Results

The LD of RanBP2 Interacts with Cox11 and HKI
The LD of RanBP2 (Figure 1A) is a large and orphan domain of ~700 residues (~80 kDa), for which no molecular partners have been identified until this date. Brain and retina yeast two-hybrid libraries were screened with LD. Cox11 was identified as a partner to this domain (Figure 1B and 1C). Cox11 is a metallochaperone implicated in cytochrome c oxidase assembly [30,31]. Structure-function analysis of the interaction between mCox11, LD of RanBP2, and subdomains thereof, with quantitative yeast two-hybrid assays [32], showed optimal interaction between the intact LD and Cox11 proteins (Figure 1C). Pull-down assays of retinal extracts with glutathione S-transferase (GST)-LD precipitates a sodium dodecyl sulfate-resistant dimer isoform of Cox11 (Figure 1D, top panel), which does not bind to GST-LDZIP alone. In addition to Cox11, we also found that other mitochondrial components such as the outer membrane-associated protein, HKI [33] (Figure 1D, bottom panel) and mHsp70 (unpublished data), associated with LD of RanBP2. This association was highly specific toward the HKI isoform, because HKII, HKIII, and glucokinase did not interact with the LD of RanBP2 (unpublished data). The interaction of Cox11, HKI, and mHsp70 with RanBP2 occurred in vivo in retinal extracts, since antibodies against these and RanBP2 coimmunoprecipitated RanBP2 (Figure 1E) and HKI (Figure 1F), respectively, and these interactions were observed across different tissues (Figure S1). Since RanBP2 exhibits chaperone activity, we assessed whether the interaction between the LD of RanBP2 and Cox11 was direct and the chaperone activity of LD toward folding species of Cox11. Reconstitution binding assays were carried out between purified LD and Cox11, fully and partially denatured with GnHCl and urea, respectively, and native Cox11 (Figure 1G and 1H). Partial denatured Cox11 exhibits significantly higher and concentration-dependent binding affinity toward LD compared with the native and fully denatured Cox11 (Figure 1G). In addition, native Cox11 purified upon expression in the presence of CuSO4 (a prosthetic group tightly bound to Cox11) [31], shows significantly higher binding activity toward the LD of RanBP2, than in the absence of CuSO4 (Figure 1H).
Figure 1  The LD of RanBP2 Interacts with Cox11 and HKI
(A) Primary structure of RanBP2 and its structural/functional domains. The N-terminal LD of RanBP2 is underlined.
(B) Sequence alignment of murine and yeast Cox11. The yeast Cox11 C- and N-terminal domains are poorly conserved. Arrow and solid line denote the predicted mitochondrial cleavage site and membrane-spanning domain. The dotted and dashed lines above the aligned sequences represent, respectively, Cox11-N and Cox11-C constructs shown in Figure 1C.
(C) Structure-function analysis of the interaction between the LD of RanBP2 and Cox11. Optimal interaction between the LD and Cox11 occurred in the presence of constructs comprising both the complete LD and Cox11. Although removal of the cytosolic N-terminal (Cox11-C) significantly decreased the interaction with LD, the mitochondrial intermembrane domain of Cox11 (Cox11-C) together with the C-terminal half of LD (LD-C) retained most of the interaction activity. LD-N and LD-C ended and began with the leucine zipper domain of RanBP2. White and black bars denote β-galactosidase activity and growth rates in selective growth medium, respectively. Results shown represent the mean ± SD, n = 3.
(D) GST pull-down assays with the LD of RanBP2 and its leucine zipper domain and retinal extracts. The LD, but not the leucine zipper domain of RanBP2, associate with Cox11 (top panel, lane 1) and HKI (bottom panel, lane 1).
(E) Coimmunoprecipitation of RanBP2 with antibodies against its molecular partners shows that RanBP2 forms a complex in vivo with HKI (lanes 1 and 2), mHsp70 (lane 3), and Cox11 (lane 4). Lanes 5, 6, and 7 are control immunoprecipitation reactions with different antibodies against the RanBP2 domains, KBD, ZnF, and XAFXFG of nucleoporins.
