Research ArticleALS

Loss-of-function mutations in the C9ORF72 mouse ortholog cause fatal autoimmune disease

See allHide authors and affiliations

Science Translational Medicine  13 Jul 2016:
Vol. 8, Issue 347, pp. 347ra93
DOI: 10.1126/scitranslmed.aaf6038

C9orf72, a suppressor of autoimmunity?

Mutations in C9ORF72 are a common contributor to amyotrophic lateral sclerosis and frontotemporal dementia, yet the function of this gene is still poorly defined. In new work, Burberry et al. demonstrate that mutations disrupting the normal function of C9orf72 cause mice to develop features of autoimmunity. They further found that transplantation of normal bone marrow into mutant animals ameliorated this phenotype, whereas transplantation of mutant bone marrow into normal animals was sufficient to cause autoimmunity. The authors conclude that C9orf72 acts through hematopoietic cells to maintain normal immune function and suggest that investigations are warranted into whether disruptions in immunity contribute to disease in patients.


C9ORF72 mutations are found in a significant fraction of patients suffering from amyotrophic lateral sclerosis and frontotemporal dementia, yet the function of the C9ORF72 gene product remains poorly understood. We show that mice harboring loss-of-function mutations in the ortholog of C9ORF72 develop splenomegaly, neutrophilia, thrombocytopenia, increased expression of inflammatory cytokines, and severe autoimmunity, ultimately leading to a high mortality rate. Transplantation of mutant mouse bone marrow into wild-type recipients was sufficient to recapitulate the phenotypes observed in the mutant animals, including autoimmunity and premature mortality. Reciprocally, transplantation of wild-type mouse bone marrow into mutant mice improved their phenotype. We conclude that C9ORF72 serves an important function within the hematopoietic system to restrict inflammation and the development of autoimmunity.


Amyotrophic lateral sclerosis (ALS) is characterized by the progressive degeneration of motor neurons, resulting in paralysis and death (1). Genetic findings suggest that ALS can result from mutations in genes acting in several cellular processes (2). To date, the most common mutation found in ALS patients, as well as in frontotemporal dementia (FTD), is a hexanucleotide repeat expansion in the first intron of C9ORF72 (3). This mutation is present in 4 to 8% of patients with sporadic ALS (4). The repeat expansion is also found in 40% of familial ALS cases from Europe and the United States and is even more prevalent in Scandinavia (3, 4). Identification of unaffected elderly carriers suggests that the mutation’s effects are incompletely penetrant (5).

The expanded GGGGCC repeat in C9ORF72 has been proposed to mediate its effects through one mechanism or a combination of three mechanisms: the creation of long repetitive RNAs (3, 6), the translation of these RNAs into toxic repetitive dipeptides (7), or the silencing of the mutant allele (3, 8, 9). Histological and gene expression studies indicate that all three effects of the repeat expansion manifest themselves in individuals harboring it. However, determining which effect or effects of the mutation contribute to the degenerative phenotypes seen in patients remains unresolved. The normal protein product of C9ORF72 contains a DENN domain (10) and may act as a guanine nucleotide exchange factor (11). Previous studies in fish and worms suggested that C9ORF72 might function in the nervous system (8, 12). We found that although the murine ortholog of C9ORF72 (3110043O21Rik, hereafter referred to as C9orf72) is widely expressed, it was enriched in motor neurons known to degenerate in ALS (13). However, a recent study demonstrated that conditional excision of C9orf72 in cells of the nervous system had no overt effects, raising the question of whether it plays any role in the nervous system or overall health (14).

Resolving the function of the endogenous C9ORF72 gene product remains important because it could provide insight into whether the reduction in gene product found in patients with the repeat expansion contributes to their disease (3, 8, 9). Furthermore, efforts are proceeding to develop therapeutic approaches for knocking down the gain-of-function products of the repeat expansion (9). Because knockdown strategies could inadvertently depress transcription of the normal gene product, it will be critical to understand the phenotypic ramifications of long-term depression of gene expression from this locus.

Here, we report an initial study of mice harboring loss-of-function mutations in C9orf72. We found that homozygous mutant animals produced through homologous recombination, as well as CRISPR (clustered regularly interspaced short palindromic repeats)/Cas9 targeting, developed several classical features of autoimmunity. Transplantation studies demonstrated that mutant bone marrow cells were sufficient to cause autoimmunity and premature mortality when placed into otherwise wild-type recipients. Thus, we conclude that the gene product encoded by C9orf72 likely acts within the hematopoietic system to play an important role in the promotion of immunological tolerance.


