Introduction

Retinol-binding protein 4 (RBP4) is a specific carrier for retinol in the blood. In hepatocytes, newly synthesized RBP4 associates with retinol and transthyretin (TTR) in the endoplasmic reticulum and is then secreted into the blood.1, 2, 3, 4, 5 Thus, RBP4 is essential for the mobilization of retinol from liver stores. Binding to TTR (a tetramer of MW 55 000) prevents renal filtration of RBP4 (MW 21 000).2

In humans, genome-wide association studies revealed new susceptibility loci for eye diseases, such as refractive error including myopia.6, 7, 8 Over 450 inbred strains of mice have been described,9 providing a wealth of different genotypes and phenotypes for studying human diseases. Keane et al10 reported the genome sequences of 17 inbred strains of laboratory mice and identified almost 10 times more variants than was previously known. Many of them result in significant tissue-specific expression bias and phenotypic variation.10, 11 C57BL/6J mice are susceptible to high-fat diet-induced type 2 diabetes.12 JF1 mice are especially sensitive to high-fat diet-induced diabetes and obesity, whereas MSM/Ms mice are resistant.13 We14 showed that the severity of acute pancreatitis experimentally induced by cerulein was highest in C3H/HeJ and CBA/J, moderate in BALB/c, and mildest in C57BL/6J and JF1 mice.

To analyze the relationship between sequence variations and phenotypic variations, it is essential to create transgenic or knockout mice with a defined genetic background. So far, Quadro et al15 generated Rbp4-deficient mice by using 129/Sv embryonic stem (ES) cells and mated these chimeric mice with C57BL/6J to obtain Rbp4-deficient mice in a mixed genetic background (129 × C57BL/6J). Although these mice showed impaired visual function, it progressively improved by 24 weeks of age.

Here, we created targeted null Rbp4 allele mice in the C57BL/6 background and showed that these Rbp4−/− mice had much more severe ocular phenotypes in terms of impaired visual function, and structural and developmental defects.

MATERIALS AND METHODS

Construction of Targeting Vectors and Isolation of Targeted ES Cells

The construction of targeting vectors and isolation of targeted ES cells were carried out as described previously,16 except for the use of a 2.8-kb 5′ Rbp4 homologous region upstream of the ATG codon and a 6.7-kb 3′ Rbp4 homologous region downstream of the ATG codon (Figure 1).

Figure 1
figure 1

Targeting vector construction. Homologous recombination between the wild-type allele and the targeting vector produced a targeted null allele carrying the PGK-neo gene flanked by lox71 and loxP. MC1-DT-A-pA: a diphtheria toxin A (DT-A) fragment with an MC1 promoter.

Full size image

Southern Blot Analyses for Isolation of Targeted ES Cell Clones

So far, all ES cell lines have been established from C57BL/6N mice, but not from C57BL/6J. Mattapallil et al17 showed that a series of commercially available ES cells of C57BL/6N origin and C57BL/6N mouse lines used to generate ES cells contained the rd8 mutation in the Crb1 gene. This mutation results in a form of retinal dysplasia and degeneration.3 To establish an Rbp4-deficient mouse in the C57BL/6 background, we used TT2 ES cells originated from an F1 embryo between C57BL/6N and CBA/JN strains.18

Targeted ES cell clones were confirmed by Southern blot analyses as described previously.16 For detection of homologous recombination in the 5′ region, DNA from ES cells was digested with XhoI/SalI or MunI and the targeted allele was detected with a neo probe and a 5′ probe; for detection of homologous recombination in the 3′ region, AgeI/HpaI or SacII/HpaI digestion and a neo probe and 3′ probe were used, respectively (Supplementary Figure S1).

