7-fold increase of at RA formation by ALDH1A2 in the presence of CRBP1 was incorporated into equation 2 to predict the contribution of each enzyme to at RA formation.
showing the specifi c quantifi ed peptides are depicted in Fig. 1 .
The expression of ALDH1A1, ALDH1A2, and ALDH1A3 was quantifi ed from 33-102 mg samples of testicular tissue collected from 15 men undergoing orchiectomy for treatment of prostate cancer and from 3 male to female observed sensitivity. In order to account for the effi ciency of the trypsin digestion, also a heavy labeled peptide for ALDH1A3 was synthesized with an extended end over the trypsin cleavage site, and hence its detection required cleavage by trypsin ( Table 1 ). Chromatograms of the digested recombinant proteins and the internal standards vitro at RA formation kinetics predict the net at RA formation velocity in the human tissue, at RA formation from at -retinal was measured and predicted at two nominal concentrations (100 and 1,000 nM) of at -retinal for each individual in the study ( Fig. 3 ). The substrate concentrations in the in vitro system were corrected for the protein binding of at -retinal measured at 100 and 1,000 nM at -retinal. The f u of at -retinal was 7.8 ± 0.7% at 100 nM and 18.1 ± 7.8% at 1,000 nM. The measured concentration of CRBP1 in the in vitro incubations was <0.5 nM (see below) and the effect of CRBP binding was considered insignificant for at -retinal in the incubations. Therefore, the free concentrations used in the predictions were 7.8 and 181 nM. Due to the sole contribution of ALDH1A1 to 13-cis RA formation from 13-cis -retinal, the formation of 13-cis RA was measured and predicted only at the nominal concentration of 1,000 nM 13-cis -retinal. The 13-cis -retinal concentration was corrected for the measured unbound fraction in the in vitro system (f u = 60.0 ± 5.1%), and the free concentration of 600 nM 13-cis -retinal was used for all predictions. The average observed and predicted velocities for at RA formation from 7.8 nM free at -retinal in the human testis were 5.6 ± 3.2 pmol/min/mg S10 protein and 2.6 ± 0.6 pmol/min/mg S10 protein, respectively. At 181 nM free at -retinal, the mean observed and predicted velocities were 33.3 ± 7.7 pmol/min/mg S10 protein and 28.9 ± 15.9 pmol/min/mg S10, respectively ( Fig. 3 ). The average observed and predicted 13-cis RA formation velocities at 600 nM 13-cis -retinal were 25.5 ± 9.1 pmol/min/mg S10 and 15.2 ± 3.2 pmol/min/mg S10, respectively. The correlations between the predicted and observed at RA and 13-cis RA formation velocities are shown in Fig. 3 . RA formation (ALDH1A activity) was accurately predicted at each retinal concentration with average fold error values of 0.6 (7.8 nM free at -retinal), 1.3 (181 nM free at -retinal), and 0.8 (600 nM free 13-cis -retinal).

Relative importance of individual ALDH1A enzymes to at RA formation in human testis samples
The relative importance of each ALDH1A enzyme in at RA formation [fraction of retinal metabolized by a given enzyme (f m )] was predicted using ALDH1A expression levels and at RA formation kinetics as a function of substrate concentration. The predicted relative importance of each ALDH1A enzyme at increasing at -retinal concentrations is shown in Fig. 4 . Overall, ALDH1A1 is predicted to be the main enzyme forming at RA in the testis with >80% contribution to the overall at RA formation. ALDH1A2 was predicted to play a role in at RA formation ( ‫ف‬ 15%), while the transgendered individuals undergoing orchiectomy as part of their gender reassignment surgery. S10 fractions containing the cytosolic and microsomal proteins were generated from each individual donor sample. The average yield of S10 protein was 48 ± 11 mg S10 protein/gram testis tissue. Eighty micrograms of this S10 protein was used for ALDH1A quantifi cation. Representative chromatograms of the detection of ALDH1A1, ALDH1A2, and ALDH1A3 in the human testis, and the quantifi cation of the three enzymes in the human testis are shown in Fig. 1 . Based on the expression levels, ALDH1A1 was the major ALDH1A enzyme present in the human testis. In the nontransgendered individuals, the average expression of ALDH1A1 (70.9 ± 15.1 pmol/mg testis) was approximately 42-fold higher than that of ALDH1A2 (1.7 ± 0.3 pmol/mg testis) and 51-fold higher than ALDH1A3 (1.4 ± 0.5 pmol/mg testis). A positive correlation was observed between the expression levels of ALDH1A1 and ALDH1A2 (R 2 = 0.27, P = 0.05) and ALDH1A1 and ALDH1A3 (R 2 = 0.30, P = 0.04). However, there was no correlation between the expression levels of ALDH1A2 and ALDH1A3.