(F) Reciprocal coimmunoprecipitation of HKI with antibodies against RanBP2 (used and shown in (E)).
(G) Reconstitution pull-down assays with purified LD and increasing concentrations of native (top panel), denatured (middle panel), and partially denatured (bottom panel) Cox11, respectively, in the absence and presence of denaturating agent, GnHCl and chaotropic agent, urea. Folding intermediates (lower panel) of Cox11 exhibit the highest binding activity toward the LD of RanBP2.
(H) Similar experiments as in (G) but in the presence of native Cox11 expressed in the absence (top panel) or presence (bottom panel) of CuSO4. The mature isoform of the metallochaperone has an increased affinity toward the LD of RanBP2. LD, leucine-rich domain; LZ, leucine zipper domain; RBD1–4, Ran-binding domains 1–4; ZnF, zinc finger cluster domain; KBD, kinesin (KIF5B/KIF5C)-binding domain; CLD, cyclophilin-like domain; IR, internal repeat domain; CY, cyclophilin domain.

Cox 11 Inhibits HKI Activity and the LD of RanBP2 Reverses the Inhibition of Cox11 over HKI
To probe whether the interaction of Cox11 and HKI with the LD of RanBP2 modulates the enzymatic activity of HKI, we first examined the effect of increasing concentrations of Cox11 on the initial rates of HKI enzymatic activity (Figure 2A). Cox11 strongly inhibits HKI activity in a concentration-dependent fashion, and at ~15 nM of Cox11, HKI activity could not be recorded (Figure 2A). Cox11 behaves as a partial noncompetitive inhibitor of HKI by affecting the V max of HKI for glucose (Figure 2B). Then, we evaluated the effect of the LD of RanBP2 on the HK activity in the presence of a fixed inhibitory concentration of Cox11, saturating concentration of glucose substrate, and increasing concentrations of LD. As shown in Figure 2C, the LD domain sharply reversed the inhibitory effect of Cox11 on HKI activity in a concentration-dependent manner, but under saturating (and stochiometric) amounts of LD, the velocity of the reaction did not reach that observed for HKI activity in the absence of Cox11 (Figure 2A), suggesting the LD by itself may also have an effect on HKI activity. Indeed, a saturating concentration of LD reduced the V max but not the K m of HKI (Figure 2D) by ~20% under similar conditions.
Figure 2  Effect of Cox11 and RanBP2 on HKI Activity
(A) Saturation kinetics, rate versus glucose of HKI (0.24 μg) in the absence (solid circles) and presence of Cox11 (open circles, 0.25 nM; solid triangles, 7.5 nM). The activity of HKI decreases with increasing concentrations of Cox11. No measurable HKI activity was recorded in the presence of 15 nM of Cox11 (unpublished data).
(B) Hanes-Wolf plot of (A) (1/rate versus glucose) in the absence and presence of fixed concentrations of Cox11. Linearity of reciprocal plots also supported the hyperbolic behavior of the reactions (unpublished data). Cox11 behaves as a noncompetitive inhibitor of HKI by reducing the V max of HKI but not its K m toward glucose.
(C) HKI rate is plotted as a function of LD concentration at saturating glucose and fixed Cox11 (7.5 nM) concentrations. Note that increasing concentrations of the LD of RanBP2 reverse the inhibition of HKI activity by Cox11. A half-maximal effect of the LD of RanBP2 on HKI activity in the presence of 7.5 nM of Cox11 was observed at a concentration of ~0.05 nM of LD.
(D) Rate versus glucose plot in the absence and presence of the LD of RanBP2. At a saturating concentration of the LD of RanBP2 (3.75 nM), the HKI activity was reduced by about 20%. v, rate; S, glucose.

