C57BL/6N Mutation in Cytoplasmic FMRP interacting protein 2 Regulates Cocaine Response

See allHide authors and affiliations

Science  20 Dec 2013:
Vol. 342, Issue 6165, pp. 1508-1512
DOI: 10.1126/science.1245503

Not All Mice Are Equal

Different laboratories often use different strains of inbred animals, but one cannot make behavioral comparisons and assume that their reaction to interventions will necessarily be similar. Kumar et al. (p. 1508) have detected differences in cocaine response between the widely used C57BL/6N and C57BL/6J mouse strains and used quantitative trait locus analysis to identify a mutation in an inducible gene, Cyfip, that interacts with the Fragile X protein (FMRP) to regulate sensitivity and sensitization to cocaine through regulation of neuronal connectivity.


The inbred mouse C57BL/6J is the reference strain for genome sequence and for most behavioral and physiological phenotypes. However, the International Knockout Mouse Consortium uses an embryonic stem cell line derived from a related C57BL/6N substrain. We found that C57BL/6N has a lower acute and sensitized response to cocaine and methamphetamine. We mapped a single causative locus and identified a nonsynonymous mutation of serine to phenylalanine (S968F) in Cytoplasmic FMRP interacting protein 2 (Cyfip2) as the causative variant. The S968F mutation destabilizes CYFIP2, and deletion of the C57BL/6N mutant allele leads to acute and sensitized cocaine-response phenotypes. We propose that CYFIP2 is a key regulator of cocaine response in mammals and present a framework to use mouse substrains to identify previously unknown genes and alleles regulating behavior.

The reference mouse strain, C57BL/6J, was established in 1921 and has been maintained at the Jackson Laboratory since 1948 (1). In 1951, a colony of C57BL/6J was shipped to the National Institutes of Health (NIH), and C57BL/6N became a second major source of C57BL/6 mice. Large-scale projects use different C57BL/6 substrains, including the International Knockout Mouse Consortium (IKMC), which uses C57BL/6N embryonic stem (ES) cells (2, 3), and the Mouse Genome Sequencing Consortium and Allen Brain Atlas, which use C57BL/6J (4, 5). Although most behavioral measurements to date have been carried out in the C57BL/6J substrain, the IKMC will shift emphasis to the C57BL/6N substrain. Although behavioral differences have been noted in C57BL/6 substrains, their genetic basis has not been elucidated (6, 7). Therefore, it becomes important to understand the genetic causes of phenotypic differences between C57BL/6N and C57BL/6J.

Drug addiction is characterized by loss of control over drug consumption and behaviors associated with drug seeking. Even though drugs of abuse fall into different pharmacological categories, they all act on the mesolimbic reward pathway in the brain (8). Long-term structural and functional changes in neuronal circuitry are thought to be a key feature of addiction (9). We investigated in detail a difference in cocaine response between the C57BL/6N and the C57BL/6J mouse substrains.

We characterized C57BL/6N and C57BL/6J animals for locomotor response to cocaine. Because of the nonlinear nature of drug response, we carried out testing at multiple doses. C57BL/6N had a 45% (1 SD) lower acute response to cocaine and methamphetamine at multiple doses (Fig. 1, A to D, and fig. S1). In psychomotor sensitization, repeated administration of the stimulus elicits a progressively larger behavioral response. It is regulated by experience-dependent neuronal plasticity and thought to be a key event leading to addiction (10, 11). At a 10-mg/kg close, C57BL/6J sensitized to cocaine much more efficiently than C57BL/6N (compare day 5 to 10, Fig. 1E), although both strains sensitized to similar levels at a higher 15-mg/kg dose (Fig. 1F).

Fig. 1 Acute and sensitized psychostimulant response, measured by locomotor hyperactivity, is lower in C57BL/6N (B6N) substrain than in C57BL/6J (B6J).