Metabolism of at -retinal and 13-cis -retinal by ALDH1As
The formation kinetics of at RA from at-retinal and 13-cis RA from 13-cis -retinal by purifi ed ALDH1A1, ALDH1A2, and ALDH1A3 were characterized to determine the specifi c activity of each of these enzymes in at RA and 13-cis RA formation ( Table 2 ). All three ALDH1A enzymes formed at RA from at -retinal, but only ALDH1A1 showed detectable formation of 13-cis RA from 13-cis -retinal ( Fig. 2 ). Of the ALDH1A enzymes, at -retinal had the highest affi nity for ALDH1A2 ( K m = 56 nM), and the affi nities of at -retinal to ALDH1A1 and ALDH1A3 were both 5-fold lower ( K m = 285 and 261 nM). In addition, ALDH1A2 had the greatest effi ciency to form at RA. The intrinsic clearance (k cat / K m ) of at RA formation by ALDH1A2 was 27-fold greater than that by ALDH1A1 and 60-fold greater than that by ALDH1A3 ( Table 2 ). Interestingly, 13-cis RA formation by ALDH1A1 demonstrated substrate inhibition kinetics ( Fig. 2 ). The k cat , K m , and K i values relevant for a substrate inhibition model together with kinetic constants obtained for at -retinal with the ALDH1A enzymes are listed in Table 2 .

Characterization of 13-cis RA and at RA formation in human testis samples
To establish the relative importance of the ALDH1A enzymes in the human testis, and to test whether the measured ALDH1A expression levels and the determined in was also evaluated. Using a pooled testis S10 fraction of fi ve men, the IC 50 for WIN 18,446 against testicular ALDH1A activity ( at RA formation) was 88 ± 1 nM. This IC 50 value is in excellent agreement with the IC 50 with recombinant ALDH1A1 suggesting that this enzyme contributes the majority of at RA formation in the in vitro system ( Fig. 4 ) ( Fig. 4 ). This observed 28% contribution of ALDH1A2 to testicular at RA formation is consistent with the average 15% contribution predicted from the recombinant enzyme data.

The effect of CRBP1 on RA formation by ALDH1A enzymes and in testes S10 fractions
To determine whether CRBPs affect the formation of at RA in human testis, the formation of at RA by recombinant ALDH1A and testes S10 protein was measured with holo-CRBP1. The formation of at RA was increased 3-fold in incubations of testis S10 protein with holo-CRBP1 in comparison to at -retinal ( Fig. 6A ). In order to determine the enzyme(s) responsible for the increase in at RA formation, the effect of holo-CRBP1 on at RA formation with each ALDH1A enzyme was determined ( Fig. 6A ). at RA formation by ALDH1A1 was decreased 52% when holo-CRBP1 was used as a substrate compared with at -retinal. In contrast at RA formation by ALDH1A2 was increased 2.7-fold in the presence of CRBP1 when compared with free at -retinal. at RA formation by ALDH1A3 was unchanged in the presence of CRBP1. CRBP1 expression in the human testis was quantifi ed in each study subject and the average concentration of CRBP1 was 445 ± 85 pmol/gram testis (445 nM) ( Fig. 6B ). The effect of CRBP1 on the contribution of each ALDH1A enzyme to at RA formation in the testis was predicted at 100 nM concentration of at -retinal. In the absence of CRBP1, ALDH1A1 is predicted to be the main contributor to at RA formation (84%). However, if all at -retinal is bound to CRBP1, ALDH1A2 is predicted to contribute to the majority of at RA formation (52%) ( Fig. 6C ).

