The neocortex consists of histochemically, connectionally, and functionally distinguishable areas. Recently, molecular biological techniques have enabled us to find rare types of genes expressed in specific neocortical areas. We previously reported occ1 gene as preferentially expressed in the primary visual cortex (V1), using the differential display method. Here, by differential display, we found selective and strong expression of the serum retinol-binding protein (RBP) gene, in higher-order association areas. In V1, RBP mRNA was expressed only in the superficial part of layer II, but its expression increased, involving deeper layers, along the visual pathway. In visual association areas such as TE, RBP mRNA was strongly expressed in both supra- and infragranular layers. In primary auditory and somatosensory areas, as in V1, RBP expression was low, and restricted to the upper part of the supragranular layers. The laminar pattern of RBP expression is in marked contrast with that of occ1; and in early visual areas where both genes are expressed, these occur in distinct sublayers within the supragranular layers. In neonatal monkeys, the area-specific expression pattern of RBP was less distinct, suggesting that the characteristic expression of RBP in higher-order association areas is mainly established postnatally.
Recent advances in molecular biological techniques allow us to screen thousands of genes for differential expression; and it is evident that this tool provides a potentially important new approach for investigating the classic questions of cortical structure, function, and arealization (Hendry et al., 1984; Levitt, 1984; Arimatsu et al., 1994). Using the differential display method, our strategy has been to identify genes that were differentially expressed in neocortical areas of adult macaque monkeys, in order to compare differences in level of expression and laminar distribution across the systems.
By this method, we previously identified one gene, gdf7, which was enriched in the primary motor cortex (Watakabe et al., 2001) and another gene, occ1, which was preferentially expressed in V1 (Tochitani et al., 2001, 2003). Within V1, the expression of occ1 was concentrated in layer 4Cβ, which receives strong direct input from the lateral geniculate nucleus. Furthermore, we demonstrated that occ1 is expressed in an activity-dependent and a developmentally regulated manner (Tochitani et al., 2001, 2003). These results are consistent with the importance of thalamic projections in establishing the functional architecture of the primary sensory cortex, in addition to the geneticaly programmed mechanisms (Rakic, 1988; O'Leary, 1989; Katz and Shatz, 1996; Donoghue and Rakic, 1999a,b; Sestan et al., 2001; O'Leary and Nakagawa, 2002).
Here, we report that serum retinol-binding protein (RBP) is highly expressed in cerebral cortex, preferentially in association areas. RBP is abundantly synthesized in liver and is secreted as a complex with retinol and transthyretin into plasma (Goodman, 1980; Blaner, 1989), but its function and distribution in the cerebral cortex has not previously been recognized.
By in situ hybridization, our study indicated that, in association and limbic cortices, RBP gene was intensely expressed in layers II, III and V, mainly in excitatory neurons. In primary sensory areas, by contrast, the expression was low and restricted to the upper part of the supragranular layers. Inspection of the well-defined visual pathway showed that neurons expressing RBP mRNA became progressively more numerous with distance from V1, and gradually extended to the deeper part of layer III and infragranular layers. The laminar distribution of RBP expression was complementary to that of occ1.
We also compared the expression of RBP with the distribution of thalamocortical terminations, and found a reverse correlation in the early sensory areas. The influence of thalamic projections is thus likely to be more indirect than for occ1. We conclude that RBP is a useful marker gene for studying the function and formation of association areas and their related subcortical structures. With further characterization, it may provide clues as to area and laminar specializations of the primate neocortex.
Materials and Methods
Two young adult crab-eating monkeys (Macaca fascicularis) were used for the differential display screening. Three young adult Japanese monkeys (Macaca fuscata) were used for histological analysis (body weight 2.9–4.5 kg). Three neonatal macaques (postnatal day 1, n =2 and postnatal day 2, n = 1) were used as previously reported (Tochitani et al., 2003). Frozen sections from two young adult animals, monocular-deprived by TTX-injection for 7 or 14 days, previously reported (Tochitani et al., 2001), were available for processing. All experiments followed the animal care guidelines of Okazaki National Research Institute, Japan, and the NIH, USA.
Five neocortical areas were sampled for the fluorescent differential display–polymerase chain reaction (PCR; Fig. 1A). The abbreviations for the neocortical areas are according to Von Bonin and Bailey (1947): FDΔ, both banks of the principal sulcus including areas 9 and 46; FA, primary motor cortex in the anterior bank of the central sulcus; PC, primary somatosensory area, in the posterior bank of the central sulcus; TE, visual association cortex, located in the superior temporal sulcus (middle of the ventral bank); OC, primary visual cortex, posterior to the lunate sulcus.