Generation of Rbp4-Knockout Mice

Chimeric mice were produced by aggregation of ES cells with eight-cell embryos from ICR mice as described previously.16 ICR originated from the CD1 strain at the Institute of Cancer Research (Philadelphia, PA, USA). ICR is an albino mouse kept in closed colony at Clea (Tokyo, Japan; http://www.clea-japan.com/en/index.html). To exclude the rd8 mutation, chimeric male mice were mated to C57BL/6J females (Clea) to obtain the Rbp4−/+ offspring. These F1 mice were backcrossed to C57BL/6J for a further six generations. At the second and sixth generation, we carried out marker-assisted breeding by using 123 polymorphic markers (111 markers on autosomes, 8 markers on X chromosome and 4 markers on Y chromosome). These markers can distinguish between C57BL/6 and CBA/JN alleles. These polymorphic markers were distributed evenly throughout the genome and detected by using a high-speed genetic profiling system (384-well plate-based TaqMan assay method) (Japan Mouse Clinic; http://ja.brc.riken.jp/lab/jmc/mouse_clinic/en/assistive/index.html) and mice with C57BL/6 alleles at most loci were selected for breeding. All mice used in our experiments were at the seventh generation and had C57BL/6-homozygous alleles at all 123 polymorphic loci. Mice were fed a CE-2 diet (Clea), which contained 1515 IU retinol/100 g (0.454 mg/100 g). To avoid possible sex-related differences, only males were used in all analyses. All experiments were performed in accordance with the Declaration of Helsinki and were approved by the Kumamoto University Ethics Committee for Animal Experiments.

Northern Blot, Semiquantitative RT-PCR, and qRT-PCR Analyses

Northern blot and semiquantitative reverse transcription-polymerase chain reaction (RT-PCR) analyses were carried out as described by Zhao et al16 and Li et al,19 respectively. The nucleotide sequence identity between the mouse and human Rbp4 coding regions is ~87%. The primer set suitable for human and mouse RBP4 coding regions was as follows: sense (mhRBP4-F), 5′-GAGTCAAGGAGAACTTCGAC-3′; antisense (mhRBP4-e4R), 5′-CAGTACTTCATCTTGAACTTGG-3′; the PCR product was expected to be 244-bp long. The primer set used for detection of the Hprt mRNA was as follows: sense, 5′-CACAGGACTAGAACACCTGC-3′; antisense, 5′-GCTGGTGAAAAGGACCTCT-3′; the PCR product was expected to be 238-bp long. PCR products were separated on 1.5% agarose gels and stained with ethidium bromide. Quantitative RT-PCR (qRT-PCR) was performed with an Applied Biosystems 7500 Real-Time PCR system (Applied Biosystems, Foster City, CA, USA). Reactions were carried out under the following conditions: 1 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 60 °C. Quantified values were normalized against the input determined by expression of the housekeeping gene Hprt.

Western Blot Analysis

Mouse serum (6 μl diluted 1:50 in 0.9% NaCl per lane) was separated in a 12% polyacrylamide gel and transferred to an Immobilon polyvinylidene difluoride membrane (Millipore, Billerica, MA, USA). The following primary antibodies were used: rabbit anti-human RBP4 antibody (1:1000; PTG, Chicago, IL, USA) and rabbit anti-bovine serum albumin antibody (1:1500; Upstate, New York, NY, USA). Anti-rabbit immunoglobulin G conjugated with horseradish peroxidase (Amersham Biosciences, Piscataway, NJ, USA) was used for detection. Band intensities were quantified by densitometry and the ImageJ ver. 1.38 software (http://rsb.info.nih.gov/ij/docs/index.html).

Functional Analysis of the Retina

Electroretinograms (ERGs) were performed as described previously.20 Left eyes of Rbp4+/+ and Rbp4−/− mice at 7, 14, 24, and 40 weeks of age were used. Mice were dark-adapted overnight and anesthetized with a mixture of ketamine (80 mg/kg; Daiichi Sankyo Propharma, Tokyo, Japan) and xylazine (16 mg/kg; Bayer Medical, Tokyo, Japan) under dim red light. Pupils were dilated with Mydrin-P ophthalmic solution (1:5 dilution) and corneas were anesthetized with 0.4% oxybuprocaine hydrochloride (Santen Pharmaceutical, Tokyo, Japan). Stimulus presentation was programmed and data acquisition was performed with the data recording and analysis software (Scope v.3.7; ADInstruments, Bella Vista, Australia). Scotopic ERG responses to single flash presentations (10-ms duration) at three increasing intensities (0.4, 1.2, and 2.0 log cd s/m2) were recorded. Triplicate responses were computer-averaged for each flash condition. The time intervals between flashes were 10, 30, and 60 s for intensities of 0.4, 1.2, and 2.0 log cd s/m2, respectively. The amplitude and implicit time of the a-wave were measured from the prestimulus baseline to the a-wave trough. The amplitude and implicit time of the b-wave were measured from the a-wave trough to the peak of the b-wave or, if the a-wave was not measurable, from the prestimulus baseline.