Alterations in the intratesticular ALDH1A activity in transgender individuals
To further evaluate the function of ALDH1A enzymes in the human testis, the ALDH1A activity and protein concentrations were measured in a sample set of three transgender individuals who had undergone hormonal sex reassignment therapy. The average protein concentrations of ALDH1A1, ALDH1A2, and ALDH1A3 were 54.0 ± 14.7 pmol/mg S10 protein, 1.3 ± 0.2 pmol/mg S10 protein, contribution of ALDH1A3 appeared insignifi cant. Due to the different K m values for at -retinal with ALDH1A2 and ALDH1A1, the predicted contribution of ALDH1A2 increases with decreasing retinal concentrations. The published concentration of retinal in the mouse testis (90.7 pmol/gram) ( 41 ) is similar to ALDH1A2 K m . However, the free concentration of retinal in the testis is unknown as well as the fraction of retinal bound to CRBP1 in the testis and hence the contribution of ALDH1A2 could be greater.
Based on the selective formation of 13-cis RA by ALDH1A1, ALDH1A1 was predicted to be the only contributor to 13-cis RA formation in incubations with at -retinal. In agreement with this, a signifi cant ( P < 0.01) correlation between ALDH1A1 protein expression and 13-cis RA formation (R 2 = 0.41) was observed ( Fig. 5 ). In addition, ALDH1A1 protein concentration had a signifi cant correlation with at RA formation velocity measured with 181 nM at -retinal ( P = 0.02, R 2 = 0.28) ( Fig. 5 ). However, only ALDH1A2 expression had a positive correlation ( P = 0.02, R 2 = 0.33) with at RA formation measured at 7.8 nM at -retinal ( Fig. 5 ).
To further evaluate the relative contributions of ALDH1A enzymes to at RA formation, an inhibitor of ALDH1A enzymes, WIN 18,446, was used as a selective ALDH1A inhibitor. WIN 18,446 was found to be a potent reversible inhibitor of ALDH1A1 and ALDH1A3 (IC 50 values 102 ± 2 nM and 187 ± 1 nM), and an effi cient time dependent inhibitor (k inact = 22.0 ± 2.4 h Ϫ 1 , K I = 1,026 ± 374 nM) of ALDH1A2 ( Fig. 4 ). Based on the strong inhibition of all three ALDH1A enzymes by WIN 18,446, the effect of WIN 18,446 on at RA formation in human testis S10 fractions Because only ALDH1A1 formed 13-cis RA from 13-cis -retinal, only 13-cis RA formation by ALDH1A1 was characterized (B). The kinetic constants obtained are shown in Table 2 .
Hence, it was hypothesized that the tissue localization of the ALDH1A enzymes within the testis may explain this phenotypic discrepancy. To test this, immunohistochemistry was performed to investigate the cell type-specifi c expression pattern of the three ALDH1A enzymes within testicular tissue from three healthy fertile men ( Fig. 8 ).