Total RNA was extracted by acid guanidinium thiocyanate–phenol–chloroform extraction method (Chomczynski and Sacchi, 1987). Reverse transcription was carried out following the protocol from the supplier (GIBCO BRL/Invitrogen: Life Technologies, San Diego, CA). The fluorescent differential display with a carboxy-X-rhodamine (ROX)-labeled anchor primer was performed as previously reported (Yoshikawa et al., 1998; Tochitani et al., 2001; Watakabe et al., 2001). The ROX-labeled anchor primer was obtained from Takara (Otsu, Shiga, Japan): 5′-AAGCTTTTTTTTTTT(C, G or A)-3′. Thirty-seven oligonucleotides of 13-mer were synthesized (GenHunter Corporation, Nashville, TN; GIBCO BRL/Invitrogen: Life Technologies, San Diego, CA). Each 13-mer oligonucleotide was used for an arbitrary primer with the anchor primer. A target band was reamplified by the same PCR primer set and inserted into the Eco RV site of pBluescript II (KS+) vector (Stratagene, La Jolla, CA) for cloning of the DNA fragment (the primer set of 5′-AAGCTTTTTTTTTTTG-3′ and 5′-AAGCTTGATCGTC-3′ was used for RBP cDNA cloning).
PCR Primers and Probe Synthesis
All PCR primers except for occ1 were designed based on the human sequence database. The DNA fragments were produced by RT (reverse transcription)–PCR from African green monkey cDNA: (1) monkey RBP gene (corresponding to GenBank X00129, nt 133–862); (2) monkey glutamate decarboxylase 1, 67kDa (GAD67; corresponding to GenBank BC037780, nt 422–1051 ); (3) monkey vesicular glutamate transporter 1 (VGluT1; corresponding to GenBank: AB032436 nt 204–1093 ). PCR fragments were ligated into Eco RV site of the pBluescript II (KS+) vector and transfected into E. coli. The plasmids were extracted and linearized by Eco RI or Sal I before being used for the template of antisense or sense probes. The occ1 probe was derived from GenBank AB039661 (Tochitani et al., 2001). The digoxigenin (DIG)-UTP labeling kit (Roche Diagnostics, Basel, Switzerland) was used for in situ hybridization. The fluorescein (FITC)-UTP labeling kit (Roche Diagnostics) was also used to produce RNA probe for double in situ hybridization. RT-PCR for tissue from the five neocortical areas was performed by using RBP primer sets as above and African green monkey cDNA in order to confirm the differential display result.
Monkeys were deeply anesthetized with nembutal (60 mg/kg body wt, i.p. injection) following ketamine pretreatment (16 mg/kg body wt, i.m. injection) and perfused through the heart with warmed 0.9% NaCl containing 2 U/ml heparin followed by ice cold 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). The brains were postfixed for 4–6 h at room temperature and then cryoprotected in 30% sucrose in 0.1 M phosphate buffer at 4°C. Sections at 40 μm thickness were sliced from a frozen block of tissue. Serial sections were used for the thionin staining, parvalbumin immunostaining, in situ hybridization and histochemistry for cytochrome oxidase (Wong-Riley, 1979). in situ hybridization was performed for RBP and occ1 probes using adjacent sections.
Mouse monoclonal anti-parvalbumin antibody (Sigma-Aldrich, St Louis, MO) was used for parvalbumin immunostaining. The sections were preincubated for 1 h in 0.1 M phosphate buffer (pH 7.4) containing 0.3% Triton X-100 and 5% normal donkey serum, then incubated in a solution of the primary antibody (1:5000) in 0.1 M phosphate buffer, 1% normal donkey serum, and 0.3% Triton X-100 for 12 h at 4°C. The sections were then washed thoroughly and incubated in 0.1 M phosphate buffer, containing 1% normal donkey serum, and 0.3% Triton X-100, a donkey anti-mouse IgG antibody (diluted 1:500; Jackson Immunoresearch Laboratories, West Grove, PA) for 2 h at room temperature. The reaction was completed by the avidin-biotin-peroxidase method (Vectastain ABC reagents; Vector Laboratories, Burlingame, CA).