(A) Baseline locomotor velocity data of B6J (blue) and B6N (green) mice were measured for 30 min, and then mice were injected with cocaine (20 mg/kg, red arrow) and recorded for 60 min further. (B) Cocaine dose response for B6J and B6N. (C) Decreased methamphetamine response in B6N. Animals were injected with methamphetamine (2 mg/kg, red arrow) and recorded for 3 hours. (D) Methamphetamine dose response for B6J and B6N. Mice were treated with 1, 2, and 4 mg/kg of the drug. Average response 60 min post injection is shown. (E and F) B6N have a lower sensitized response to cocaine at 10 mg/kg (E) but respond similarly at 15 mg/kg dose (F). Mice were measured for baseline and postinjection response for 1 hour and injected with saline (S) or cocaine (C). The 30-min postinjection response is shown. All data are shown as mean ± SEM; the number of animals in each group is indicated. *P < 0.05,**P < 0.01,***P < 0.001,****P < 0.0001 from Tukey post hoc test following two-way analysis of variance (ANOVA) [(B) and (D)] or pairwise t test [(E) and (F)].

To determine which genetic loci contribute to reduced cocaine response seen in C57BL/6N, we carried out quantitative trait locus (QTL) analysis (fig. S2). Parental strains, F1, and F2 animals were tested concurrently. C57BL/6N had an ~1-SD lower response to cocaine. The F1 animals exhibited a cocaine response similar to that of C57BL/6N, indicating dominance of the C57BL/6N allele, and the F2 progeny were intermediate in response when compared with the parental strains (Fig. 2A). These differences were seen in all of our measures and in both sexes (figs. S3 and S4).

Fig. 2 QTL on chromosome 11 regulates cocaine response.

(A) Hyperactivity after intraperitoneal injection of cocaine (20 mg/kg) was quantified. B6J (n = 73), B6N (n = 44), F1 (n = 124), and F2 (n = 244) were tested. The blue boxes represent mean ± 1 SD ranges, and the histograms show normal distributions of the measures. (B) Locations of polymorphic markers between B6J and B6N. (C) Genome-wide QTL scan revealed a single highly significant QTL on chromosome 11. The significance thresholds are established through 100,000 permutation tests. (D) The chromosome 11 QTL is significant for multiple cocaine response measures. (E) Genotype effect plot at marker rs13481014 shows the B6N allele has lower response than B6J allele, and the F1 is intermediate. Data are mean ± SEM.

To perform QTL analysis, we identified polymorphic markers between C57BL/6J and C57BL/6N (supplementary materials, fig. S5, and tables S1 and S2). QTL analysis for cocaine response yielded a single QTL on chromosome 11 [log of odds (LOD) = 6.8] (Fig. 2C). This peak is highly significant on the basis of permutation tests (Fig. 2C), is specific to cocaine response, and was not seen for baseline activity (fig. S6). The peak linkage occurs at marker rs13481014 for 30 min, 60 min, and net response (Fig. 2D), and the genotype effect plot indicates that the B6N/B6N allele elicits a lower response than the B6J/B6J allele. The F1 response is similar to that of C57BL/6N, implying dominance of B6N allele (Fig. 2E). This QTL accounts for 11% of the total phenotypic variance (genetic, environmental, and error) and 61% of the genetic variance (fig. S7). The QTL support interval translates to a 22-Mb interval between 35 to 57 Mb of chromosome 11 (12) (supplementary materials and fig. S8).

Phenotypic differences between related strains may occur because of three possibilities: residual heterozygosity at the time of separation, genetic contamination after separation, or genetic drift. Genotyping indicated that inbreeding was essentially complete and that there was no genetic contamination or residual heterozygosity in C57BL/6N and C57BL/6J substrains (13). Because C57BL/6N and C57BL/6J diverged 62 years ago, there may be a limited number of polymorphisms accumulated through genetic drift that accounts for the phenotypic difference. To identify the causative mutation, we performed whole-genome sequencing runs and combined the data with published and unpublished C57BL/6NJ sequences for 99.8-fold coverage of chromosome 11 (table S3) (14). We reanalyzed the data sets with identical parameters (fig. S9 and tables S4 and S5) and found a single nonsynonymous single-nucleotide polymorphism (SNP) in the QTL interval (Fig. 3A and fig. S10). No indels or structural variants cause protein-coding changes (Fig. 3, B and C). This nonsynonymous SNP changes G to A at 46,036,117 base pair (bp) of chromosome 11 (mm9) and produces a serine-to-phenylalanine missense mutation at position 968 (S968F) in Cyfip2 (Fig. 3D). This variant had a high Polyphen2 score of 0.957/1 and was predicted to be damaging (15) (fig. S11). CYFIP1 and CYFIP2 are highly conserved proteins implicated in Fragile X-mental retardation protein (FMRP) function and regulation of actin polymerization through the WAVE regulatory complex (WRC) (16, 17). Cyfip2 is widely expressed throughout the brain, whereas Cyfip1 is more restricted (fig. S12). FMRP is the most common cause of mental retardation in humans, and WAVE complex members have been shown to regulate neuronal connectivity and behavior in mice and Drosophila (1823).