The three ALDH1A enzymes were expressed in distinct cell types and the expression patterns were different for each isozyme. Sertoli and peritubular myoid cells were all positive for ALDH1A1, but this enzyme could not be detected within any germ cell type ( Fig. 8A ). In contrast, ALDH1A2 localized to pachytene spermatocytes and round spermatids with signal also present in some spermatogonia and peritubular myoid cells ( Fig. 8B-D ). The third isoform, ALDH1A3, displayed a similar expression pattern, with protein detected in pachytene spermatocytes, spermatogonia, and Sertoli cells ( Fig. 8E ). Under these conditions, cells of the interstitium are positive without primary antibody, and therefore are possibly nonspecifi cally stained in all. Taken together, the localization patterns of the individual ALDH1A enzymes suggest that and 1.3 ± 0.5 pmol/mg S10 protein in the transgender donors ( Fig. 7 ). ALDH1A2 expression was signifi cantly ( P < 0.05) decreased in the testes of the transgendered individuals compared with the men. Additionally, the ALDH1A1 expression level was 20% lower in these donors than in the other donors, but this difference did not achieve signifi cance ( P > 0.05). The ALDH1A mediated at RA formation was signifi cantly ( P < 0.02) lower in the testis from transgender donors than in the men undergoing orchiectomy at both concentrations of at -retinal (1.8 ± 0.6 pmol/min/mg S10 protein at 7.8 nM and 9.0 ± 3.4 pmol/min/mg S10 protein at 181 nM) ( Fig. 7 ). Additionally, the formation of 13-cis RA from 13-cis -retinal was signifi cantly ( P < 0.01) decreased (8.9 ± 4.1 pmol/min/mg S10 protein) in these individuals when compared with the other donors.

ALDH1A localization in the human testis
The predicted relative contribution of ALDH1A1 to at RA formation in the testis must be reconciled with the fact that male ALDH1A1 knockout mice are fertile ( 14 ). Fig. 3. Comparison of the predicted and measured ALDH1A activity in human testis. at RA formation was measured and predicted for the 18 subjects with at -retinal as the substrate at free concentrations of 8 nM (red circles) and 181 nM (blue circles) (A). Each dot represents an individual in the study. The measured activity was plotted as a function of predicted activity in each subject with 7.8 nM free at -retinal (B), 181 nM free at -retinal (C), and 600 nM free 13-cis -retinal (D). The blue dotted line represents an exact correlation between predicted and measured RA formation (ALDH1A activity). The red lines represent the range between 2-fold overprediction and 0.5-fold underprediction. the unknown peak, the 159 + fragment was extracted from the MS/MS/MS data ( Fig. 9C ). The extraction of the fragment allows for clear identifi cation of RA isomers from matrix components observed using MRM. The ALDH1A activity ( at RA formation) measured at the biologically relevant 7.8 nM concentration of at -retinal correlated positively with intratesticular 13-cis RA concentrations (R 2 = 0.40, P = 0.01) and total RA concentration (R 2 = 0.30, P = 0.03). However, intratesticular at RA concentrations did not correlate signifi cantly with ALDH1A activity at either of the at-retinal concentrations ( P = 0.22 at 7.8 nM and 0.94 at 181 nM) ( Fig. 9D ).

DISCUSSION
The current study describes the development of novel methods necessary to study the formation of RA by ALDH1A enzymes and the application of these methods to characterization of RA formation in human testis. While the indispensible nature of RA in the testis is well-established, the complement of human enzyme(s) responsible for RA formation has been poorly characterized. The methods developed in this study provide the foundation to determine the tissue-specifi c importance of each ALDH1A enzyme to RA formation, and establish novel tools to evaluate the various cell types of the testis can all synthesize RA, but do so utilizing distinct ALDH1A isoforms.