In Situ Hybridization
In situ hybridization was carried out as previously described (Liang et al., 2000). Briefly, free-floating sections were treated with 1–6 μg/ml proteinase K for 30 min at 37°C, acetylated, then incubated in hybridization buffer containing 0.5–1.0 μg/ml DIG-labeled riboprobes at 60°C. The sections were sequentially treated in 2 × SSC/50% formamide/0.1% N-lauroylsarcosine for 15 min at 55°C, two times; 30 min at 37°C in RNase buffer (10 mM Tris–HCl, pH 8.0, 1 mM EDTA, 500 mM NaCl) containing 20 μg/ml RNase A (Sigma-Aldrich); 15 min at 37°C in 2 × SSC/0.1% N-lauroylsarcosine, two times; 15 min at 37°C in 0.2 × SSC/0.1% N-lauroylsarcosine, two times. The hybridized probe was detected by alkaline phosphatase conjugated anti-DIG antibody using DIG nucleic acid detection kit (Roche Diagnostics). There were no apparent signals detected in control sections with the sense probes throughout the brain areas examined in macaque monkeys.
Double In Situ Hybridization
Sections (15 μm thickness) were used for double in situ hybridization. Pre-hybridization and washing were the same as the single hybridization protocol. Hybridization was performed at 65°C in buffer containing 500 ng/ml of DIG-labeled and FITC-labeled probes. The immunodetection by anti-DIG antibody conjugated with horseradish peroxidase (Roche Diagnostics) was enhanced with TSA Plus DNP System (PerkinElmer Life Sciences, Boston, MA). The fluorodetection was done by anti-DNP antibody conjugated with Alexa 488 (Amersham Biosciences, Piscataway, NJ). The FITC probe was detected by using Fast Red (Roche Diagnostics) after immunolabeling by anti-FITC antibody conjugated with alkaline phosphatase.
Isolation of RBP cDNA by Differential Display
We compared mRNA expression profiles of five distinct neocortical areas (Fig. 1A; see also Materials and Methods) by differential display–PCR. The differential display–PCR was simultaneously carried out using cDNA from two different monkeys (Fig. 1B). We found one band that was abundant in lanes of FDΔ (prefrontal association areas) and TE (visual association areas), moderate in FA (primary motor area), weak in PC (primary somatosensory areas) and very faint in OC (primary visual area). The DNA fragment derived from this band was purified and cloned. The sequence analysis revealed that the fragment was highly homologous to the 3′ region of human serum retinol-binding protein (RBP) sequence (GenBank accession No. X00129 nt 599–879, with the 10 bp gap sequence between nt 817 and 818). Based on this sequence, we designed the PCR primers; and the differential expression of RBP was confirmed by reverse transcription-PCR (data not shown).
RBP Expression in the Cortical Areas of Monkeys
Next, in order to examine the expression pattern of RBP in more detail, we performed in situ hybridization using the entire cerebral hemisphere. The in situ hybridization experiments basically confirmed the differential display results, showing high expression in association areas and low expression in primary sensory areas. The sites of high RBP expression generally corresponded to association areas as follows: the prefrontal higher-order association areas located in sections 1 and 2 of Figure 2 (e.g. square 3A in section 1 is enlarged in Fig. 3A), the dysgranular insular cortex located in the bottom of the anterior part of the lateral fissure (Fig. 2, sections 5–8, e.g. square 3C in section 7), the anterior part of the inferior temporal cortex which contains visual association (TE) and limbic areas (Fig. 2, sections 4–10; e.g. square 3I in section 7), and the dorsal bank of the superior temporal sulcus corresponding to the superior temporal polysensory area (STP; Fig. 2, sections 4–12; e.g. square 3B in section 7). In addition, the parietal association areas (PG) showed a high level of RBP expression (Fig. 2, sections 12–15; e.g. square 3E in section 12). Intense expression was also observed in the limbic cortices, for example the cingulate cortex and the entorhinal cortex (Fig. 2, sections 1–12; e.g. square 3D in section 7).
In these areas, strong RBP expression was observed in layers II, III and V (Fig. 3A–D). Although this pattern was typical of various association and limbic cortices, we noticed area-specific differences. For example, in PG and TE (Fig. 3E,I), RBP signals were more obvious in layer VI than in other association and limbic cortices (Fig. 3A–D).
In contrast, RBP expression was very weak and restricted to the upper part of the supragranular layers in primary sensory areas. Expression in V1 showed the lowest level of any cortical area (Fig. 3M). In V1, expression was mostly restricted to a narrow region at the top of layer II, with a few very weak signals in layer VI. Area 3b which is part of the primary somatosensory area also showed a low level of RBP expression (Fig. 3G). As in V1, the signals in the supragranular layers were mostly restricted to layer II, but weak signals were observed in infragranular layers. The primary auditory area, AI, showed similar laminar distributions as area 3b (Fig. 3H).
