Address correspondence and reprint requests to Arturo Ortega, PhD, Departamento de Genética y Biología Molecular, Cinvestav-IPN, Apartado Postal 14–740 México DF 07000, México. E-mail: firstname.lastname@example.org
Glutamate, the main excitatory neurotransmitter in the vertebrate brain, is critically involved in gene expression regulation in neurons and in glia cells. Neuron–glia interactions provide the framework for synaptic plasticity. Retinal and cerebellar radial glia cells surround glutamatergic excitatory synapses and sense synaptic activity through glutamate receptors expressed in their membranes. Several glutamate-dependent membrane to nuclei signaling cascades have been described in these cells. Octamer DNA binding factors, namely Oct-1 and Oct-2 recognize similar DNA sequences on regulatory regions, but their final transcriptional effect depends on several factors. By these means, different responses can be achieved in different cell types. Here, we describe a comparison between the glutamate-induced DNA binding of octamer factors and their functional activities in two important types of radial glia, retinal Müller and cerebellar Bergmann glial cells. While Oct-1 is expressed in both cell types and in both glutamate treatments results in an increase in Oct-1 DNA binding, this complex is capable of transactivating a reporter gene only in Müller glia cells. In contrast, Oct-2 expression is restricted to Bergmann glia cells in which glutamate treatment results in an augmentation of Oct-2 DNA binding complexes and the repression of kainate binding protein gene transcription. Our present findings demonstrate a differential role for Oct-1 and Oct-2 transcription factors in glial glutamate signaling, and further strengthen the notion of an important role for glial cells in glutamatergic transactions in the central nervous system.
Radial glia cells play an important role during development, supporting neuronal migration and laminar patterning. Unlike other glia cells, Bergmann and Müller radial glia are not converted into astrocytes after birth (Cameron and Rakic 1991). Cumulative evidence indicates the important role of glia cells in the adult brain as modulators of neuronal excitability (Hansson and Ronnback 1995; Anthony et al. 2004). Müller glia cells (MGC) are the predominant glial cells in the retina, comprising about 90% of the retinal glia, and their processes interdigitate with the perikarya and processes of neurons. These cells participate in basic and essential retinal functions such as neurotransmitter uptake and inactivation, K+ homeostasis, nutrient supply, pH regulation and retinoid metabolism (White and Neal 1976; Rauen and Kanner 1994; Derouiche and Rauen 1995; Francke et al. 1995).
Cerebellar Bergmann glia cells (BGC) are radial glia cells that extend their processes through the molecular layer of the cerebellar cortex, enveloping the excitatory and inhibitory synapses present in this layer (Somogyi et al. 1990). BGC have recently been regarded as a neuronal reservoir (Malatesta et al. 2003; Anthony et al. 2004).
The octamer motif (ATGCAAT), present in the promoter region of numerous genes, acts as an activator or repressor cis element depending on the octamer-binding proteins expressed in a particular cell type and their interaction with other co-factors. For example, the immunoglobulin genes are specifically expressed in B lymphocytes due to the fact that Oct-2 is present in these cells (Latchman 1996). Furthermore, in embryonic carcinoma cells, an octamer-dependent inhibition of the immunoglobulin enhancer takes place, again due to the presence of octamer DNA binding factor 2, Oct-2 (Latchman 1996). Therefore, it is clear that Oct-2 is a transcription factor that exhibits cell specificity and cell-determined interaction with co-factors (Schreiber et al. 1988; Lillycrop et al. 1991; Lillycrop and Latchman 1992; Latchman 1996). Within the central nervous system, Oct-2 is apparently restricted to neurons, although we have reported its expression in BGC (Mendez et al. 2004). Neuronal isoforms, like Oct-2.5, are typically repressors (Dent and Latchman 1991; Lillycrop et al. 1991). In BGC, activation of Oct-2 by glutamate through AMPA receptor activation negatively regulates the chkbp promoter via a signaling pathway that includes protein kinase C (PKC) and NF-κB (Mendez et al. 2005).
As Oct-1 is widely expressed, and MGC and BGC are in the vicinity of glutamatergic synapses, we investigated whether Glu receptor activation is linked to Oct-1 DNA binding in these cells. We detected a Glu-dependent increase in Oct-1 DNA binding activity, and we established that Oct-1 is capable of transactivating a reporter gene in a cell-specific fashion. These results suggest that the POU family of transcription factors plays an important and active role in Glu-dependent transcriptional control in glial cells.