Correlation of intratesticular RA concentrations with ALDH1A activity
To test whether tissue at RA and 13-cis RA concentrations were affected by ALDH1A activity in the same tissue, intratesticular at RA and 13-cis RA concentrations were measured in the testes samples from the 15 donors with prostate cancer ( Fig. 9 ). The intratesticular RA concentrations were 4.3 ± 0.5 pmol/gram for at RA and 5.1 ± 0.6 pmol/gram for 13-cis RA, and concentrations of at RA and 13-cis RA were positively correlated (R 2 = 0.36, P = 0.02). An unknown peak was observed in each sample that eluted between the at RA and 13-cis RA isomers. Based on the fact the 9-cis RA isomer elutes between the at RA and 13-cis RA isomers (35), MS/MS/MS was used to determine whether the unknown peak was a RA isomer. The fragmentation patterns of the at RA and 13-cis RA isomers were identical and generated MS/MS/MS fragments of m/z 159 + , 119 + , 131 + , and 105 + ( Fig. 9B ). An identical fragmentation pattern was observed for the at RA and 13-cis RA peaks measured from human testicular tissue. However, the fragments generated from the unknown peak were distinctly different and suggest it is not a RA isomer. Based on the abundance of the m/z 159 + fragment formed from RA and not  WIN 18,446, the rate of inactivation was determined with increasing concentrations of inhibitor (D, inset) and plotted as a function of inhibitor concentration (D) to determine the K I and k inact . In order to measure the f m for ALDH1A2 in the testis, S10 protein was preincubated with inhibitor for 0.25, 10, and 30 min to inactivate any ALDH1A2 protein (E). The preincubation was diluted out to remove reversible inhibition, and the AL-DH1A activity was determined. The 25% decrease in ALDH1A activity (E) can be attributed to the contribution by ALDH1A2. The predicted relative contribution of each ALDH1A enzyme (green, ALDH1A1; blue, ALDH1A2; red, ALDH1A3) to at RA formation as a function of at -retinal concentration is shown in (F). > ALDH1A1 (29%) > ALDH1A3 (12%) ( 18 ). This is in contrast to the current study showing ALDH1A1 being the predominant enzyme and emphasizes the need to measure protein expression levels instead of mRNA. Another previous study used in situ hybridization to determine the localization of ALDH1A enzymes in the mouse testis. In that study, ALDH1A1 was found in Leydig cells and Sertoli cells, ALDH1A2 in pachytene spermatocytes and step 1-6 spermatids, and ALDH1A3 in Leydig cells ( 12 ). The protein localization reported here is in generally good agreement with the previous mouse mRNA data, although, unlike in mouse mRNA studies, ALDH1A2 protein was found in the spermatogonia of the human testis. The distinct ALDH1A localization in human testicular tissue suggests that each ALDH1A may play a different role in RA formation in different cell types of the testis. Previous work in knockout mice has shown that ALDH1A enzymes in Sertoli cells are responsible for the initial source of RA during spermatogenesis, while ALDH1A in spermatocytes may help maintain RA formation after the fi rst wave of spermatogenesis ( 30 ).
This study shows that overall in the human testis, ALDH1A1 and ALDH1A2 contribute signifi cantly to formation of at RA from at -retinal and their relative contributions will depend on the interactions of the ALDH1A enzymes with CRBP1 in the specifi c cell types. The WIN 18,446 inhibition data unequivocally shows that RA formation in the testis is dependent on ALDH1A enzymes and not other enzymes such as aldehyde oxidases or xanthine oxidases. Together with the correlation between ALDH1A expression and at RA and 13-cis RA formation in vitro, the WIN 18,446 inhibition confi rms the validity of the protein quantifi cation and recombinant enzyme kinetic data. The correlation between ALDH1A2 expression and at RA formation together with the predicted importance of ALDH1A2 in the presence of CRBP, suggests that ALDH1A2 makes a signifi cant contribution to at RA formation, and that its expression is at least in part responsible for the interindividual variability in RA formation in the testis. As shown by the expression of CRBP1 in the human testes and the ALDH1A isoform specifi c effects of CRBP1 on RA formation, prediction of the in vivo contribution of each individual ALDH1A enzyme is complex and likely also affected by the fraction of CRBP1 that is bound by retinol in vivo. The expression of cellular RA binding proteins that may affect product release from ALDH1As may also alter the predicted contributions in vivo. Previously, rat ALDH1A1 has been shown to recognize holo-CRBP1 as a substrate ( 9 ), but the effect of CRBP1 on at RA formation by human ALDH1A enzymes has never been reported. Similarly CBRP1 expression in the Sertoli cells of rats ( 42 ) has been shown suggesting that CRBPs play an important role in modulating intratesticular RA metabolism. While ALDH1A1 was predicted to be the major contributor to at RA formation in the absence of CRBPs, when the inhibition of ALDH1A1 and activation of ALDH1A2 by CRBP1 was accounted for, ALDH1A2 was predicted to be the predominant enzyme forming at RA in the testis. The observed increase in RA formation in the presence of CRBP1 in the testis S10 fractions is in the cell and tissue-specifi c mechanisms that allow for generation of RA gradients required for the regulation of biological processes. The developed methods are particularly useful for evaluating the role of altered RA homeostasis in human diseases, as all these methods can be used to study ALDH1A localization, expression level, activity, and RA concentration using a single tissue biopsy of 100 mg.