Along the sensory pathway progressing from the primary sensory toward the sensory association areas, we observed an increase of RBP expression in both intensity and width of laminar distribution. This was obvious along the well-characterized ventral visual pathway. RBP expression was restricted to the upper part of the supragranular layers in V2 (Fig. 3L), but extended to the deeper part of layer III in V4 (Fig. 3K). In TEO, RBP expression was observed in both supra- and infragranular layers although its expression in the infragranular layers was much weaker than that in the supragranular layers (Fig. 3J). In TE, as described above, intense signals extended throughout the supragranular layers; and the expression in the infragranular layers was much higher than that in TEO (Fig. 3I). The change in the laminar distribution within the supragranular layers was gradual and did not observe any sharp border between layers II and III (also see next section).
In the primary motor cortex (area 4), intense signals were located in the upper part of the supragranular layers (Fig. 3F). In addition, RBP signals in giant pyramidal cells were conspicuous in layer V. In the premotor areas, compared to the primary motor area, RBP expression extended more deeply into the supragranular layers and was more apparent in layer V (data not shown).
In summary, RBP expression was high in association and limbic cortices, with strong expression in layers II, III and V, and was low in primary sensory areas, with a restricted distribution in the supragranular layers. Furthermore, the laminar distribution increased from the primary areas toward association areas, as was most obvious in the visual pathway. Additionally, we noted scattered cells that showed relatively strong RBP expression in the white matter under every cortical area.
Selective Expression of RBP in other Brain Regions
High expression of RBP was observed in subcortical limbic structures and striatum (Fig. 4). In the amygdala, the dorsal intermediate subdivision of the lateral nucleus showed a relatively low level of RBP expression (Fig.4A), while this was relatively high in the central nucleus (compare Fig. 4A with Fig. 4B). The hippocampal formation also showed high RBP expression (Fig.4C). Although we observed weaker sense probe signals in the dentate gyrus, the intense labeling in this region is specific to the antisense probe of RBP mRNA. CA2, where there is occ1 expression (Tochitani et al., 2003), showed a relatively lower expression of RBP among the hippocampal subdivisions (Fig.4C). The caudate and putamen showed equal expression levels of RBP anteriorly (Fig.4D), but in the posterior part, expression in the caudate was higher than that of the putamen (Fig. 4E,F). In posterior striatum we also noticed that the ventral region of the putamen showed more intense expression than did the dorsal region (data not shown).
Choroid plexus is reported to express RBP mRNA in rodents, ruminants and carnivores (Duan and Schreiber, 1992; Aldred et al., 1995). We also detected some expression of RBP in the choroid plexus of monkeys, but the level was much fainter compared to the expression described here (data not shown).
Complementary Expression of RBP and occ1
Previously, we identified the occ1 gene that is preferentially expressed in area V1 (Tochitani et al., 2001). When we compared the expression of RBP with that of occ1 in the adjacent sections, the distribution of these two genes showed striking complementarity (Fig. 5A). occ1 expression appeared almost absent or negligible in the prefrontal cortex and anterior part of the temporal cortices where RBP expression was high (Fig. 5A and area 46 in Fig. 5B). In the posterior region, occ1 was expressed heavily in V1 and moderately in extrastriate areas, with gradual decrease toward the anterior region, along the dorsal and ventral visual pathways (Fig. 5A, inset). In adjacent sections, RBP mRNA by contrast was almost absent in V1, and gradually increased along both the visual pathways (Fig. 5A).
In these occipital regions, both dorsal and ventral, RBP and occ1 were expressed in separate sublayers. For example, in V2, occ1 expression was relatively strong in layer III, whereas RBP expression was mostly restricted to the uppermost part of the supragranular layers (Fig. 5D).
In the more anterior regions, occ1 expression gradually decreased in intensity and width, and was restricted to the very bottom of layer III (Fig. 5B, lower panels a-3–a-1). Conversely, the intensity of RBP signals increased (Fig. 5C) and the expression extended deeper into layer III (Fig. 5B, upper panels a-3–a-1). Although RBP and occ1 signals remained largely segregated in the upper and lower compartments of the supragranular layers, in a complementary fashion, our double in situ hybridization showed that, in the deeper part of layer III in V2, occ1 positive cells expressed RBP mRNA at a low level (data not shown).
In the infragranular layers of the posterior regions, we observed that a few intense signals of occ1 existed predominantly in layer V, and that the signals seemed invariant across areas (Fig. 5B, lower panels a-3–a-1). In contrast, we observed a gradual increase of RBP signals in the infragranular layers (Fig. 5B, a-3–a-1) in both layers V and VI from posterior (Fig. 4B, a-3) to anterior regions (Fig. 5B, a-1). We did not see clear laminar segregation between RBP and occ1 signals in the infragranular layers.