Retinal Histology

After the ERG was performed, mice (40 weeks of age) were anesthetized with an intraperitoneal injection of pentobarbital sodium (90 mg/kg; Kyoritsu Seiyaku, Tokyo, Japan). Left eyes were enucleated and fixed overnight in 2.0% paraformaldehyde/2.0% glutaraldehyde in 0.1 M cacodylic acid, pH 7.4. The eye was dehydrated by immersion in a series of increasing ethanol concentrations. After three changes of xylene, the eyes were embedded in paraffin wax. A mid-horizontal series of sections (3-μm-thick) through the optic nerve head was prepared. Sections were stained with hematoxylin and eosin (Muto Pure Chemicals, Tokyo, Japan) and viewed with an AX80 microscope (Olympus, Tokyo, Japan).

Measurement of Retinal Properties

The number of cells in the ganglion cell layer and the thickness of the inner nuclear layer (INL), outer nuclear layer (ONL), and photoreceptor outer segment layer (OSL) were measured by using light microscopy. The measurements were performed at the points where the ratios of the distance from the optic disc to the point to the distance from the optic disc to peripheral end of the retina were 0.2 and 0.8.20, 21 The number of cells in the ganglion cell layer was estimated at the central retina (220 μm from the optic nerve head) and peripheral retina (220 μm from the peripheral edge).21 The distance from the retinal surface to the end of the photoreceptor outer segments was measured as the retinal thickness. The distance from the peripheral edge of the retina to the central end was measured as the distance from the central to peripheral retina. The values recorded for each parameter for the nasal and temporal sides were averaged to obtain a single value for each eye. For consistency, only retinal sections with optic nerve stumps were used.

Transmission Electron Microscopy

Mice at 40 weeks of age were used for this study. The eyes were fixed as described above at 4 °C overnight, washed three times with 0.1 M cacodylate buffer for 30 min each, and postfixed with 2% osmium tetroxide in 0.1 M cacodylate buffer at 4 °C for 2 h. The samples were then dehydrated in a graded ethanol series: 50 and 70% for 30 min each at 4 °C, 90% for 30 min at room temperature, three changes of 100% for 30 min each at room temperature, and 100% ethanol overnight at room temperature. The samples were infiltrated with propylene oxide (PO) two times for 30 min each and were placed into a 70:30 mixture of PO and Quetol-812 resin (Nisshin EM, Tokyo, Japan) for 1 h; the tube was kept open overnight to evaporate PO. The samples were transferred to fresh 100% resin and polymerized at 60 °C for 48 h. Ultrathin sections (70 nm) were cut with a diamond knife using an ultramicrotome (Ultracut UCT; Leica, Vienna, Austria) and mounted on copper grids. They were stained with 2% uranyl acetate at room temperature for 15 min, washed with distilled water, and stained with lead stain solution (Sigma-Aldrich, Tokyo, Japan) at room temperature for 3 min. The grids were examined under a transmission electron microscope (JEM-1400Plus; JEOL, Tokyo, Japan) at an acceleration voltage of 80 kV. Digital images (2048 × 2048 pixels) were taken with a CCD camera (Veleta; Olympus Soft lmaging Solutions GmbH, Münster, Germany).