RanBP2 Colocalizes with HK1, Cox11, and mHsp70 in the Retina and Cultured Neurons
In addition to its presence at the vicinity of NPCs, we have shown previously that RanBP2 is localized to and abundant in the mitochondria-rich ellipsoid subcellular compartment of photosensory (photoreceptor) neurons of the retina [19]. We extended the subcellular colocalization studies on RanBP2 and its novel partners by immunocytochemistry to determine if RanBP2, Cox11, HKI, and mHsp70 colocalize in hippocampal neurons (Figure 3A–3C), cerebral cortex neurons (Figure 3D–3F), ellipsoid (mitochondria-rich) subcellular compartments of photosensory neurons of the retina (Figure 3G–3O), and dissociated primary glia and neuron cultures from the brain (Figure 3P–3Z). Double immunostaining with antibodies against these proteins showed that they colocalize to the mitochondria (Figure 3A–3Z).
Figure 3  Localization of RanBP2 and Its LD Molecular Partners
(A–F) are thin cryosections of an area of the hipocampus (CA1 neurons) and cerebral cortex, respectively, immunostained against HKI (A and D), RanBP2 (B and E), and merged images thereof (C and F). Note that while RanBP2 and HKI are widely expressed among and colocalize to hippocampal neurons (C), HKI expression and localization with RanBP2 is restricted to a subset of cortical neurons (likely interneurons) (F). Images of the distal region of bovine retinal cryosections comprising part of the nuclear layer of photoreceptor neurons and their inner (myoid and ellipsoid) segment compartment (G–O) are immunostained against mHsp70 (G) and RanBP2 (H), mHsp70 (J) and Cox11 (K), HKI (M) and Cox11 (N), and merged images thereof (I–O). Note the prominent localization of RanBP2, mHsp70, and Cox11 at the mitochondria-rich ellipsoid compartment of photoreceptors and the colocalization of RanBP2 and Cox11 with mHsp70 (I and L), while HKI colocalization with Cox11 was limited to restricted foci (R, arrowheads). High-resolution images of dissociated primary cerebral neurons and glial cells confirmed that the colocalization of HKI and Cox11 was highly restricted (P–R), while RanBP2 extensively colocalized with HKI (S–U) and mHsp70 (V–Z). Scale bars in A–O and P–Z are 40 and 10 μm, respectively. ONL, outer nuclear layer.

Haploinsufficiency of RanBP2 Causes Decreased Levels of HKI and Partial Mislocalization of HKI
To determine the physiological implications of the interaction between RanBP2 and HKI (and other partners), we employed a murine embryonic stem 129Ola cell line with a targeted RanBP2 locus produced by gene-trapping to produce stable inbred (coisogenic) and mixed background lines, respectively (Figure 4A and 4B). Genotyping of 299 F2 offspring revealed a RanBP2+/+:RanBP2+/−:RanBP2−/− distribution (89:210:0) that deviated from the expected Mendelian ratio and supports that the RanBP2−/− mice were embryonically lethal. E12.5 embryos show strong expression of RanBP2 in the optic vesicle and throughout much of the embryo, which had no apparent developmental abnormalities (Figure 4C). In agreement with previous immunocytochemistry analysis [19], the RanBP2 gene is expressed across mature retinal neurons, but expression in ganglion cells of the adult retina was extremely strong (Figure 4D and 4E).
Figure 4  Insertion Mutagenesis of the Murine RanBP2 Gene
(A) Diagram of the genomic region of RanBP2 disrupted by insertion trap mutagenesis with a bicistronic reporter vector between exon 1 and 2. The bicistronic transcript produces two proteins under regulation of RanBP2. Upon splicing of RanBP2, a fusion between exon 1 and β-geo (a fusion between the β-gal and neo genes) is generated, while human placental alkaline phophatase (PLAP) is independently translated using the internal ribosome entry site. Consistent with previous studies, the expression of the former is directed to cell bodies, while expression of the latter is targeted to the axonal processes [67,68]. Transcriptional 5′ RACE analysis detects a fusion between exon 1 and β-geo.