C9orf72 loss-of-function mutations cause premature mortality

We recently identified the mouse ortholog of C9ORF72 and generated heterozygous mice harboring a LacZ insertion replacing exons 2 to 6 of the gene on an inbred C57BL/6 background (13) (Fig. 1A). Because this targeting event eliminated highly conserved 5′ exons of C9orf72, we reasoned that it might result in a loss of gene function, allowing the importance of this gene to be investigated. After intercrossing heterozygous (+/−) animals, we found that mice of the three expected genotypes were recovered at expected Mendelian frequencies (Fig. 1B). We designated this strain “KOMP” because the gene targeting to create this allele was originally performed by the Knock Out Mouse Project (KOMP) Consortium.

Fig. 1. A loss-of-function mutation in the C9orf72 ortholog.

(A) The schematic illustrates replacement of exons 2 to 6 to generate the KOMP allele and crosses with Sox2-cre mice to generate the Neo-deleted allele. 3′ HA, 3′ homology arm; FRT, flippase recognition target. (B) Frequency of genotypes born from KOMP +/− crosses. (C) C9orf72 expression in KOMP whole blood by quantitative reverse transcription polymerase chain reaction (RT-PCR). **P < 0.01, Tukey multiple comparisons. (D) Western blotting of cortical tissue from the three KOMP genotypes using anti-C9ORF72 antibodies. (E) Quantification of Western blot bands. *P < 0.05, **P < 0.01, Tukey multiple comparisons. ns, not significant. Ab, antibody. (F) Frequency of genotypes born from Neo-deleted +/− crosses. (G) C9orf72 expression in Neo-deleted whole blood by quantitative RT-PCR. **P < 0.01, Tukey multiple comparisons.

We next sought to use animals from the +/− intercrosses to determine whether the targeted mutation reduced expression of C9orf72. Using exon-spanning primers, we detected a significant, dose-dependent reduction in the abundance of transcript sequences from exons 6 through 8 in whole blood of +/− and −/− animals relative to that found in wild-type controls (Fig. 1C; **P < 0.01). To assess whether the mutation also resulted in decreased protein levels, we used two anti-C9orf72 antibodies, one raised against C-terminal regions of the protein and the other against its N terminus. In each case, a band with the predicted mass of C9orf72 was significantly depleted in mutant animals (Fig. 1, D and E; *P < 0.05, **P < 0.01).

There are historical examples of the transcriptional units encoding drug resistance used in gene targeting leading to off-target phenotypes (15). We therefore removed the selection cassette used for gene targeting by outcrossing C9orf72-targeted KOMP mice with outbred animals that expressed the Cre recombinase under the control of the Sox2 promoter (16) (hereafter referred to as “Neo-deleted”) (Fig. 1A and fig. S1A). Intercrosses of +/− Neo-deleted animals resulted in recovery of +/+, +/−, and −/− animals at the expected Mendelian frequencies (Fig. 1F). Neo-deleted +/− and −/− mice exhibited a significant and dose-dependent reduction in C9orf72 transcript abundance (Fig. 1G; **P < 0.01). Because proper interpretation of mouse models relies on the fidelity of the genetic modifications introduced, we performed whole-genome sequencing of representative +/+, +/−, and −/− KOMP animals, as well as −/− Neo-deleted mice. These studies revealed proper gene targeting in each strain and the absence of off-target insertions (fig. S1B). We were also able to confirm the C57BL/6 inbred nature of KOMP animals and outbred nature of Neo-deleted animals (fig. S1, C to F).

To determine the effects of C9orf72 loss of function, we intercrossed +/− KOMP animals and monitored littermates for 400 days (Fig. 2A; n = 163; +/+, n = 50; +/−, n = 84; −/−, n = 29). At 70 days of age, we noted no obvious phenotypes (Fig. 2A). However, as the cohort continued to age, the risk of death in both −/− and +/− animals grew to a high level of significance (Fig. 2A; *P < 0.05). We also examined Neo-deleted mice generated by +/− intercrosses (Fig. 2B; n = 50; +/+, n = 15; +/−, n = 21; −/−, n = 14). At later ages, we similarly observed significantly increased mortality in +/− and −/− mice (Fig. 2B; *P < 0.05). Because C9orf72 has been genetically implicated in ALS, we asked whether increased mortality in mutant animals was associated with neural degeneration. However, we did not find obvious changes in the number of spinal motor neurons (fig. S2). Nor did we observe differences in the gross histology of the mutant motor cortex (fig. S3) or other brain regions of mutant animals (fig. S4). We did, however, observe evidence for increased neural inflammation in the cortex of mutant animals as evidenced by focal staining with anti–GFAP (glial fibrillary acidic protein) antibodies (fig. S5).