Fig. 3 Next-generation sequencing identifies a single nonsynonymous polymorphism in Cyfip2.

Classification of SNPs (A), indels (B), and structural variants (C) sequencing reveals only a single SNP [top row of (A)] in Cyfip2 (D) that changes Ser968 to phenylalanine in B6N. Only the QTL support interval is shown. UTR, untranslated region. (E and F) A comparison of C57BL/6 substrains shows that S968F variant arose in B6N lineage between 1961 and 1974.

Because many C57BL/6 substrains are readily available with information about the time of separation from the founder colony, we constructed a phylogenetic timeline of the S969F variation (Fig. 3, E and F). The Cyfip2 polymorphism was fixed in the C57BL/6N colony sometime between 1961 and 1974 and is present in most commercially available sources of C57BL/6N, including Charles River (NCrl) and Taconic (NTac) (Fig. 3F).

The Cyfip gene family is highly conserved in metazoans, and a multiple-sequence alignment indicated that CYFIP2 S968 is 100% conserved in orthologs from Caenorhabditis elegans to humans (fig. S13A). The crystal structure of CYFIP1 has been solved as part of a pentameric 400-kD WRC (16). Molecular modeling of the S969F mutation using the WRC homology model revealed that the mutation leads to substantial steric hindrance with adjacent residues (figs. S13B and S14). We measured protein stability after cycloheximide treatment to estimate the relative half-life of wild-type and mutant CYFIP2. The wild-type and mutant proteins had half-lives of 8.5 and 2.8 hours, respectively, indicating that the S968F mutation greatly destabilizes CYFIP2 (Fig. 4A) (supplementary text). Stably transfected HEK293 cell lines expressing equivalent levels of the wild-type and mutant CYFIP2 proteins were used for biochemical and proteomic analyses. Quantitative co-immunoprecipitation (IP) assays showed no difference in interactions between wild-type and mutant CYFIP2 with NAP1, WAVE1, ABI2, and HSPC300, the core components of the WRC (Fig. 4B). We did not observe differences in our proteomic analysis (fig. S15) and saw no interaction of wild-type or mutant CYFIP2 with FMRP, FXR1, or FXR2 under our conditions.

Fig. 4 CYFIP2 S968F causes destabilization, and knockout analyses of Cyfip2 show that heterozygous mice have rescued cocaine response phenotypes.

(A) Protein stability assay shows CYFIP2 S968F is less stable than wild type (WT). Human 293T cells transfected with FLAG-CYFIP2 (green bands) were treated cycloheximide for the indicated times, and Western blot was performed (bottom). β-tubulin was used as a control (red bands). Replicate experiments were quantitated, and half-life was calculated by using nonlinear one phase decay. K, the half-life parameter, is highly significant P < 0.0001 (top). (B) Quantitative IP experiments show CYFIP2 S968F interacts with WAVE complex members. IP (left) was conducted in replicate, and binding was quantitated (right). (C) Mice generated from ES cells harboring the knockout (KO) first allele of Cyfip2 from the IKMC are on B6N background (designated Cyfip2B6N/B6N for WT and Cyfip2B6N/–- for heterozygous knockout). The homozygous deletion is lethal at birth and therefore not shown. Cyfip2B6N/B6N animals have lower cocaine response than the Cyfip2B6N/– mice. (D) The Cyfip2 heterozygous mice show higher sensitized response to cocaine than WT. Cyfip2B6N/B6N (orange) and Cyfip2B6N/– (purple) were injected with saline (S) or cocaine (C, 10 mg/kg). Pairwise comparisons for genotype for each day were significant as indicated. (E) Diolistic labeling of nucleus accumbens neurons shows decrease in dendritic spine density in B6N. Sample images are shown with representative location of neurons (yellow on coronal section of brain). Quantitation and classification of spines are shown. At least six animals were used in each group with over 10 images collected from each animal. (F) B6N has a decrease in frequency of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor mEPSCs. Sample traces of mEPSCs in nucleus accumbens shell medium spiny neurons from B6J and B6N mice. Dot plots for mEPSC amplitude and frequency. Data are mean ± SEM and Student’s t test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). Two-way ANOVA with Tukey post hoc test was used for (A).