In addition to visual related areas, occ1 is shown to be preferentially expressed in the middle layers of AI and area 3b compared with the surrounding areas (Tochitani et al., 2001). In these sensory areas, we observed complementarity in the layer distribution between RBP and occ1 signals. Consistent with complementary expression in other areas, occ1 showed much weaker expression in AI and SI, than in V1, whereas RBP expression was higher in AI and area 3b than in V1 (Fig. 3G,H,M).
Complementary Expression of RBP and Thalamic Terminations in Primary Sensory Areas
We previously found that occ1 expression in V1 is dependent on retinal activity and thalamocortical connections (Tochitani et al., 2001, 2003). To investigate the relationship between RBP expression and the thalamocortical projection, we compared the RBP expression around primary sensory areas, V1, AI and SI, with two other markers. One was parvalbumin immunoreactivity (PV-IR), which demonstrates dense thalamic terminals in the middle layers of sensory cortex (Steriade et al., 1997). The other was cytochrome oxidase (CO) whose activity is high in primary sensory cortices (Wong-Riley, 1989; Morel et al., 1993; Jones et al., 1995, 2002) and which is related to sensory input from specific thalamic nuclei in V1 and SI (Wong-Riley, 1989, Jones et al., 2002). Since a subset of cortical interneurons also expresses PV, we aimed to distinguish the thalamic fibers from the intrinsic PV-IR by using a short incubation and reaction time. We felt that this approach was successful in delimiting the primary auditory and somatosensory areas, and our data are limited to these regions.
Within somatosensory areas, area 3b and area 1 have both been considered as primary areas. Area 3b received denser PV-IR fibers in the middle layers and exhibited higher CO activity than area 1 (Fig. 6B,C,D). In comparison, RBP expression was lower and more restricted in layers in area 3b than in area 1 (Fig. 6A). We observed a similar relationship from AI to secondary auditory cortex; namely, a decrease of PV-IR and CO staining, and an increase of RBP mRNA (Fig. 6F,G,H). In both these regions, the laminar expression of RBP was complementary to that of the PV-IR plexus in the middle layers (Fig. 6F,G). Thus, the laminar distribution of RBP mRNA seemed to be reversely correlated to that of PV-IR in both auditory and somatosensory cortices. We also observed abundant PV-IR fibers in the middle layers of other areas; for example, area 5 (Fig. 6E) and area 9 (data not shown). However, a complementarity to RBP signals was not clear in these higher-order sensory and association areas.
RBP expression is complementary with PV-IR fibers and occ1 expression in the early sensory areas. In view of this result, we wanted to investigate the relationship between RBP expression and thalamic activity, as this is known to be important for occ1 expression. To test this possibility, we examined RBP expression in V1 of monocularly deprived monkeys which were repeatedly injected with TTX into one eye for 7 or 14 days. As we reported previously, monocular deprivation leads to a clear reduction of occ1 in the ocular dominance columns of the deprived eye. This did not occur for RBP expression which, in contrast, was restricted to a narrow layer at the top of layer II in both deprived and non-deprived columns (data not shown). We observed little difference if any in this thin layer or in any other layer in V1. Thus, the low level of RBP expression in V1 did not seem to be attributable to an influence of thalamic activity.
Cell Types that Express RBP in Supragranular Layers
In order to examine the cell types that express RBP mRNA, we performed double in situ hybridization using cell-type specific markers. We used VGluT1 as a glutamatergic neuronal marker (Fujiyama et al., 2001) and GAD 67 as a GABAergic neuronal marker. In our experimental condition, these two probes labeled distinct populations (data not shown). Here we illustrate examples from two different areas (TE and area 1), in the same section, where the signals were processed exactly in the same condition. In TE, the majority of RBP signals within supragranular layers colocalized with VGluT1 mRNA signals (Fig. 7A); but only a restricted population of GAD67 mRNA positive cells expressed RBP signals (Fig. 7B). This bias was also observed in area 1 where the band of RBP mRNA positive neurons was thinner than in TE: most of the RBP mRNA positive cells were VGluT1 mRNA positive (Fig. 7C), while only some were GAD67 mRNA positive (Fig. 7D). We confirmed essentially the same results in V1, V2, STP and area 4 as well (data not shown).
RBP Expression in Neonatal Monkey
Our next question was whether the graded pattern of RBP expression along the anteroposterior axis might already be present during development (Donoghue and Rakic, 1999a,b) and retained throughout maturation. Thus, we examined the expression pattern of RBP in newborn monkeys (P1 and P2 monkeys) and compared it with that of the adult.