Fig. 2. C9orf72 mutations lead to premature death of mice.

Mice harboring loss-of-function mutations in C9orf72, generated by homologous recombination using a targeting vector in embryonic stem cells on a C57BL/6 background (KOMP) and those outcrossed with Sox2-cre–expressing mice to remove the neomycin cassette (Neo-deleted), were aged for survival studies. (A and B) Kaplan-Meier survival curves for (A) KOMP and (B) Neo-deleted mice. *P < 0.05, **P < 0.01, generalized Wilcoxon test. (C and D) Body weight of female (C) KOMP mice and (D) Neo-deleted animals over time. *P < 0.05, Dunnett’s multiple comparisons. (E) Cause of death or obligatory euthanasia in KOMP and Neo-deleted mice. “Days of decline” indicates the time between maximum body weight (onset) and death or obligatory euthanasia.

We next sought to establish the onset of decline in these mutant animals, to quantify the duration of their decline, and to better understand the causes of their premature death. We weighed mice weekly and performed necropsy on animals found dead or that we were obliged to euthanize because of preestablished animal welfare criteria. We noted that although animals of all genotypes initially exhibited similar weights, as they aged, the weight of both KOMP and Neo-deleted −/− animals plateaued and became significantly lower than that of wild-type controls (Fig. 2, C and D; *P < 0.05). By defining phenotypic onset for each animal as the age at which it reached its maximum body weight, we were then able to quantify the number of days that a given animal was burdened by decline before its death. Heterozygous animals were burdened by decline for, on average, 118 ± 67 days (KOMP; n = 13) and 56 ± 38 days (Neo-deleted; n = 10), whereas −/− animals were burdened by decline for, on average, 76 ± 51 days (KOMP; n = 21) and 37 ± 29 days (Neo-deleted; n = 9) (Fig. 2E). During these studies, we determined that the causes of demise included cachexia, respiratory failure, hepatomegaly, dermatitis, internal hemorrhage, lymphocytic stromal cell hyperplasia, severe ataxia, and prolapse (Fig. 2E and fig. S6).

Upon necropsy of end-stage animals, we observed enlarged spleens in KOMP −/− (n = 10 of 11) and Neo-deleted −/− (n = 6 of 6) mice (Fig. 3, A and B). Splenomegaly can result from any one of several underlying pathologies, including infection, myeloproliferative disease, chronic lymphocytic leukemia, lymphoma, and autoimmunity. To investigate whether any of these was the cause of splenomegaly in mutant animals, we visually inspected day 300 spleens from Neo-deleted animals (+/+, n = 6; +/−, n = 3; −/−, n = 7) and then subjected them to histological analysis. We observed a clear demarcation between red and white pulp regions in +/+ and +/− spleens, whereas −/− spleens displayed a disruption of red and white pulp boundaries (Fig. 3B). We next asked when splenomegaly first appeared in these animals. We found that day 25 Neo-deleted −/− animals exhibited normal spleen sizes, whereas day 50 −/− animals had developed significantly enlarged spleens relative to their littermates (Fig. 3D; **P < 0.01). This significant increase in spleen weight was maintained in day 100, day 200, and day 300 −/− animals (Fig. 3D; **P < 0.01). We reasoned that splenic vein thrombosis and spontaneous microbial infection were less likely causes of splenomegaly in −/− animals because of the synchronicity and penetrance at which this phenotype developed.

Fig. 3. Identity of cells within the enlarged mutant spleens of Neo-deleted mice.

(A) Spleens from day 300 Neo-deleted animals. Scale bar, 1 cm. (B) Hematoxylin and eosin staining of spleens from Neo-deleted mice at day 300. Scale bar, 500 μm (i) and 50 μm (ii). (C and D) Quantification of spleen weight in (C) end-stage KOMP and Neo-deleted animals and (D) aged Neo-deleted animals. (E) Splenocyte counts from day 200 Neo-deleted mice. (F) Quantification of splenocyte subsets in day 200 Neo-deleted mice. NKs, natural killer cells. DCs, dendritic cells. (C, E, and F) *P < 0.05, **P < 0.01, Tukey multiple comparisons. (D) *P < 0.05, **P < 0.01, Student’s t test. ns, not significant. (G) PCR analysis of T and B cell clonality in the spleens of day 400 Neo-deleted mice. LN2 and LN3 represent embryonic stem cells generated by nuclear transfer from lymph node–derived T cells that harbor monoclonal TCRβ (T cell receptor β) rearrangements.