Tissue culture reagents were obtained from Invitrogen (Gaithersburg, MD, USA). Glutamatergic agonists [Glu, 1-amino-4, 5-cyclopentane-trans-1, 3-dicarboxylate (trans-ACPD), AMPA, l-(+)-2-amino-4-phosphobutyric acid (L-AP4), (RS)-3,5-dihydroxyphenylglycine (DHPG)] and Glu antagonists [6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), 6,7-dinitroquinoxaline-2,3-dione (DNQX)] were obtained from Tocris-Cookson (St. Louis, MO, USA). Kainate was obtained from Ocean Produce (Shelburne, NS, Canada). Protease inhibitors were purchased from Roche (Indianapolis, IN, USA). The rabbit polyclonal antibodies used were: anti-Oct-1 sc-232, anti-Oct-2 sc-233 and anti-p53 sc-6243 (obtained from Santa Cruz Biotechnology, Santa Cruz, CA, USA), and monoclonal anti-actin antibody (kindly donated by Dr Manuel Hernández, Cinvestav). Horseradish peroxidase-linked anti-mouse or anti-rabbit, and the enhanced chemiluminescence reagent (ECL), were obtained from Amersham Biosciences (Little Chalfont, UK). All other chemicals were from Sigma (St. Louis, MO, USA). Fluoresceinated anti-rabbit antibodies were purchased from Zymed Laboratories (San Francisco, CA, USA).
Cell culture and stimulation protocol
Primary cultures of MGC were obtained as described previously (Lopez-Colome and Romo-de Vivar 1991). Retinas from 7-day-old chick embryos (Alpes, Puebla, México) were dissected and washed in Hank's solution free from Ca2+ and Mg2+ (g/100 mL): NaCl 0.8, KC1 0.04, KH2PO4 0.006, Na2HPO4 0.0125, phenol red 0.002, glucose 0.1. Tissue was dissociated in 0.25% trypsin followed by filtration through a 50 µm mesh nylon net, then resuspended in minimum essential medium (MEM) containing 0.05% glucose, 25 mm NaHCO3, 0.0125% gentamycin, 0.0125% penicillin, 0.0125% streptomycin, 0.025% neomycin and 10% fetal bovine serum (FBS). Cells were seeded onto 6-well plates, at a density of 7.5 × 105 cells per well, and maintained at 37°C in a humidified atmosphere of 5% CO2 : 95% air. For immunostaining, cells were seeded on #1 glass coverslips (1.75 × 105 cells per well). Purity of the culture was assessed by glial fibrillary acidic protein (GFAP) (Bjorklund et al. 1985) and neuron-specific enolase (NSE) antibodies (Schmechel et al. 1980); 95% of the cells were GFAP+ and NSE– at day 12 in vitro, at which time cultures had formed a confluent monolayer. The medium was changed every other day, and 80% confluent cell cultures were used for all experiments.
Primary cultures of cerebellar BGC were prepared from 14-day-old chick embryos as described previously (Ortega et al. 1991). Cells were plated in 60 mm diameter plastic culture dishes in DMEM containing 10% FBS, 2 mm glutamine and gentamycin (50 mg/mL). Cells were used on the fourth or fifth day of culture. Prior to any drug exposure, confluent monolayers from both cell cultures were incubated for 7 h in 0.5% serum medium and then treated as indicated. Antagonists were added 30 min before the agonists. HeLa cells were grown under the same media conditions as those used for BGC.
Electrophoretic mobility shift assays (EMSA)
Nuclear extracts were prepared as described previously (Lopez-Bayghen et al. 1996). All buffers contained a protease inhibitor cocktail to prevent nuclear factor proteolysis. Protein concentration was measured by the Bradford method (Bradford 1976). Nuclear extracts (approximately 5 µg) from cells were incubated on ice with 1 µg poly[dI-dC] as non-specific competitor (Amersham Biosciences) and 1 ng [32P]-end-labeled double-stranded oligonucleotides: Oct-1: 5′-CTAGTGTCGAATGCAAATCACTAGAA-3′; Oct-2/SIL-2: 5′-AGCTTTATCTGTATTTTCCGAGTC-3′; OctM: 5′-TGTCGAATGCAAGCCACTAGAA-3′; Sp-1: 5′-CTAGATTCGATCGGGGCGGGGCGA-3′.