In neonatal monkeys, the V1–V2 border was clearly visible because the RBP expression in V1 was lower than that in V2 (Fig. 8A). However, RBP expression in both V1 and V2 of neonatal monkeys was broader than that in the adult monkey (Fig. 8A, upper panels). Intense RBP signals in V2 were widely distributed within supragranular layers, including the deeper part of layer III, whereas RBP signals were almost restricted to layer II in the adult. Similarly, in V1, the signal, even though weaker than in V2, extended beyond layer II. In contrast with RBP expression, however, the occ1 expression was low in V1 and was not observed in V2 of newborn monkeys (Fig. 8A, lower panels), as previously reported (Tochitani et al., 2003).
In the prefrontal cortex (area 11) of neonatal monkeys, RBP signals were mostly observed in layers II and III, and only weakly expressed in infragranular layers (Fig. 8B, P1). This is in contrast with adult monkeys, where the RBP expression was widely present in layers II, III and V (Fig. 8B, adult). In newborns, the layer distribution of RBP signals in area 11 was similar to that in area V2; in other words, without the characteristic adult regional gradation. These results suggested that the graded pattern of RBP expression in adult monkeys is mainly established during postnatal development by processes that enhance its laminar distribution in higher-order association areas but diminish this in early sensory areas.
Using the differential display method, we found that the RBP gene was differentially expressed in neocortical areas, being strongly expressed in higher-order association and limbic areas and only slightly in primary sensory areas. This differential mapping reinforces the traditional distinction between sensory and association areas, from a new perspective which may allow a further means for investigating mechanisms of area specialization. The differential expression levels among cortical areas were also associated with differences in laminar distributions. Beyond the primary sensory areas, there was a gradual recruitment of neurons through the supragranular and then infragranular layers.
In the early sensory areas, RBP expression was complementary to that of occ1 and to PV-IR fibers in the middle layers. From this, we infer that it does not seem to be directly related to the thalamocortical projection. Consistent with this, its distribution in area V1 is not sensitive to monocular deprivation. Finally, the area-specificity of RBP expression is likely to result from maturational processes during the postnatal period. This is suggested by the developmental shift in laminar distributions. In neonates, RBP mRNA is distributed in a similar laminar pattern in higher-order association and early visual areas, but this is not the case in adults.
In addition, to our knowledge, this is the first report of RBP mRNA expression in characterized neuronal populations in the primate central nervous system (CNS).
Possible Relations of Retinoic Acid and RBP in the Brain
Vitamin A (retinol) is dietarily obtained as β-carotene in plants and retinyl esters in animal tissues and stored in liver (Goodman, 1984; Malik et al., 2000a). For delivery to peripheral tissues, vitamin A is mobilized from storage and effectively transported into the plasma as retinol bound to RBP (Ong, 1994; Malik et al., 2000a). Once retinol is received by the cell, alcohol dehydrogenase can catalyze the oxidation of retinol into retinaldehyde, abbreviated retinal (Duester, 2000, 2001). Retinal is used for visual pigments (Applebury and Hargrave, 1986) and as the substrate of retinal aldehyde dehydrogenase (RALDH) which irreversibly converts retinal into retinoic acid (RA) (Duester, 2000, 2001). Both ADH and RALDH are expressed in neural tissues (Wagner et al., 2002; Galter et al., 2003).
RA is a biologically active metabolite of retinoid, and transduces its effect through the RA receptor (RAR α, β and γ) and the retinoid X receptor (RXR α, β and γ) which act as RA-dependent transcriptional regulators. The binding of these receptors to the RA response element (RARE) in promoter regions activates a large number of gene transcriptions (Gudas et al., 1994; Clagett-Dame and Plum, 1997). Thus, RA is important for morphogenesis and differentiation in a number of tissues including the CNS; for example, for hindbrain segmentation (Gavalas, 2002) and possibly for regionalization of the forebrain (LaMantia et al., 1993; Smith et al., 2001).
While these reports focus on embryonic development, a recent study suggests that RALDH is transiently expressed even in the postnatal brain (Wagner et al., 2002). In mice, RALDH3, one of the isotypes of RALDH, is transiently expressed in the nuclear accumbens, the olfactory bulb and the cerebral cortex during the early postnatal period (Wagner et al., 2002). Interestingly, RALDH3 expression in the cerebral cortex resembles the RBP expression in neonatal monkeys. That is, RALDH3 is differentially expressed in layers II and III of cortical regions; and its expression is high in V2 (both medial and lateral regions), but appears to be absent in V1. Since RBP has the RARE sequence in the promoter region and its transcription could be strongly regulated by RA (Mourey et al., 1994; Panariello et al., 1996; Clagett-Dame and Plum, 1997), RBP mRNA expression could depend on RA activity.