We next considered the possibility that splenomegaly was the result of lymphoma. In lymphoma, individual clonal populations of either B or T cells can dominate the splenic compartment. However, mutant animals displayed an increase in splenocytes derived from several lineages, including CD19+ B220+ B cells, CD3+ CD4+ T cells, and CD11b+ Ly6G+ neutrophils (Fig. 3, E and F, and fig. S7; *P < 0.05, **P < 0.01). Furthermore, analysis of V(D)J recombination in spleens of mutant animals (17) revealed polyclonal populations of B and T cells, arguing against the type of clonal expansion observed in lymphoma (Fig. 3G). We also did not observe histological changes in the cellularity of −/− bone marrow (fig. S8), which can be a site of lymphocytic transformation (18).

In addition to the nearly ubiquitous splenomegaly observed in −/− animals, we also noted that a subset of end-stage KOMP −/− (n = 4 of 21) and Neo-deleted −/− (n = 3 of 9) animals developed grossly enlarged cervical lymph nodes (fig. S9). Enlarged lymph nodes can result from inflammation, lymphoma, autoimmunity, or an inability of lymphocytes to escape this compartment (19, 20). Although we noted modest changes in the relative proportion of B and T cells, both of these populations were greatly expanded within enlarged cervical lymph nodes of mutant animals (fig. S9), again arguing against clonal transformation as an explanation. Instead, we observed an increase in T cell activation state, with a significant reduction in CD62Lhi CD44lo cells and a significant increase in CD62Lhi CD44hi and CD62Llo CD44hi cells when compared to controls (fig. S9; *P < 0.05, **P < 0.01).

Finally, we noted that a small but significant subset of both KOMP and Neo-deleted mutant animals developed hepatomegaly (Fig. 2E and fig. S10). Histological analysis suggested that this liver overgrowth was likely the result of immune cell infiltration (fig. S10). Such immune infiltrates in liver can be seen in the context of both leukemia and autoimmunity (21). In this case, infiltrating hematopoietic cells appeared morphologically diverse, rather than clonal in nature, which again seemed less parsimonious with the diagnosis of leukemia and more consistent with chronic inflammation (fig. S10).

Changes in peripheral blood were consistent with autoimmunity not leukemia

Assessing the composition of peripheral blood can yield insight into pathways that disrupt the hematopoietic system. We therefore performed whole blood cell counts on mutant mice greater than 300 days of age (Fig. 4, A to H). Numbers of white blood cells, platelets, and RBCs in KOMP +/+ and Neo-deleted +/+ animals were similar to historic ranges in inbred mouse strains (22). In contrast, total white blood cell counts were modestly but significantly elevated in KOMP −/− and Neo-deleted −/− mice (Fig. 4A; **P < 0.01), which was due to a significant increase in the number of circulating neutrophils (Fig. 4B; **P < 0.01). We observed no significant changes in the relative proportion of peripheral lymphocytes (Fig. 5C) or monocytes in mice of differing genotypes (Fig. 4D).

Fig. 4. Mice with C9orf72 mutations develop hematological phenotypes.

(A to H) Peripheral blood counts assessed for day 300+ KOMP and Neo-deleted animals. *P < 0.05, **P < 0.01, Tukey multiple comparisons. ns, not significant. (A) White blood cells. (B) Neutrophils. (C) Lymphocytes. (D) Monocytes. (E) Platelets. (F) Red blood cells (RBCs). (G) Hematocrit. (H) Mean corpuscular volume. (I to K) Peripheral blood counts assessed for aged Neo-deleted animals. (I) White blood cells. (J) Neutrophils. (K) Platelets.

Fig. 5. Mice with C9orf72 mutations show increased cytokines, chemokines, and autoantibodies.

(A and B) Analysis of 36 plasma cytokines and chemokines in day 300+ (A) KOMP and (B) Neo-deleted animals. *P < 0.05, **P < 0.01, Tukey multiple comparisons. (C and D) Anti-dsDNA antibody reactivity in plasma of (C) day 300+ KOMP and Neo-deleted animals and (D) aged Neo-deleted animals. *P < 0.05, **P < 0.01, Tukey multiple comparisons. ns, not significant. (E and F) Plasma from day 300+ Neo-deleted mice assessed for (E) IgM and (F) IgG reactivity against 124 self-antigens (26). Unsupervised hierarchical clustering grouped individual animals (x axis) and self-antigens (y axis).