Funduscopic Examination

Pupils were dilated with 0.5% tropicamide and 0.5% phenylephrine (Santen Pharmaceutical, Osaka, Japan) diluted 5- to 7.5-fold with physiological saline. Fundi were examined by indirect ophthalmoscopy and photographs were taken with a Genesis hand-held funduscope camera (Kowa, Nagoya, Japan) with a Volk 90D condensing lens (Volk Optical, Mentor, OH, USA) as described previously.20 The camera was fitted to a dissection microscope base for stability and a foot pedal was used to operate the shutter. Photographs of conscious mice were taken.

Genotyping of the Crb1 Locus

DNA samples isolated from tail biopsies were amplified separately for the wild-type allele and the mutant rd8 allele as described previously;3, 17 the size of the amplified wild-type allele was 220 bp and the mutant rd8 allele was 244 bp.

Determination of Serum and Liver Retinol Concentrations

Serum and liver retinol concentrations were determined by high-performance liquid chromatography when the mice were 8 and 30 weeks of age. SRL determined serum retinol as described.22 Japan Food Research Company determined liver retinol as described.23 As retinyl esters are converted to retinol in this method, the amount of retinol was expressed as a total retinol.

Number of Mice Used for Each Experiment

Number of mice used for each experiment was shown in Supplementary Table S1.

Statistical Analysis

At least three independent experiments were carried out. Unpaired Student’s t-tests were used for the statistical analysis. Results are expressed as mean±s.e. P<0.05 was considered to indicate a significant difference.

RESULTS

Establishment of ES Cell Clones and Mouse Strains with the Targeted Null Allele

After electroporation with the targeting vectors, 98 neo-resistant clones were screened for targeted recombination by using Southern blot analysis. With a neo probe, three clones (nos. 19, 83, and 87) gave a 5.1-kb band when digested with XhoI/SalI and an 8.8-kb band when digested with AgeI/HpaI (Supplementary Figure S1). All three clones digested with MunI gave 8.2- and 10.1-kb bands with the 5′ probe; the same clones digested with SacII/HpaI gave 7.5- and 9.3-kb bands with the 3′ probe (Supplementary Figure S1). These patterns indicated that the presence of the targeted allele. Clones 19 and 87 showed germline transmission and were used to create the homozygous mutant mice (Rbp4−/−). These mice were born at the expected Mendelian frequency, indicating that Rbp4 is not essential for normal prenatal development with a diet containing 1515 IU/100 g food. Rbp4−/− adults were viable and fertile.

Confirmation of the Null Mutation in Rbp4−/− Mice

No Rbp4 mRNA was detected in the brain, lung, liver, and kidney of Rbp4−/− mice by using RT-PCR (Figure 2a). Northern blots of liver RNA and western blots of serum confirmed the absence of Rbp4 mRNA (Figure 2b) and RBP4 protein (Figure 2c), respectively, in these mice.

Figure 2
figure 2

Confirmation of null mutation. (a) Semiquantitative reverse transcription-polymerase chain reaction (RT-PCR) analysis. The expected 244-bp band was detected in tissues from Rbp4+/+ and Rbp4+/− mice but not in those from Rbp4−/− mice. (b) Northern and western blot analyses. No Rbp4 mRNA or RBP4 protein was detected in the liver or serum, respectively, from Rbp4−/− mice.

Full size image

Retinal Function

A long-term ERG study was performed to assess age-dependent changes in retinal function. Representative ERG waveforms are shown in Figure 3a.

Figure 3
figure 3

Abnormal retinal function. (a) Representative electroretinogram (ERG) waveforms. ERG responses in Rbp4−/− mice had recovered moderately at 40 weeks of age, but were still lower compared with those in Rbp4+/+ mice. (b) ERG responses to varying light intensity. The a- and b-wave amplitudes were strongly reduced in Rbp4−/− mice during the 40-week observation period. The a- and b-wave implicit times were significantly shorter in Rbp4−/− mice than at 7 weeks of age, but had recovered to normal by 40 weeks of age.

Full size image

The a- and b-wave amplitudes were severely reduced in Rbp4−/− mice throughout the 40-week observation period. The a- and b-wave implicit times were longer in Rbp4−/− mice at 7 weeks of age, but recovered to normal at 40 weeks of age. Both strains showed increased a- and b-wave amplitudes and decreased a-wave implicit time with increased light intensity (Figure 3b).