The postnatal remodeling of the cortex is considered to be necessary in order for neuronal circuits in the brain to adapt to sensory stimuli (Rakic, 1988; Batardiere et al., 2002). The significant differences in expression level among early postnatal cortical regions suggests that RA transcriptional activity may play a role in the postnatal differentiation of higher-order sensory and association cortices, in response to environmental inputs.
The characteristic RBP mRNA distribution shown in this study overlaps some of adult brain regions in which RA transcriptional activity may be involved in neuronal plasticity. One such region is the hippocampus where RBP mRNA expression is high. In rodents, RAR and RXR are highly expressed in the hippocampus (Krezel et al., 1999; Zetterstrom et al., 1999); and RALDH 1 and 2 are expressed in hippocampal meninges. Vitamin A deficiency in adult rodents lowers or eliminates synaptic plasticity of CA1 (Misner et al., 2001) and results in a relational memory deficit (Cocco et al., 2002; Etchamendy et al., 2003).
The activity of RA in the mature CNS is more directly shown using RA-reporter mice (Haskell GT et al., 2002). A neuronal population presumably activated by RA can be visualized in transgenic mice that are introduced with lacZ gene fused to minimally required promoter at the down stream of five direct repeats of the RARE sequence. These transgene-expressing cells are located in the dorsal horn of the spinal cord, the olfactory bulb, habenular complex, the amygdala and cerebral cortex. In our material, RBP expression in the spinal cord and the olfactory bulb remains to be studied, but we confirmed RBP expression in the amygdala (Fig. 4A) and the lateral habenular nucleus (data not shown).
It should be noted that there is some discrepancy between the RA-active cells in mice and the RBP-mRNA-positive cells in monkeys. The study using the RA-reporter mice indicated that the RA active cells are fusiform or possibly stellate cells in the neocortex (Haskell et al., 2002); but our in situ hybridization results showed that RBP expression is usually observed in pyramidal cells in the primate. Future studies will be needed to determine what accounts for the discrepancy.
Roles of RBP and RA in the Brain Suggested by Pathological Studies
Human neuropathologies provide further evidence that the RBP-mediated RA synthesis pathway may play a role in maintaining higher brain function. Possible disease links include Alzheimer's disease (see below), schizophrenia (see below), and fronto temporal dementia (Davidsson et al., 2002a). In Alzheimer's disease patients, RBP secretion into cerebrospinal fluid is reduced (Davidsson et al., 2002b; Puchades et al., 2003) and its immunoreactivity is observed in amyloid extracts (Maury and Teppo, 1987). In addition, the human RBP locus in chromosome 10q24 is near to a marker sequence whose recombination correlates with Alzheimer's disease onset (Bertram et al., 2000; Goodman and Pardee, 2003). Further analysis of the loci of retinoid-related genes and those related to late-onset Alzheimer's disease suggests that retinoid metabolism is implicated in the disease (Goodman and Pardee, 2003). RA synthesis from retinaldehyde is higher in the hippocampus and parietal cortices of the Alzheimer's disease cohort than in controls (Conner and Sidell, 1997).
The pathology of Alzheimer's disease suggests a further relationship with RA. That is, Alzheimer's disease involves the degeneration of cholinergic neurons whose differentiation and maintenance can be induced by RA. RA promotes expression of choline acetyltransferase (ChAT) and the vesicular acetylcholine transporter, ChAT activity and intracellular ACh concentration in vitro studies (Berse and Blusztajn, 1995; Pedersen et al., 1995; Malik et al., 2000b). The cholinergic neurons and innervations that are affected in Alzheimer's disease patients are distributed in basal forebrain, limbic and association cortex (De Lacalle and Saper, 1997; Dickson, 1997; Wisniewski et al., 1997; Clippingdale et al., 2001; Selkoe, 2001; Kar and Quirion, 2004). Given the apparent expression of RBP in these areas (Fig. 2), cell sorting by using RBP expression may provide further linkage of retinoid metabolism with Alzheimer's disease and cholinergic subpopulations.
The relationship between schizophrenia and RA is somewhat controversial. Macroarray analysis suggests that there is no significant difference in the expression of genes related to retinoid metabolism (Middleton et al., 2002). However, three independent lines of evidence suggest some involvement of RA in schizophrenia (Goodman, 1998; Citver et al., 2002): (i) congenital anomalies similar to those caused by retinoid dysfunction are found in schizophrenics and their relatives; (ii) those loci suggested as linked to schizophrenia are also the loci of the genes of the retinoid cascade (convergent loci); and (iii) transcription of the dopamine D2 receptor and numerous schizophrenia candidate genes is regulated by RA. Assuming there to be a high expression of RBP mRNA in human association cortices, disruption of RBP-mediated retinoid metabolism might specifically damage neuronal circuits of prefrontal and other association and limbic cortices due to the strong effects of RA transcriptional activity (Weiberger and Berman, 1998; Buchsbaum et al., 2002; Kurachi, 2003).