We also noted that platelet count was significantly reduced in KOMP −/− and Neo-deleted −/− animals (Fig. 4E; **P < 0.01) and that there was a modest but significant reduction of RBCs in KOMP −/− and Neo-deleted −/− (Fig. 4F; **P < 0.01) animals that was accompanied by a modest but significant reduction in hematocrit (Fig. 4G; **P < 0.01). Mean corpuscular volume was also reduced in KOMP −/− and Neo-deleted −/− animals (Fig. 4H; **P < 0.01), consistent with modest microcytic anemia. Although there can be many causes for changes in red cell and platelet counts, similar phenotypes are seen in autoimmune hemolytic anemia and thrombocytopenic purpura, in which RBCs or platelets, respectively, are targeted by autoantibodies, leading to their destruction (23).

To understand how changes in the peripheral blood developed, we bled Neo-deleted animals (n = 163) ranging from 50 to 200 days of age (Fig. 4, I to K; +/+, n = 49; +/−, n = 65; −/−, n = 49). We noted that platelet count was already reduced by day 50 in −/− mice and was persistently depressed in older animals (Fig. 4K; *P < 0.05). Although young animals displayed normal neutrophil and white cell numbers, at 100 days of age, −/− mice began to exhibit a persistent and significant increase in neutrophil numbers (Fig. 4J; **P < 0.01), which was the primary contributor to a significant increase in total white cell counts (Fig. 4I; *P < 0.05). In contrast, numbers of lymphocytes in the blood remained unchanged relative to wild-type controls. Thus, neutrophilia, thrombocytopenia, and splenomegaly were highly penetrant phenotypes, which all developed between days 50 and 100 in −/− animals, suggesting an interconnected pathological relationship. In contrast, enlarged lymph nodes and immune cell infiltrates in the liver were observed later and at lower frequencies in mutant animals, suggesting that they could be secondary effects of changes to the hematopoietic compartment.

Loss of C9orf72 causes early and chronic cytokine induction

If the splenic and blood phenotypes that we observed in −/− animals developed in response to a chronic inflammatory process, we reasoned that their emergence might be associated with an increased abundance of inflammatory chemokines and cytokines. To test this idea, we measured the levels of 36 chemokines and cytokines in plasma from day 300 KOMP (Fig. 5A; +/+, n = 18; +/−, n = 8; −/−, n = 11) and Neo-deleted animals (Fig. 5B; +/+, n = 3; +/−, n = 11; −/−, n = 4). Eighteen of 36 cytokines and chemokines were significantly elevated in both KOMP −/− and Neo-deleted −/− animals relative to +/+ controls, including IL-22 (interleukin-22), IL-28, IL-23, IL-6, MCP-1 (monocyte chemoattractant protein 1), IL-31, IL-5, IL-10, IL-1β, IL-15/IL-15R (IL-15 receptor), IFN-γ (interferon-γ), IL-3, GM-CSF (granulocyte-macrophage colony-stimulating factor), IL-17A, IFN-α, MIP-1B (macrophage inflammatory protein-1B), LIF (leukemia inhibitory factor), and GROα (growth-related oncogene α) (Fig. 5, A and B; *P < 0.05, **P < 0.01). No cytokines or chemokines were significantly changed in KOMP +/− mice relative to controls (Fig. 5A), although we did note that IL-28 was significantly elevated and IL-2 was significantly reduced in +/− Neo-deleted animals relative to controls (Fig. 5B; *P < 0.05). We next analyzed cytokine levels in cohorts of mice at earlier time points. We found that already at 50 days of age, IL-31, IL-15/IL-15R, and GM-CSF were significantly elevated in Neo-deleted −/− mice relative to wild-type mice (fig. S11; *P < 0.05). By 100 days of age, Neo-deleted −/− animals displayed significantly increased levels of 19 of 36 cytokines and chemokines tested, fully recapitulating the pattern observed in older animals at both days 200 and 300 (fig. S11; *P < 0.05, **P < 0.01). Thus, the development of an inflammatory cytokine profile was an early event in mutant animals that either preceded or was coincident with other blood phenotypes and a decline in health.