To analyze the ERG changes during growth, the same data were plotted against mouse age (Figure 4). In Rbp4+/+ mice, there were almost no changes in the intensities of the a- and b-wave amplitudes with age. The a-wave implicit time was a little longer and the b-wave implicit time was a little shorter at 40 weeks than at 7 weeks of age (Figure 4a); however, the reasons for this are unclear. Both the a- and b-wave amplitudes were severely reduced in Rbp4−/− mice throughout the 40-week observation period. However, the delays in a- and b-wave implicit times were slight at 7 weeks and were within the normal range at 14 weeks (Figure 4b).

Figure 4
figure 4

Electroretinogram (ERG) changes during growth. (a) ERG response in Rbp4+/+ mice. (b) ERG response in Rbp4−/− mice. In Rbp4−/− mice, the a- and b-wave amplitudes had not recovered to within the normal range even at 40 weeks of age.

Full size image

Histological Comparison of the Retinas of Rbp4+/+ and Rbp4−/− Mice

We examined whether the impairment of retinal function was caused by pathologic changes in the eyes. The entire retina of an Rbp4+/+ and an Rbp4–/– were shown in Figures 5a and b and Supplementary Figure S2. A hyaloid artery remnant extended from the optic disc towards the vitreous humor (arrowhead) (Figure 5b). There were no obvious differences in the central retinas of Rbp4+/+ and Rbp4−/− mice (Figures 5c and d); however, the peripheral retinas of most Rbp4−/− mice showed choroidal loss (5/7 eyes; Figure 5e). Furthermore, the Rbp4−/− peripheral retina showed various morphological defects (Figures 5f–i). Vacuolation (2/7 eyes) (Figure 5f) and formation of multiple layers of hypopigmented retinal pigment cells (2/7 eyes) (Figure 5g) were observed in the Rbp4−/− retinal pigment epithelium. Large, irregular, macrophage-like cells (3/7 eyes)24 appeared between the INL and retinal pigment epithelium in the regions where severe photoreceptor loss developed (2/7 eyes) (Figures 5g and h). One of the Rbp4−/− mice had retinal structure collapse including disappearances of both retinal and choroidal layers in the peripheral retina (Figure 5i).

Figure 5
figure 5figure 5

Histological analysis of the retina. Representative hematoxylin and eosin–stained sections of the retinas of 40-week-old mice at low (a and b) and high magnifications (c and d) and lesions observed in the Rbp4−/− retinas (e–i). (a) The entire retina of an Rbp4+/+ mouse. (b) The entire retina of an Rbp4−/− mouse. A hyaloid artery remnant extended from the optic disc towards the vitreous humor (arrowhead). (c) Central retina: the boxed area in (a) at higher magnification. (d) Central retina: the boxed area in (b) at higher magnification. (e) A representative section of an Rbp4−/− peripheral retina showing choroidal loss. (f) Vacuolation in the retinal pigment epithelium. (g) Hypopigmented retinal pigment cells (arrow) or focal loss of the photoreceptor layer with appearance of large, irregular, macrophage-like cells (arrowheads and the areas bounded by a dashed line). (h) Focal loss of the photoreceptor layer and appearance of large, irregular, macrophage-like cells (arrowheads). (i) Focal collapse of the retinal structure with disappearance of the retinal and choroidal layers. Ch, choroid; GCL, ganglion cell layer; INL, inner nuclear layer; IS/OS, inner segments/outer segments; ONL, outer nuclear layer; RPE, retinal pigment epithelium; Sc, sclera. Scale bar: 500 μm in panels a and b, 25 μm in panels c–f, 50 μm in panels g–I.

Full size image

Measurement of Retinal Properties

To evaluate the quantitative differences in the retinas from the Rbp4+/+ and Rbp4−/− mice, various analyses were performed. The number of neurons in the ganglion cell layer and the thicknesses of the ONL, OSL, and whole retina were decreased in the peripheral region of retina from Rbp4−/− mice (Figure 6).