To generate a second mutant allele of Cyfip2, we produced mice using ES cells carrying the “knockout first” (null) allele from the IKMC (3) and maintained them in C57BL/6N background (fig. S16). The homozygous knockouts die within a few hours of birth; however, the heterozygous mice are viable with no overt phenotypes. Because the ES cells used in IKMC are of C57BL/6N origin, the wild-type mice carry the mutant C57BL/6N alleles of Cyfip2 (designated Cyfip2B6N/B6N), and the heterozygotes carry one copy of the C57BL/6N allele (designated Cyfip2B6N/–). We compared mice heterozygous for the knockout allele with the homozygous WT mice for cocaine response (Fig. 4C). As expected the C57BL/6N mice have lower cocaine response than C57BL/6J mice at all doses (Fig. 4C). The Cyfip2B6N/B6N mice also have low response to cocaine, similar to that of C57BL/6N; however, heterozygous Cyfip2B6N/– mice have higher cocaine response than homozygous Cyfip2B6N/B6N mice at all three doses (Fig. 4C). The deletion of one mutant allele alleviates the severity of the phenotype seen with two mutant alleles and explains the dominant phenotype of the F1 progeny and the genotype effect plot of linked QTL maker seen previously (Fig. 2, A and E). The sensitized response is also higher in heterozygous mice (Cyfip2B6N/–), confirming that CYFIP2 regulates acute and sensitized response to cocaine (Fig. 4D). We did not see any changes in cocaine metabolism between the two strains (fig. S17).

Next, we explored the functional effect of the CYFIP2 S968F mutation. Drug-induced structural plasticity is thought to be key in addiction. We measured and classified dendritic spine morphology of medium spiny neurons in the nucleus accumbens. C57BL/6N have a lower number of total spines with significantly lower spines of each class (Fig. 4E). Because dendritic spines are the major sites of excitatory postsynaptic current, we predicted a deficit in excitatory glutamatergic signaling in C57BL/6N. We measured the frequency and the amplitude of mini-excitatory postsynaptic signaling currents (mEPSCs) in the nucleus accumbens shell and found a decrease in frequency, but not amplitude, of mEPSCs (Fig. 4F). Lowered frequency of mEPSCs is a plausible functional consequence of reduced dendritic spines, although we cannot rule out other adaptations.

Our work has three major implications. First, we identify a mutation at the nucleotide level by using QTL analysis and provide evidence that CYFIP2 is a regulator of acute and sensitized cocaine response. Second, with over 20 commercially available C57BL/6 substrains, we describe a framework to use mouse substrains as a rich genetic source for identifying previously unknown genes and alleles that regulate behavior. If other substrains such as C3H/He and DBA/2 are included, there may be hundreds of mouse substrains available for analysis. Third, care must be taken when comparing new behavioral data from C57BL/6N substrains with existing data from C57BL/6J.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S17

Tables S1 to S5

References (2460)

References and Notes

  1. Acknowledgments: We thank A. Khan and W. Khan for genotyping help; A. M. Wissman, J. Gibson, and K. Huber for help with diolistics; C. Olker for behavioral testing; C. Cowan for advice on sensitization; M. Strobel for C57BL/6 substrain DNA; S. Padrick, B. Chen, and M. Rosen for WRC reagents and advice; T. Keane (Sanger Centre) for access to sequencing data; N. Kumar and I. Kornblum for technical help; and O. Hermanson and P. Lowrey for advice on manuscript. V.K. was funded through National Research Service Award F32DA024556 from NIDA. NIH grant U01MH61915 funded J.S.T. J.S.T. is an Investigator and S.H.Y. and V.K. were associates in the Howard Hughes Medical Institute.
View Abstract

Navigate This Article