Mutant animals rapidly develop persistent autoimmunity

We next considered the possibility that the changes in cytokine levels that we found were due to autoimmunity. Autoantibodies targeting double-stranded DNA (dsDNA) are a common diagnostic marker of autoimmune syndromes, including systemic lupus erythematosus (24). We found significant accumulation of anti-dsDNA antibody activity between 50 and 100 days of age in Neo-deleted −/− animals (Fig. 5D; **P < 0.01), which was sustained in plasma from KOMP −/− and Neo-deleted −/− animals greater than 300 days of age (Fig. 5C; *P < 0.05, **P < 0.01).

Although the accumulation of anti-dsDNA antibodies occurs under conditions of autoimmunity, it can also be indicative of rapid but transient cell death (25). To distinguish between these two possibilities, we used established autoantigen microarrays to test the plasma of day 300 Neo-deleted littermates for the presence and abundance of immunoglobulin M (IgM) and IgG antibodies targeting 124 autoantigens from disparate cell types (26) (Fig. 5, E and F). In −/− plasma, IgM autoantibodies targeting 117 of 124 antigens were significantly elevated relative to controls. Similarly, IgG autoantibodies targeting 113 of 124 antigens were significantly elevated in −/− animals (table S1; *P < 0.05). In +/− plasma, IgM autoantibodies targeting 1 of 124 antigens [liver kidney microsomal type 1 (LKM1)] were significantly elevated and IgG autoantibodies targeting 2 of 124 antigens (Collagen II and Mi-2) were significantly elevated relative to controls (table S1; *P < 0.05). Hierarchical clustering revealed that the pattern of IgM autoantibody activity in five of six −/− animals was significantly distinct from +/+ controls and two of four +/− animals (Fig. 5E). Intriguingly, the other two of four +/− animals displayed an intermediate pattern of IgM autoantibody reactivity, clustering with the remaining −/− mice (Fig. 5E). Clustering based on IgG-recognized autoantigens placed all mutants in one distinct group, with no obvious distinction between +/− and +/+ found in this case (Fig. 5F). Thus, we conclude that loss of C9orf72 results in the accumulation of a wide array of self-reactive antibodies, which is indicative of an autoimmune phenotype.

Because T regulatory cells restrict inflammation, we measured the percentage of CD25+ CD4+ cells, which are enriched for this T cell subtype, in spleens from 400-day-old Neo-deleted animals and found that CD25+ CD4+ cells were significantly elevated in −/− spleens relative to +/+ animals (fig. S12; **P < 0.01).

C9orf72 promotes tolerance in the immune system

The ImmGen database (27) indicates that C9ORF72 and its murine ortholog are expressed in blood cells. We therefore asked whether C9orf72 acts within the hematopoietic compartment. To address this, we performed bone marrow transplantation (BMT) between animals of distinct genotypes. Because of their inbred nature, we performed reciprocal BMT experiments using KOMP mice. To serve as phenotypic negative and positive controls, we transplanted −/− bone marrow into −/− recipients (−/−→−/−) (n = 6) and +/+ marrow into +/+ recipients (WT→WT) (Fig. 6A). To test whether a −/− genotype in bone marrow–derived cells was sufficient to cause autoimmunity, we transplanted −/− bone marrow into +/+ recipients (−/−→WT) (n = 12). Reciprocally, we asked whether loss of C9orf72 in the blood was necessary for the development of autoimmunity by attempting to ameliorate the phenotype of −/− animals with transplantation of +/+ marrow (WT→−/−) (n = 7). B6.SJL-PtprcaPepcb/BoyJ mice were used as wild-type bone marrow donors and recipients to facilitate tracking of reconstitution efficiency (Fig. 6A). No animals perished immediately after BMT. By 17 weeks, 29 of 30 transplanted animals exhibited >90% reconstitution of CD45+ cells, including CD11b+ Ly6G monocytes, CD11b+ Ly6G+ neutrophils, and B220+ B cells (Fig. 6, B and C, and fig. S13), indicating that engraftment of donor hematopoietic cells was efficient and sustained. However, we did note that CD3+ T cell reconstitution was incomplete, ranging from 40 to 80% donor-derived cells (Fig. 6, B and C, and fig. S13), which can be observed because of the relative radio resistance and long-lived nature of this cell type (28).

Fig. 6. C9orf72 acts in bone marrow–derived cells to prevent autoimmunity.