Figure 6
figure 6

Quantitative comparison of retinal properties. (a) Cell number in the ganglion cell layer (GCL). Significant retinal ganglion cell loss was observed in the peripheral region of retinas from Rbp4−/− mice. (b) Thickness of the inner nuclear layer (INL). No significant differences were observed. (c) Thickness of the outer nuclear layer (ONL). The ONL in peripheral retinas was significantly thinner in Rbp4−/− mice than in Rbp4+/+ mice. (d) Thickness of the outer segment layer (OSL). The OSL in peripheral retinas was significantly thinner in Rbp4−/− mice than in Rbp4+/+ mice. (e) ONL/INL ratio. No significant differences were observed. (f) Retinal thickness from GCL to OSL. Retinal thickness was significantly decreased in the peripheral retinas of Rbp4−/− mice. (g) Distance between central and peripheral retina. No significant difference was observed. Data are presented as means±s.e.

Full size image

Transmission Electron Microscopy

Similar to histological analyses, transmission electron microscopy showed that Rbp4−/− mice had thinner photoreceptor layer, loss of retinal pigment cell, and thinner choroidal layer in the peripheral region (data not shown). We also found some abnormalities in the central retina. In Rbp4−/− mice, the distance between the inner limiting membrane and outer plexiform layer was much shorter than that in Rbp4+/+ mice (Figure 7a), with fewer retinal ganglion cells (Figure 7b). Much fewer synapses were found in the inner plexiform layer (Figure 7b). Electron density in the outer plexiform layer (a layer of neuronal synapses) was very low in Rbp4−/− mice (Figure 7c). Although the number of mitochondria in photoreceptor cell of Rbp4−/− mice and Rbp4+/+ mice was similar, there were very few cristas in Rbp4−/− mice (Figure 7d). In addition, the Golgi apparatus was swollen (Figure 7e).

Figure 7
figure 7

Transmission electron microscopy. (a) Distance between the inner limiting membrane and outer plexiform layer (↔□). This distance was much shorter in Rbp4−/− mice than in Rbp4+/+ mice. (b) Ganglion cell and inner plexiform layer. In Rbp4−/− mice, ganglion cells (asterisks) were lost and much fewer synapses were found in inner plexiform layer (bounded by a dashed line □). (c) Outer plexiform layer. In Rbp4−/− mice, electron density in the outer plexiform layer (surrounded by white line) was very low. (d) Mitochondria in a photoreceptor cell. Although the number of mitochondria (asterisks) was similar in Rbp4−/− and Rbp4+/+ mice, there were very few cristas in Rbp4−/− mice. (e) Golgi apparatus (asterisks) was swollen in Rbp4−/− mice.

Full size image

Defects in the Ocular Fundus

We examined the ocular fundus by indirect ophthalmoscopy. Developmental defects such as persistent hyaloid artery, retinal depigmentation, and optic disc abnormalities were found in Rbp4−/− mice (Figure 8). These defects did not recover by 40 weeks of age (Figure 8).

Figure 8
figure 8

Defects in the ocular fundus. Persistent hyaloid artery, retinal depigmentation, and optic disc abnormalities were observed as early as 7 weeks of age in Rbp4−/− mice. Black arrow: hyaloid artery; white arrow: retinal depigmentation; white arrowhead: optic disc abnormality.

Full size image

Rd8 Mutation

As reported by Mattapallil et al,17 the rd8 mutation in the Crb1 gene is found in the C57BL/6N substrain but not in the C57BL/6J substrain. This mutation results in ocular spots and retinal lesions such as retinal folding, dysplasia of the nuclear layers, retinal degeneration, and vacuolation of the retinal pigment epithelium. To rule out this possibility that the defects we observed in Rbp4−/− mice were caused by the rd8 mutation, we analyzed this mutation in our mice. C57BL/6N mice were obtained from Clea, or F1 hybrid between C57BL/6J and C57BL/6N carried the rd8 mutation as expected (Figure 9a). However, we confirmed that none of the mice used in our morphological and functional analyses of the retina did so (Figure 9b).