Implication of RBP Distribution in the Cerebral Cortex
A striking characteristic of RBP expression in the supragranular layers is its expansion from early visual areas (V1 and V2) to higher visual areas. The interpretation of this pattern is not clear; but the concentration of RBP-mRNA-positive neurons in the upper layers might suggest a preferential association with corticocortical projections.
Layer II in visual and other areas is one source of feedback projections; and the dendrites of layer II neurons are among the likely targets of feedback and other terminations layer I (Rockland and Pandya, 1979; Felleman and Van Essen, 1991). In this regard, it is interesting that the laminar distribution of both feedback terminations and cells of origin broadens in anterior areas (i.e. TEO to V4; Rockland et al., 1994, fig. 21), a broadening which parallels the shift in laminar expression of RBP. In addition, layer II neurons are involved in horizontal intrinsic connections (Lund et al., 1981; Yoshioka et al., 1992; Fujita and Fujita, 1996). These connectivity features are suggestive of some involvement in integrative processing.
A distinctive role for layer II is suggested by recent electrophysiological results of Shipp and Zeki (2002). That is, their study indicates that, in each type of CO stripe (thick, thin and pale), the specificity of cell response for visual stimuli was to some degree laminar-specific; and in particular, the responses in layer II were more generalized, in terms of directional, orientation, and spectral sensitivity, than those deeper in layer III within V2.
The laminar expansion of RBP expression occurs gradient-wise, seemingly across area borders. Other features, such as pyramidal neuron dendritic architecture, are known to systematically change along the visual and somatosensory pathway. Increases in the size of the basal dendritic field, in the degree of branching, and in the spine density have been demonstrated by intracellular injection of Lucifer yellow, and interpreted as signifying differential integrative capacity of neurons in early and higher-order visual areas (Elston et al., 1999; Elston, 2001). Further morphological elaboration is reported for neurons in the prefrontal cortex (Elston, 2000; Elston et al., 2001), a region with strong RBP expression.
In the amygdala complex, which is heavily connected with cortical areas, RBP mRNA is expressed at low levels in the lateral nuclei and higher levels in the accessory and central nuclei (Fig. 4). These nuclei are organized in a largely unidirectional intrinsic circuit and have distinctive connections with cortical areas: the lateral nucleus is preferentially innervated by visual cortices and projects to basal accessory nuclei (Stefanacci and Amaral, 2002). The basal accessory nucleus is densely interconnected with frontal and limbic cortices and projects to the central nuclei (Stefanacci and Amaral, 2002). Amaral and his coworkers hypothesize that the sequential processing of these nuclei is a substrate for the contextual fear response for novel or potentially dangerous objects (Stefanacci and Amaral, 2002). An intriguing possibility is that RBP mRNA selective expression in the amygdala is related to an anatomical and functional hierarchy within the amygdala and its connectivity with cortical areas.
RBP mRNA is preferentially distributed in the caudate and ventral putamen in the posterior striatum (Fig. 4). The caudate and ventral putamen are connected to higher-order and visual association areas (Alexander et al., 1986; Webster et al., 1993; Yeterian and Pandya, 1994), whereas the dorsal putamen is more directly connected to the motor cortex (Alexander et al., 1986).
In summary, the expression of RBP mRNA in primate cortical structures, as well as its complementary expression with occ1 and PV-IR in the cerebral cortex, suggest that RBP mRNA is preferentially distributed in association areas and related subcortical structures rather than the early sensory and motor areas. This further suggests that RBP and RA may play important roles in selectively influencing executive functions and social behavior in primates. We hope that further studies on the function of RBP and retinoids in the brain may provide clues to the organization of association areas at the molecular level.
We thank Drs Fumiko Ono of the Corporation for Production and Research of Laboratory Primates, Keiji Terao, Tsukuba Primate Center, National Institute of Infectious Diseases, H. Horie, S. Abe and S. Hashizume of the Japan Poliomyelitis Research Institute for supplying monkey tissues, and Kathleen Rockland for critical reading and valuable discussions. This research was support by a Grant-in-Aid for Scientific Research on Priority Areas (A) from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to T.Y.).
1Division of Speciation Mechanisms 1, National Institute for Basic Biology, Aichi 444-8585, Japan and 2Laboratory for Neural Architecture, Brain Science Institute, RIKEN, Wako, Saitama 351-0198, Japan