(A) Wild-type (WT) or C9orf72-deficient animals were lethally irradiated at day 110 and reconstituted with WT or mutant bone marrow. Recipient mice were regularly weighed, monitored for survival, and bled for whole-blood cell counts and plasma analyses, and animals were necropsied at end stage. (B and C) Quantification of flow cytometry–based assessment of 17-week posttransplant peripheral blood reconstitution. (B, C, and F to K) Each dot represents one mouse. (D) Survival curves for transplanted mice. *P < 0.05, generalized Wilcoxon test. IR, irradiation. (E) Average body weight ± SEM. *P < 0.05, Dunnett’s multiple comparisons. (F) End-stage spleen weight. (G) End-stage peripheral blood neutrophil counts. (H) End-stage peripheral blood platelet counts. (I) End-stage hematocrit. (J and K) Plasma anti-dsDNA antibody activity in animals at (J) day 230 (D230) and (K) end stage. (F to K) *P < 0.05, **P < 0.01, Tukey multiple comparisons. ns, not significant.

All WT→WT animals survived beyond 460 days of age (Fig. 6D) and progressively gained weight (Fig. 6E). In line with our previous findings (Fig. 2), we noted that mice lacking C9orf72 (−/−→−/−) failed to gain weight as fast as their wild-type (WT→WT) counterparts, a difference that became significant by 174 days of age (Fig. 7E; *P < 0.05). Moreover, all −/−→−/− animals (n = 6 of 6) died by 350 days, with a median survival of 293 days (Fig. 6D; *P < 0.05), and the causes of death or obligatory euthanasia in −/−→−/− animals were consistent with those observed previously (Fig. 2E, fig. S13, and video 1). At end stage, −/−→−/− mice had developed splenomegaly (Fig. 6F; **P < 0.01), neutrophilia (Fig. 6G; **P < 0.01), thrombocytopenia (Fig. 6H; **P < 0.01), and anemia (Fig. 6I; **P < 0.01). Anti-dsDNA antibody activity was significantly elevated in −/−→−/− at an age that preceded mortality (day 230) (Fig. 6J; **P < 0.01), as well as at end stage (Fig. 6K; **P < 0.01). Thus, lethal irradiation and BMT did not affect the development of autoimmunity in −/− animals.

Fig. 7. CRISPR/Cas9-induced mutations in C9orf72 lead to autoimmunity.

(A) Schematic showing the CRISPR/Cas9-targeting strategy that causes DNA double-strand breaks in exon 4 of C9orf72, resulting in several distinct mutations. (B) DNA sequences from CRISPR/Cas9-targeted mice indicated that 23 of 24 animals harbored a stop codon in exon 4 of the C9orf72 gene. (C) Survival of CRISPR/Cas9-targeted mice. (D) Spleens from a CRISPR/Cas9-targeted mouse at end stage and an age-matched C57BL/6 control. (E) Spleen weights for CRISPR/Cas9 mutant and KOMP mice. (F) Hematocrit for day 300+ CRISPR/Cas9-targeted animals. (G) Plasma anti-dsDNA antibody activity in day 300+ CRISPR/Cas9-targeted animals. (E to G) Each dot represents one mouse. *P < 0.05, **P < 0.05, Dunnett’s multiple comparisons. ns, not significant. (H) Mice harboring a 38-nucleotide deletion in exon 4 of C9orf72 were bred to heterozygosity (+/del) and homozygosity (del/del), as visualized by PCR of tail DNA using primers flanking the deletion site. (I) Plasma anti-dsDNA antibody reactivity in day 150 +/+, +/del, and del/del animals. **P < 0.01, Tukey multiple comparisons. (J) Proposed model for how C9orf72 may act in bone marrow–derived cells to limit fatal immune deregulation.

We asked which of the phenotypes found in −/− animals could be observed in +/+ animals receiving −/− bone marrow. We found that −/−→WT animals were significantly smaller than WT→WT animals at 206 days of age, after which their body weight plateaued (Fig. 6E; *P < 0.05). Strikingly, 9 of 12 −/−→WT mice died by 460 days of age with a median survival of 443 days (Fig. 6D; *P < 0.05 relative to WT→WT; video 2). At end stage, all −/−→WT (n = 9 of 9) had larger spleens (Fig. 6F; **P < 0.01), whereas a subset of −/−→WT mice (3 of 12) had developed enlarged cervical lymph nodes (fig. S13). Relative to WT→WT mice, −/−→WT animals had similar numbers of neutrophils (Fig. 6G) but displayed a significantly lower platelet count (Fig. 6H; **P < 0.01) and had significantly reduced hematocrit (Fig. 6I; **P < 0.01). IL-22, IL-31, and MIP-1B were significantly elevated in −/−→WT animals relative to WT→WT at end stage (fig. S13; *P < 0.05). Anti-dsDNA antibody activity was similar in −/−→WT and WT→WT animals at day 230 (Fig. 6J), yet by end stage, −/−→WT displayed significantly elevated anti-dsDNA antibody activity (Fig. 6K; *P < 0.05).