Figure 9
figure 9

Absence of the mutant rd8 allele of the Crb1 gene. Representative gels show the wild-type (220 bp) and rd8 (244 bp) bands. (a) The mutant rd8 allele was detected in C57BL/6N (B6N) but not in C57BL/6J (B6J) mice. (b) None of the Rbp4−/− mice used in our morphological and functional assessment of the retina carried the rd8 mutation.

Full size image

Retinol Concentration in the Plasma and Liver

As various eye defects are caused by vitamin A deficiency,25 we determined retinol concentrations in the plasma and liver. Serum retinol concentration in Rbp4+/+ mice was 48.5±14.3 at 8 weeks of age and 97±3.6 IU/dl at 30 weeks (Figure 10a). As expected, serum retinol in Rbp4−/− mice was below the detection threshold (<5 IU/dl) (Figure 10a).

Figure 10
figure 10

Retinol concentration. (a) Serum retinol concentration. Serum retinol was not detected in Rbp4−/− mice. (b) Liver total retinol content. In both strains, retinol accumulated in the liver by 30 weeks of age.

Full size image

Liver total retinol content at 8 weeks was 58.2±5.2 mg/100 g in Rbp4+/+ and 58.3±4.0 mg/100 g in Rbp4−/− mice (Figure 10b). Thus, liver total retinol content was similar in the two strains, suggesting that the absence of RBP4 markedly reduces serum retinol levels, but does not impair accumulation of hepatic retinol stores at this age. At 30 weeks, liver total retinol content was 153.3±2.9 mg/100 g in Rbp4+/+ and 188.8±5.6 mg/100 g in Rbp4−/− mice (Figure 10b). This difference at 30 weeks indicated an inverse relation between the liver retinol content and serum RBP4 concentration. It is of interest that liver total retinol concentration was increased with age even in Rbp4+/+ mice.

Time Course of Rbp4 Expression

To examine whether the increase in serum retinol concentration with age in Rbp4+/+ mice resulted from an increased expression of Rbp4 in the liver, we examined the time course of Rbp4 mRNA expression by using qRT-PCR. The level of Rbp4 mRNA at 12 and 24 weeks of age was 133% and 131%, respectively, of the level at 6 weeks (Figure 11a). The difference between 6 and each of the later time points was significant; the absence of a significant difference between 12 and 24 weeks suggested that Rbp4 mRNA expression in the liver reached a maximum level by 12 weeks.

Figure 11
figure 11

Rbp4 expression and retinol concentration. (a) Expression of Rbp4 mRNA in the liver assessed by means of quantitative reverse transcription-polymerase chain reaction (qRT-PCR). Expression level at 6 weeks of age was set as 1.0. (b) Serum level of RBP4 protein. Expression profiles of Rbp4 in the liver and serum were similar.

Full size image

At 12 and 24 weeks, the serum level of RBP4 was 276% and 304%, respectively, of the level at 6 weeks (Figure 11b). Again, there was a significant difference between 6 weeks and each of the two other time points, but not between 12 and 24 weeks. These results are consistent with those for mRNA expression levels in the liver in that the maximum was reached at 12 weeks. These data suggest that the increase in serum retinol with age is due to the increase in Rbp4 mRNA expression.

DISCUSSION

Here, we established that Rbp4−/− mice in the C57BL/6 background show much more severe phenotypes in terms of ERG abnormalities than Rbp4−/− mice reported previously.15 In accordance with these abnormalities, structural changes were observed, such as loss of photoreceptor and choroidal cells, fewer synapses in the inner plexiform layer, and the shorter distance between the inner limiting membrane and the outer plexiform layer with fewer ganglion cells. Furthermore, our mice had unexpected developmental ocular abnormalities, such as persistant hyaloid artery, retinal depigmentation, and persistent optic disc abnormality.