The testis ALDH1A expression levels reported in this study present, for the fi rst time, the quantitative expression levels of ALDH1A1, ALDH1A2, and ALDH1A3 in any human tissue. Human tissue-specifi c mRNA expression profi les of each ALDH1A enzyme have been previously shown and based on the mRNA levels, the rank order of ALDH1A expression in the testis is ALDH1A2 (59%) Fig. 5. The correlation between ALDH1A protein expression and RA formation velocity measured in S10 fraction from testes from each individual in the study. The correlations between 13-cis RA (A) and at RA (B) formation with ALDH1A1 expression are shown with nominal 1,000 nM substrate. The correlation between at RA formation and ALDH1A2 expression at 100 nM nominal substrate concentration is shown in (C). determined from the substrate inhibition model could be used to predict 13-cis RA formation velocity in individual human samples in vitro. The fact that ALDH1A1 was the one enzyme able to form 13-cis RA is of interest because in a preliminary study of 24 men, an association between intratesticular 13-cis RA and abnormal semen was revealed, suggesting that altered intratesticular RA homeostasis may be present in infertile men ( 43 ). In addition, 13-cis RA treatment has been shown to increase sperm production in humans ( 44 ). Based on the results of this study, the decreased intratesticular 13-cis RA concentration may suggest a decreased expression of ALDH1A1 in men with abnormal agreement with the effects observed in the recombinant enzymes and supports the role of ALDH1A2 in RA formation in the testis. The predicted primary contribution of ALDH1A2 to RA formation in the presence of CRBP1 is also in agreement with the fact that ALDH1A1 Ϫ / Ϫ mice are fertile and no obvious defects are detected in these mice at adulthood.
Interestingly, ALDH1A1 was the only ALDH1A enzyme capable of forming 13-cis RA from 13-cis -retinal. The 13-cis RA formation by ALDH1A1 displayed substrate inhibition kinetics suggesting multiple 13-cis -retinal molecules may simultaneously bind ALDH1A1. Despite this, the kinetics Fig. 6. The effect of CRBP1 on at RA formation and CRBP expression level in the human testis. The effect of CRBP1 on at RA formation by testicular S10 protein and by recombinant ALDH1A1, ALDH1A2, and ALDH1A3 was determined and the formation of at RA from 300 nM holo-CRBP compared with free at -retinal (300 nM) is shown (A). The concentration of 300 nM was chosen to mimic expected in vivo concentrations as closely as possible. CRBP1 expression in the testis S10 fractions was quantifi ed in all subjects by ELISA and the expression levels are shown (B). The relative contribution of each ALDH1A enzyme to at RA formation in the presence or absence of CRBP1 was predicted from a substrate ( at -retinal or holo-CRBP) at 100 nM as described in the Materials and Methods and the predicted percent contribution by each ALDH1A is shown in (C). populations of men, including young healthy men and infertile men are needed to fully characterize the role of the different ALDH1A enzymes in male reproduction. Taken together, these data demonstrate the critical importance of understanding the specifi c enzymology of RA synthesis in the testis.