The reaction mixtures were incubated for 10 min on ice, then electrophoresed through 6% polyacrylamide gels using a low ionic strength 0.5× TBE buffer (89nM Tris-Borate and 2nM EDTA, pH 8.3). The gels were dried and exposed to an autoradiographic film, or scanned with a Typhoon Optical Scanner (Amersham Biosciences). For competition assays, the reaction mixtures were pre-incubated with the non-labeled unrelated probes for 10 min before adding labeled DNA. For gel supershift assays (ImmunoEMSA), reaction mixtures were incubated at 4°C with either anti-Oct-1 or anti-p53 antibodies for 16 h prior to electrophoresis.
Cytoplasmic extracts were obtained from the same cell preparations as used for obtaining nuclear extracts (Schreiber et al. 1989). We carefully re-centrifuged the samples (10 min in a microfuge, 4°C) to eliminate all remaining nuclei, collected the supernatant fluids and ran a nuclear 4′,6-diamidino-2-phenylindole (DAPI) staining test to detect any contamination by nuclei. With nuclei-free fractions, cytoplasmic proteins (approximately 25 µg) were incubated on ice with 3 µg poly[dI-dC] as non-specific competitor (Amersham Biosciences) and 1 ng of [32P]-end-labeled double-stranded Oct-1 and Sp-1 oligonucleotides. The reaction mixtures were incubated for 10 min on ice and electrophoresed as noted above.
Cell-culture staining with polyclonal antibodies was performed. BGC and MGC were grown on glass coverslips (22 × 22 mm) under the same culture conditions as described above. Cells were fixed by exposure to methanol/acetone at − 20°C for 5 min, washed twice with phosphate-buffered saline (PBS) and exposed to anti-Oct-1 or anti-Oct-2 rabbit polyclonal antibodies for 16 h at room temperature (22°C). Binding of the primary antibodies was visualized using fluorescein-labeled (1 : 100) goat anti-rabbit antibodies. Control of immunolabeling was performed with the same staining procedure, using the visualizing reagents in the absence of the primary antibodies. The coverslips were mounted with Vectashield (Vector Laboratories, Burlingame, CA, USA) and the fluorescence examined using a Zeiss Axioscope 40 immunofluorescence microscope and Axiovision software (Carl Zeiss, Inc., Thornwood, NY, USA). Nuclei were counter-stained using propidium iodide.
Plasmids, transient transfections and chloramphenicol-acetyl-transferase (CAT) assays
The plasmid p3XOCT-CAT contains three copies in tandem of the Oct-1 oligonucleotide cloned in front of the promoter of Simian virus 40 and the CAT reporter gene in the pCAT-PROMOTER vector (Promega, Madison, WI, USA). Transient transfections and CAT assays were performed in 80% confluent Müller cultures using a calcium phosphate protocol with 3 µg plasmid reporter construct. Under such conditions, the transfection efficacy is approximately 30%, determined in every cell batch by a transfection control (β□-gal). Treatment with Glu, agonists and antagonists was performed 24 h post-transfection for the times and concentrations indicated in the Figure legends. Protein lysates were obtained as follows: cells were harvested in TEN buffer (40 mm Tris-HCl pH 8.0, 1 mm EDTA, 15 mm NaCl), lysed with three freeze-thaw cycles in 0.25 m Tris-HCl pH 8.0 and centrifuged at 12 000 g for 3 min. Equal amounts of protein lysates (about 80 µg) were incubated with 0.25 µCi [14C]-chloramphenicol (50 mCi/mmol, Amersham Biosciences) and 0.8 mm acetyl-CoA (Sigma) at 37°C. Acetylated forms were separated by thin layer chromatography and quantified using a Typhoon Optical Scanner (Molecular Dynamics). CAT activities were expressed as the acetylated fraction corrected for the activity in the pCAT-BASIC vector (no-promoter) and are expressed as activities relative to non-treated control cell lysates. Transient co-transfections for the expression of the full-length Oct-1 (Wong et al. 1998) or Oct-2.5 isoform (Chapman and Latchman 1998) were performed using the same procedure with the indicated amounts. Where specified, empty vector was co-transfected using the same plasmidic DNA concentrations.
Sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and western blots
Equal amounts of nuclear protein extracts or whole extracts (approximately 50 µg per lane) were denatured in Laemmli's sample buffer, resolved through 10% SDS polyacrylamide gels and electroblotted onto polyvinylidene difluoride (PVDF) membranes. Blots were stained with Ponceau S to confirm that protein loading was equal in all lanes. Membranes were soaked in PBS to remove the Ponceau S, then incubated in PBS containing 5% dried skimmed milk and 0.1% Tween 20 for 2 h to block the excess of non-specific protein binding sites. They were incubated overnight at 4°C with the primary antibodies diluted in 5% dried skimmed milk and 0.1% Tween in TBS buffer, followed by secondary antibodies. Finally, the proteins were detected using an ECL western blot detection kit (Amersham Biosciences). Whole cell extracts were obtained as follows: cells from confluent monolayers were harvested and washed several times in PBS (10 mm K2HPO4/KH2PO4, 150 mm NaCl, pH 7.4); they were then lysed in 50 mm Tris-HCl, pH 7.5, with protease inhibitors [0.5 mm phenylmethylsulfonyl fluoride (PMSF), 1 mg/mL aprotinin and 1 mg/mL leupepetin], and aliquots of this suspension were used for protein concentration determination (Bradford 1976).
Data are expressed as the mean ± SE. A one-way anova was performed to determine whether there were significant differences between conditions. When this analysis indicated significance (at the 0.05 level), the post-hoc Student–Newman–Keuls test was used to analyze which conditions were significantly different from each other (Prism, GraphPad Software, San Diego, CA, USA).
Glu induces Oct-1 DNA binding in Bergmann and Müller glial cells
Oct-1 is a ubiquitous transcription factor whereas Oct-2 is characterized by its restricted cell-specific expression. The primary regulating factor for lymphocytic gene expression is Oct-2.1, while Oct-2.4/2.5 are neuronal-specific isoforms with negative regulatory roles (Latchman 1996). We have recently reported that Oct-1 and Oct-2 are present in cultured BGC from chick cerebellum. Interestingly, following Glu receptor stimulation, Oct-2 is activated and down-regulates the chkbp promoter (Mendez et al. 2004, 2005). These results prompted us to extend our search for activities and responses of octameric factors in MGC, another type of radial glia cell. To this end, we end-labeled a consensus octameric sequence (Oct-1) and used it as a probe in mobility gel shift assays (EMSA). We analyzed the protein DNA complexes obtained after incubation of nuclear extracts, prepared from control or stimulated cells, with the labeled double-stranded probe. The results are presented in Fig. 1; a slow migrating complex, labeled complex A, is significantly augmented after Glu or KA exposure in both radial glial cells. This complex is very similar in migration to that reported previously in rodent fibroblasts, and may possibly represent Oct-1 (Dent and Latchman 1991). As expected, nuclear extracts from chick fibroblasts show this complex.
It is clear from our assays that complex A is the most abundant complex in MGC, especially if we compare it with the Oct-2-associated complexes detected either as the faster migrating complexes in Fig. 1a, or those obtained using as a probe the octamer sequence from the chkbp promoter, Oct-2/Sil2 (Fig. 1b). We have previously characterized these complexes as comprising Oct-2 isoforms in BGC (Mendez et al. 2004). Note that Oct-2 complexes are barely present in MGC nuclear extracts, suggesting that in these cells Oct-2 is expressed at lower levels.
Competition experiments, using nuclear extracts from MGC, demonstrate that complex A disappears with a 100-fold excess of the cold probe, and that it is unaffected by a mutated octamer version or by an unrelated probe such as Sp-1 (Fig. 1c). Using immunoEMSA, we noticed that, together with the appearance of a clear, slower-migrating complex, complex A disappears, indicating that anti-Oct-1 antibodies recognize only the upper complex, the complex we re-named Oct-1 (Fig. 1d).