Quadro et al15 found that Rbp4−/− mice had impaired visual function, as evidenced by abnormal ERG. They had decreased sensitivity to light only in the b-wave amplitude, with progressive improvement such that this sensitivity approached that of wild-type mice by 24 weeks of age; no abnormalities in other parameters were reported. Our Rbp4−/− mice showed decreased sensitivity in both a- and b-wave amplitudes, and retinal impairment did not recover completely, even at 40 weeks of age. Thus, our mice showed more severe retinal impairment than the mice described by Quadro et al.15

As the rd8 mutation was not observed in our mice, retinal impairment could be caused either by retinol deficiency in the blood or by loss of Rbp4 expression in the eyes. Rbp4 is expressed in the retinal pigment epithelium.26, 27, 28, 29 However, little is known about the function of RBP4 in the eye. Quadro et al30 generated a transgenic mouse strain (hRBP4: mRbp4−/−) that expressed human RBP4 under the control of the muscle creatine kinase promoter in the Rbp4−/− background. Although no human RBP4 was detected in the retinal pigment epithelium of hRBP4: mRbp4−/− mice, a normal ERG profile and normal retinol levels were observed in the eyes. These data suggest that circulating RBP4 efficiently delivers retinol to the eye, and that RBP4 present in the eye appears to have no role in maintaining the visual function. Thus, ocular phenotypes in our Rbp4−/− mice may be caused by loss of circulating RBP4 and retinol. In other words, deficiency of circulating retinol may cause abnormalities of photoreceptor cells and synapses among ganglion cells and Muller cells/bipolar cells, because the a- and b-wave amplitude and implicit time are indicators of the functions of these cells.

In human patients with no detectable serum RBP4 because of RBP4 gene mutations, atrophy of the retinal pigment epithelium and iris coloboma are common.31, 32 Although iris coloboma was not observed in our Rbp4−/− mice, retinal depigmentation in these mice might be similar to atrophy of the pigment epithelium in humans with RBP4 gene mutations. In contrast, Quadro et al15 did not observe such ocular abnormalities in their histological analysis under a light or electron microscope. Our Rbp4−/− mice have the C57BL/6 genetic background as confirmed by using 123 polymorphic markers, whereas Quadro et al15 used mice in a mixed background (129xC57BL/6 J). This strain difference might be the reason for the more severe phenotypes of our Rbp4−/− mice compared with those of the mice studied by Quadro et al.

In Rbp4−/− mice, liver retinol content was the same as that in Rbp4+/+ mice at 8 weeks, but became higher than that in Rbp4+/+ mice at 30 weeks. This is consistent with the previous findings15, 33 that there was no significant difference in total hepatic retinol between Rbp4−/− and Rbp4+/+ mice at 3 and 13 weeks of age, and a significantly higher hepatic retinol level in Rbp4−/− compared with that in Rbp4+/+ mice at later stages. In the present study, we found that the amount of retinol in the liver was increased, even in Rbp4+/+ mice, at 30 weeks compared with at 8 weeks of age. Quadro et al15 examined individual levels for hepatic retinol (28 μg/g) and hepatic retinyl ester levels (1800 μg/g). As the average mol% for retinol in retinyl esters is 0.53, hepatic retinyl ester levels were equivalent to 954 μg/g retinol. Thus, the liver total retinol level in their mice was 982 μg/g (954 μg/g+28 μg/g) and was higher compared with our mice (583 μg/g). This is in good agreement with the fact that Quadro et al15 used a diet containing 0.66 mg/100 g, which is ~145% of the retinol than the one used in our study (0.454 mg/100 g).

Structural variation is widespread in mammalian genomes34, 35 and is an important cause of disease.36 Genome-wide association studies identified single-nucleotide polymorphisms in several genes, such as GRIA4, KCNQ5, RDH5, LAMA2, BMP2, SIX6, PRSS56, GJD2, RASGRF1, ZC3H11B, and WNT7B, as risk factors for refractive error including myopia.6, 7, 8 Although some of these genes (SIX6, PRSS56, and WNT7B) function in eye development,7, 8 it remains unclear how such sequence variants shape phenotypic variation. Phenotyping studies by using genetically engineered mice with different genetic backgrounds in conjunction with sequence analyses will provide useful information on the association between genomic and phenotypic variations.