(B) Southern analysis of the RanBP2 locus of wild-type and heterozygous genomic DNA of tails of F1 mice digested with PpuMI (left panel) and HindIII (right panel) with probes at the 3′ (left panel) and 5′ (right panel) flanking regions of the insertion breakpoint. Q1 is a cosmid containing the RanBP2 gene up to exon 20 [4].
(C) Lateroventral view of a whole-mount stain of a ~12.5 dpc heterozygous embryo for PLAP and β-gal (inset picture) activities. Although PLAP was broadly expressed (e.g., somites, limbs, and CNS), the PLAP and β-Gal (inset picture) expression was particularly high in the optic vesicle (arrow). X-gal single (D) and combined staining with PLAP (E) of a retinal section of a 3-mo-old RanBP2+/− mouse. Consistent with previous immunocytochemistry studies, β-Gal activity is detected in the neuroretinal bodies and inner segment compartment of photoreceptors with conspicuously strong expression in ganglion cells. PLAP expression is found throughout the plexiform/synaptic layers and outer segment of photoreceptors (E). GC, ganglion cell; PLAP, human placental alkaline phophatase; ROS, rod outer segment; RIS, rod inner segment; ONL, outer nuclear layer; OPL, outer plexiform (synaptic) layer; INL, inner nuclear layer; IPL, inner plexiform (synaptic) layer; GC, ganglion cell layer. In light of the association in vivo of RanBP2 with Cox11 and HKI (Figures 1 and 3), profound in vitro modulation of HKI enzymatic activity by RanBP2 and Cox11 (Figure 2), and the critical role of HKI in catalyzing a rate-limiting step of glycolysis, we probed whether RanBP2+/− mice presented disturbances in HKI, Cox11, and energy homeostasis. Monoallelic expression of RanBP2 does not affect the number of NPCs and their distribution in hippocampal neurons (Figure 5A, unpublished data), but it led to consistent lower expression of RanBP2 by more than 50% in the CNS (brain and retina) in 129Ola (Figure 5B–5D), but not in C57BL/6J/129Ola backgrounds (unpublished data). Hence, we focused our analysis on the inbred RanBP2+/− 129Ola mouse line. Although the levels of RanBP2 were decreased in the retina, brain, and hippocampus of RanBP2+/−mice by ~50%–60% (Figure 5B and 5C), the levels of others nucleoporins, Nup153 and Nup62, and mHsp70 and Cox11, remained unchanged (Figure 5B, unpublished data). In addition, we observed a strong decrease of the levels of HKI (3- to 4-fold) (Figure 5B and 5C). This decrease was selective to the CNS, since HKI levels remained largely unaffected in the skeletal muscle, spleen, and liver (Figure 5D). Because HKI plays a key role in the production of energy intermediate substrates and HKI is virtually the sole HK isoform expressed in the CNS [33,34], we probed the impact of HKI and RanBP2 reduction in the levels of ATP. As shown in Figure 5E, there was significant and concordant reduction in levels of ATP in the CNS (brain and retina), but not in non-neuronal tissues. Finally, we also observed partial and selective delocalization of HKI, but not of Cox11, from the ellipsoid (mitochondria-rich) to the adjacent myoid subcellular compartment of rod photosensory neurons (Figure S2A–S2E). This was also accompanied by reduced HKI levels in the inner retina, in particular in the inner plexiform (synaptic) layer (Figure S2A–S2E).
Figure 5  Haploinsufficiency of RanBP2 Causes a Decrease in HKI Protein and ATP Levels
(A) Quantitative analysis of NPCs in dissociated hippocampal neurons of wild-type (+/+) and heterozygote (+/−) mice upon immunostaining with mAb414. No difference in the density of NPCs (3–4 NPC/μm2) at the nuclear envelope was found between RanBP2+/+ and RanBP2+/− mice.
(B) Immunoblots with anti-RanBP2/Nup153/Nup62 (mAb414), −HKI, −mHsp70, and −Cox11 antibodies of retinal (top panel) and hippocampal homogenates of +/+ and +/− mice. In comparison to RanBP2+/+, RanBP2+/− mice exhibit a reduction in the expression levels of RanBP2 and HKI but not of other proteins.