Assuming that the increase in Oct-1 DNA binding upon Glu exposure could be the result of a Glu-dependent augmentation in Oct-1 nuclear translocation that may or not be linked to higher Oct-1 protein levels, we explored these (Oct-1) levels in MGC nuclear and total extracts via western blots. As shown in Fig. 2(a), Glu receptor activation induces an accumulation in Oct-1 nuclear levels without any significant change in the overall Oct-1 content. In fact, we corroborated the lower level of Oct-1 in cytoplasmic extracts obtained from the same treated cells when we used EMSA for comparative detection of Oct-1 in nuclear versus cytoplasmic extracts (Fig. 2b). No changes were detected in the abundant and ubiquitous Sp-1 factor, which we have already reported as insensitive to Glu stimuli (Mendez et al. 2004).
As already mentioned, Oct-2 can scarcely be detected in MGC, and Glu exposure does not elevate Oct-2 levels either in nuclear or in total extracts (Fig. 2a). For BGC, the increase in Oct-1 DNA binding might result from nuclear Oct-1 accumulation as in MGC, because protein levels do not change after Glu treatment as we have already reported (Mendez et al. 2004). These results were corroborated by immunostaining experiments and are clearly shown in Fig. 2c. Note that these two POU proteins are not present in equal amounts in MGC and BGC. In both cell types, anti-Oct-1 antibodies clearly decorate the nuclei, and this signal is clearly enhanced by Glu. In contrast, anti-Oct-2 antibodies only generate a very lightly staining in MGC.
Oct-1 and Oct-2 have different transcriptional activities in Müller and Bergmann glial cells
Glu stimulation of MGC triggers multiple signaling responses through the several glutamatergic receptors expressed in these cells, which can eventually modulate a variety of transcriptional responses. We explored the transcriptional activity of Oct-1 and Oct-2 factors in both cell types using the construct pOCT3xCAT, obtained by cloning the Oct-1 oligonucleotide linked to a heterologous promoter specifically designed and constructed for detecting functional activity of these two factors (Fig. 3a). When confluent monolayers of MGC were exposed to Glu, a clear increase in reporter activity was found. Moreover, Oct-1 overexpression mimics the Glu effect while Oct-2.1 increased, and Oct-2.5 repressed, the construct activity (Fig. 3b). These results, together with those illustrated in Fig. 1, strongly suggest that Glu receptor activation is linked to Oct-1 activation and transcriptional control in MGC. On the other hand, in BGC, the construct was not responsive either to Glu or to the overexpression of Oct-1 or Oct-2 isoforms, marking a sharp contrast between these two radial glial cell types. Additional elements may be needed for Oct-2 to fully repress transcriptional activity under Glu treatment in BGC, as we have already reported for the chkbp promoter (Mendez et al. 2005).
It is clear then that in MGC, the response to Glu consists of an increased binding of Oct-1 due, at least in part, to a higher nuclear translocation of the factor after stimuli. In order to characterize the glutamatergic response in MGC, we analyzed the time dependence of the construct activity under a 1 mm Glu treatment. The results are shown in Fig. 4a. A sustained increase in reporter activity was obtained when the cells were exposed to Glu for up to 12 h. In order to identify the Glu receptors involved in the Oct-1 transcriptional response, we exposed confluent MGC to increasing Glu concentrations. An EC50 of 193.5 µm was obtained, demonstrating a receptor-mediated effect and suggesting the involvement of iGluRs (Fig. 4b). A pharmacological profile was obtained in additional transfection assays in which the Glu agonists, AMPA and KA, elicited a similar response in reporter activity to that obtained after Glu treatment. Therefore, we pre-incubated the cells with the specific antagonists for AMPA receptors, CNQX (100 µm) and DNQX (100 µm). As expected, both antagonists completely blocked the Glu response,as well as the activation elicited by AMPA and KA agonists (Fig. 4c). A role for mGluRs in this effect was ruled out after treatment with the mGluR agonists trans-ACPD, LAP-4 and DHPG, none of which modified the expression of the reporter gene (Fig. 4d).
Taken together, the results indicate that the POU family of transcription factors is differentially involved in Glu-dependent transcriptional control in radial glia cells.
Activity-dependent transcriptional control in neurons and glial cells is definitely involved in the plastic changes of the nervous system that underlie higher brain functions. Glu, the main excitatory transmitter, is involved in this process. In order to expand our current knowledge of the molecular mechanisms by which Glu controls gene expression at the transcriptional level in glial cells, we analyzed the DNA–protein interactions and the transcriptional activity of two members of the POU family of transcription factors, Oct-1 and Oct-2, in two subsets of radial glial cells: MGC and BGC. These transcriptional factors are of significant interest in the context of radial glial cells as we have previously shown that they participate in Glu-dependent gene expression regulation (Mendez et al. 2004). Moreover, Glu acting through NMDA receptors increases Oct-1 DNA binding and thereby represses gonadotropin-releasing hormone gene expression in a hypothalamic cell line (Belsham and Mellon 2000).