The correlation discovered between RA formation activity and RA concentration in the human testis described here is, to our knowledge, the fi rst published example of a correlation between human tissue concentrations of an endogenous substance and the activity of the enzyme responsible for its formation. With the evidence suggesting intratesticular RA is formed within the tissue and assuming no interindividual differences in the clearance of intratesticular RA, a correlation between the RA synthesis rate and the measured intratesticular RA is expected. The reason intratesticular 13-cis RA correlated with ALDH1A activity and at RA did not, may be due to the fact 13-cis RA is a poor substrate for CYP26 enzymes compared with at RA ( 47 ). Therefore, interindividual differences in CYP26 expression are more likely to affect the intratesticular concentration of at RA.
The results of this study demonstrate the usefulness of novel MS approaches to model RA formation in human tissue, using the testis as an example. RA formation kinetics was accurately characterized using two novel methods to measure ALDH1A concentrations and determine RA formation by ALDH1A in human tissue biopsies. ALDH1A1 semen. However, at present the source of intratesticular 13-cis RA is not known and it could be formed either via isomerization from at RA or enzymatically from 13-cis -retinal if it is present in human tissues. Also, it is unknown if 13 -cis RA is necessary for spermatogenesis, and the cause of the differences in RA concentrations with different fertility diagnoses has not been identifi ed.
A signifi cant decrease in the RA formation velocity was observed in the testicular tissue of transgender individuals, possibly due to the signifi cant decrease in ALDH1A2 expression. It seems likely that the presurgical hormonal therapy with estradiol commonly given prior to gender reassignment surgery may have affected ALDH1A2 expression. In support of this notion, a mouse model lacking a functional androgen receptor has signifi cantly decreased mRNA for both Aldh1a1 and Aldh1a2 ( 45 ). Additionally, ALDH1A2 expression is signifi cantly decreased in infertile men diagnosed with Sertoli cell only syndrome on testicular biopsy ( 46 ). The reduced expression of ALDH1A2 in these men is likely related to the reduction in the numbers of developing sperm cells, consistent with our observation that spermatogonia, spermatids, and spermatocytes were the main cell type staining for ALDH1A2 in the histological analysis. Although the donors in this study were men with prostate cancer undergoing orchiectomy, the good agreement with this data and prior clinical fi ndings suggests that these men refl ect healthy, fertile men in terms of testicular function. Yet, further studies in different A negative control (F) used rabbit IgG instead of primary rabbit antibody. The original magnifi cation was 225× for each. The bar at lower left of each panel represents 10 microns. Cell types are indicated by arrows as follows: blue, Sertoli cells; yellow, spermatogonia; red, primary spermatocytes; green, round spermatids; black, peritubular cells; gray, interstitial cells.
is the predominant enzyme in the formation of RA in the Sertoli cells, while the specifi c localization of ALDH1A2 in the developing sperm indicates a critical role for ALDH1A2 in these cell types. The distinct localization of the ALDH1A enzymes, together with their different interactions with CRBP1, suggests distinct cell type-specifi c roles of the different ALDH1A enzymes in intratesticular at RA formation. Further studies are needed to establish the full biological relevance of these fi ndings. These results illustrate the need to determine the contribution of each ALDH1A to RA formation and the localization of each enzyme in individual target tissues. Tissue RA concentrations by themselves will not give an accurate representation of how RA formation is controlled within the tissue. Furthermore, given the indispensible role of RA in many other biological functions, the methods described will provide a valuable tool in understanding how RA homeostasis is altered by or potentially leads to disorders associated with changes in ALDH1A expression.
The authors wish to thank Dr. Mary Ellen Zoulas, DVM, at the Seattle Animal Shelter for donating the dog testicular tissue, Reinhild Sandhowe for technical support in processing the human tissue samples, and Dr. Maureen Kane for helpful discussion regarding RA quantifi cation.