(C) Quantitative analysis of relative protein expression levels of RanBP2, Cox11, HKI, and mHsp70 in the hippocampus of RanBP2+/+ and RanBP2+/−mice. There is ~2- and 4-fold reduction of RanBP2 and HKI in heterozygote mice.
(D) The level of HKI is reduced in the brain but not in other non-neuronal tissues tested (muscle, spleen, and liver).
(E) The total ATP level is reduced in the CNS tissues (brain and retina) but not in non-neuronal tissues tested (e.g., spleen).

Metabolic Disturbances Caused by Haploinsufficiency of RanBP2
The growth rates of inbred RanBP2+/− mice on high-fat (~10% fat) diet were significantly slower than RanBP2+/+ mice (Figure 6A). Beginning at around 4 mo of age, RanBP2+/− mice exhibit a significant slower gain in body mass than wild-type mice (Figure 6A). In addition, RanBP2+/− inbred mice presented deficits in body mass that were erased by changing the genetic background to a mixed C57BL/6J/129Ola (Figure 6B). Food consumption did not account for the body weight differences observed (Figure 6C).
Figure 6  RanBP2+/− Mice on High-Fat Diet Exhibit Deficits in Growth
(A) In comparison to wild-type mice, RanBP2+/− mice show slower growth rates beginning at 4 mo of age (arrow), and the difference in body weight between these is maintained afterward. Note that RanBP2+/− mice lack the growth spur observed in wild-type mice between 3 and 4 mo of age.
(B) In comparison to inbred RanBP2+/−mice (129Ola genetic background), the difference in body weight between RanBP2+/+ and RanBP2+/− mice is masked upon placing these on a mixed 129Ola/C57Bl6 genetic background.
(C) RanBP2+/+ and RanBP2+/−inbred mice exhibit similar rates of food consumption. Mice in (A), (B), and (C) were placed on a high-fat diet since birth (n = 5). HKI in the CNS (brain and retina) accounts virtually for all expression of HK isozymes and glucose utilization in the CNS [33,34]. Moreover, glucose is the sole reliance source of energy in the CNS under normal conditions, the CNS lacks glucose storage sources, and despite the disproportionate mass of the CNS to the rest of the body, the CNS consumes daily about 60% of the body's glucose and 25% of the total oxygen [35,36]. To determine the impact of RanBP2 haploinsufficiency on the utilization, formation, and uptake of glucose, we carried out several physiological assays. In contrast to mice placed on a normal chow diet (~5% fat; unpublished data), RanBP2+/−mice on a higher fat diet (~10% fat) performed significantly worse in the glucose tolerance test beginning at 6 mo of age (Figure 7A and 7B), thus supporting that the RanBP2+/− mice exhibited a deficit in glucose clearance. This deficit was rescued in RanBP2+/− mice of mixed C57BL/6J/129Ola background (Figure S3). Glucose clearance was not affected due to a disturbance in insulin-mediated glucose uptake (Figure 7C). Then, we probed whether RanBP2 induces impairment of gluconeogenesis, which could contribute to the pathophysiological production and clearance of glucose. To this end, the administration of the gluconeogenic substrate precursor, pyruvate (pyruvate tolerance test), showed that there was no difference in glucose production in RanBP2+/− mice (Figure 7D). Hence, partial loss-of-function of RanBP2 had no impact on the gluconeogenesis pathway. However, upon glucose production (15 min), the clearance rates of glucose were again significantly slower in RanBP2+/− than in RanBP2+/+ mice (Figure 7D), confirming an impairment in glucose breakdown.
Figure 7  Metabolic Phenotypes of RanBP2+/− Inbred Mice on High-Fat Diet
(A) 3-mo-old inbred RanBP2+/− mice (n = 5) have normal glucose clearance rates upon glucose challenge and overnight fasting.
(B) In contrast, 6-mo-old inbred RanBP2+/− mice (n = 5) have significantly decreased glucose clearance rates upon glucose challenge and overnight fasting.