When confluent monolayers of both radial glia cultures were exposed to Glu or KA a clear increase in Oct-1 DNA binding was observed (Fig. 1a). This increased DNA binding is the result of a Glu-dependent increase in Oct-1 nuclear translocation. Western blot analyses, EMSA and immunostaining with anti-Oct-1 antibodies demonstrated that total levels of Oct-1 are not modified upon exposure to this amino acid transmitter (Fig. 2).
In sharp contrast, Oct-2 was barely detected in MGC and no increase in Oct-2 was detected after Glu treatment (Figs 1 and 2). It was notable that Oct-2 DNA binding, as well as its protein levels, were augmented after Glu treatment in BGC, results that are in line with our previous findings (Mendez et al. 2004). POU factors participate in a broad range of biological processes, ranging from housekeeping gene functions (Oct-1) to development of immune responses (Oct-1, Oct-2). Therefore, in order to perform a multitude of varied tasks, members of this family must rely on multi-level control mechanisms such as post-translational modifications, interactions with heterologous transcriptional regulators, changes in their oligomeric assemblies and flexible DNA binding, with various possible arrangements on the DNA (Sytina and Pankratova 2003).
An unexpected finding was that Oct-1 is fully responsive to Glu in MGC but not responsive at all, in terms of transcriptional regulation, in BGC (Fig. 3). This is particularly relevant because both cell types express Ca2+-permeable AMPA receptors that trigger similar signaling cascades (Lopez-Colome and Ortega 1997). A plausible explanation for this is the existence of a specific transcriptional co-factor that is only present in MGC. In support of this view is the fact that in B cells, the transcriptional activity of the Oct-1 is regulated by the lymphoid-specific co-activator, OBF-1 (Sauter and Matthias 1998; Chasman et al. 1999). In its dimeric complex form, Oct-1, can interact and synergize with this co-activator (Botquin et al. 1998). However, it has also been documented that the Oct-1 dimer bound to immunoglobulin heavy chain promoters (VH) fails to interact with OBF-1. Therefore, besides the presence of the co-factor, differential DNA–protein complexes can elicit differential transcriptional regulation. In the case of MGC, Glu signaling might be favoring the co-factor availability or the necessary modifications for protein–protein interactions compulsory for transcriptional regulation.
On the other hand, in BGC, a differential conformation of the DNA-bound factor might be responsible for the lack of the transcriptional response. Distinct transcription factor dimerization could be critically dependent on the sequence and spacing of the protein domain binding motifs of each DNA response element. In agreement with this interpretation, recent results have demonstrated that Oct-1 and Oct-2 can function as transcriptional repressors by recruiting, and physically interacting with members of the Groucho family of co-repressors that can discriminate between both Oct-1 and Oct-2 and the monomeric or dimeric nature of the POU/DNA complex (Malin et al. 2005).
The Glu-mediated increase in Oct-1 nuclear localization, DNA binding and activity is a receptor-mediated effect with a pharmacological profile corresponding to AMPA receptors (Fig. 4). One can be confident that mGluR receptors are not involved because the Glu effect is not mimicked by trans-ACPD, L-AP4 or DHPG. Furthermore, the Glu effect is sensitive to CNQX and DNQX and reproduced by AMPA or KA (Fig. 4).
In summary, we have provided evidence for a Glu-dependent AMPA-mediated increase in Oct-1 DNA binding and transcriptional activity in MGC. This effect is clearly cell specific as it is absent in another subtype of radial glia. Work is in progress in our laboratory aimed at establishing the signaling cascades that result in Oct-1 nuclear translocation, as well as identifying the genes that are transcriptionally regulated by Glu in an Oct-1-dependent manner in MGC.
This work was supported by grants from Conacyt to ELB (41273-A), ALC (42640-Q) and AO (43164-Q). The technical assistance of Clara Hernández-Kelly, Edith López and Blanca Ibarra is acknowledged.