(C) Fasted 6- to 8-mo-old RanBP2+/+ and RanBP2+/− mice have no difference in insulin-mediated glucose uptake as assayed by insulin tolerance test (n = 5).
(D) Pyruvate tolerance test shows normal rise in glucose but decreased glucose clearance between inbred RanBP2+/+ and RanBP2+/− mice (n = 5).

Haploinsufficiency of RanBP2 Causes Deficits in the Electrophysiological Output of Receptoral and Postreceptoral Retinal Neurons
In light of the prominent expression of RanBP2 and HKI in retinal neurons [1,19], the vital dependence of the neuronal retina (and brain) on glucose as the main substrate source for energy production, and the determinant impact of metabolic disorders, such as diabetes, in retinal function (e.g., diabetic retinopathy) [37], we probed the impact of deficits in RanBP2, HKI, and ATP, on the electrophysiological responses of subclasses (rod and cone) photoreceptor and postreceptor retinal neurons of RanBP2+/− and in RanBP2+/+ mice. The scotopic (dark-adapted) responses mediated by the rod photoreceptor pathway at low-stimulus intensities and mixed rod and cone pathways at high-stimulus intensities were substantially reduced in RanBP2+/− mice (Figure 8A). The differences in the photopic (light-adapted) responses, initiated by cone photoreceptors, which make up 3% of the photosensory neurons in the mouse retina [38], were less obvious but still exhibited a trend toward reduced amplitudes across a range of increasing light stimulus intensities (Figure 8B). The reduction in the scotopic responses included decreases in both b-wave (Figure 8C) and a-wave (Figure 8D) amplitudes mediated by postreceptoral and receptoral neurons, respectively. Postreceptoral second-order neuron responses, represented by the b-wave, tended to be more consistently and substantially reduced than the a-waves, which directly reflect photoreceptor activity. Since second-order neuron responses depend on input from photoreceptors, this suggests that reduced b-wave amplitudes are the result of the accumulation of decreases in the light response of both photoreceptors and postreceptoral neurons. Anesthetics, particularly ketamine, can cause sustained elevation of glucose in mice, which in turn affects electroretinogram responses [39]. Thus, we were concerned that differences in electroretinogram amplitudes between RanBP2+/− and RanBP2+/+ may reflect differences in glucose level changes in response to anesthesia. However, we found no significant differences in glucose levels measured before and every 15 min during 75 min of anesthesia (n = 4–5). Glucose rose at the same rate and reached a maximum of approximately 3.3 times the pre-anesthesia level in both genotypes (unpublished data).
Figure 8  Electroretinograms from 6-Mo-Old RanBP2+/−and RanBP2+/+ Inbred Mice Showing Photoreceptor and Postreceptor Neuron Electrophysiological Response Phenotypes
(A) Scotopic (dark-adapted) responses from RanBP2+/− mice to light stimuli of increasing intensity, beginning at threshold, have reduced amplitudes compared to those observed in RanBP2+/+ mice. The three lower intensities represent responses generated in the rod photoreceptor neuronal pathway. The upper intensities are comprised of responses generated in both the rod and cone pathways.
(B) Photopic (light-adapted, cone photoreceptor pathway) responses of RanBP2+/− mice to increasing light stimulus intensities also exhibited reduced amplitudes compared to those observed in RanBP2+/+ mice.
(C) Average ± SE (n = 9) scotopic b-wave amplitudes from RanBP2+/− (open circles) and RanBP2+/+ (filled squares) mice representing postreceptoral neuron function. (Note: log amplitude scale.)
(D) Average ± SE (n = 5) scotopic a-wave amplitudes, representing photoreceptor function, for RanBP2+/− and RanBP2+/+ mice in response to bright flashes. Amplitudes of responses from RanBP2+/− mice were lower over the entire range of stimulus intensities for both b- and a-waves. Asterisks represent significant differences between the groups (Student's t test, p < 0.05). Statistical significance was found across all intensities for b-wave amplitudes (2-way ANOVA, p < 0.0001), but not for the a-wave.