Next, we asked whether transplantation of WT bone marrow into −/− animals could improve their phenotype. By 230 days of age, WT→−/− mice had gained more weight than their −/−→−/− counterparts (Fig. 6E; *P < 0.05). Moreover, WT→−/− animals lived significantly longer than −/−→−/− animals, with the longest lived WT→−/− animals surviving to 452 days of age and with a median survival of 340 days, representing a 43-day extension of life span (Fig. 6D; *P < 0.05 relative to −/−→−/−; video 3). Splenomegaly was apparent in all WT→−/− animals at end stage, but on average, their spleens were smaller than −/−→−/− mice (Fig. 6F; **P < 0.01). Relative to −/−→−/− animals, WT→−/− mice had significantly fewer neutrophils (Fig. 6G; **P < 0.01) and significantly elevated platelet counts (Fig. 6H; **P < 0.01), but hematocrit was not significantly improved (Fig. 6I). IL-31, IL-6, MIP-1B, IL-10, IL-17A, and IL-15/IL-15R were significantly reduced in WT→−/− relative to −/−→−/− at end stage (fig. S13; *P < 0.05). Anti-dsDNA activity was significantly reduced in WT→−/− relative to −/−→−/− animals at day 230 (Fig. 6J; **P < 0.01) and at end stage (Fig. 6K; *P < 0.05).

C9orf72 mutant mice generated by CRISPR/Cas9 develop autoimmunity

We noted that some noncoding regions within the deleted locus showed a degree of conservation that raised the possibility that their disruption could contribute to the phenotypes observed in KOMP and Neo-deleted animals. To rule out this possibility, we used CRISPR/Cas9 technology to induce deletion mutations in exon 4 of C9orf72 (Fig. 7A), an exon present in all transcribed isoforms of both human and murine orthologs. Next-generation sequencing of PCR amplicons spanning the target site revealed disruption of C9orf72 exon 4 in 23 of 24 CRISPR/Cas9-targeted mice (Fig. 7B, fig. S14, and table S2). In 11 of 24 CRISPR/Cas9-targeted animals, we identified only premature stop codon mutations and no wild-type sequences, suggesting that these mice were compound heterozygous loss-of-function mutants for C9orf72 (Fig. 7B, fig. S14, and table S2).

All CRISPR/Cas9-targeted mice survived to adulthood, yet as they aged, their propensity to die increased, with only 10 of 24 animals surviving beyond 500 days of age (Fig. 7C). Splenomegaly was apparent in all mice (n = 7 of 7) necropsied (Fig. 7, D and E), whereas protrusions in the neck area indicative of enlarged cervical lymph nodes were apparent in 3 of 24 animals (fig. S14). Analysis of blood from mutant mice at day 310 revealed a significant reduction in hematocrit (Fig. 7F; **P < 0.01) and a significant increase in anti-dsDNA antibody activity (Fig. 7G; *P < 0.05).

Because CRISPR/Cas9-targeted animals were mosaic for mutations in C9orf72, we crossed F0 mice with C57BL/6 to isolate F1 progeny harboring one wild-type allele and one mutated allele (fig. S14). We then interbred F1 progeny containing a 38-nucleotide deletion in C9orf72 exon 4, predicted to generate a premature stop codon after 183 amino acids, to generate animals that were wild-type (+/+), heterozygous (+/del), and homozygous (del/del) for this particular mutation (Fig. 7H and fig. S14). We again found that C9orf72 del/del F2 mutants displayed significantly elevated anti-dsDNA antibody activity by day 150 (Fig. 7I; **P < 0.01).


Here, we report that eliminating function of mouse C9orf72 predisposes animals to fatal immune defects and autoimmunity. We observed that these classic features of autoimmunity (24, 29) developed synchronously, well before the onset of premature mortality. Furthermore, heterozygous animals were more susceptible than control mice to the generation of a limited repertoire of autoantibodies and were at increased risk of early mortality. Chronic inflammation is a known contributor to most, if not all, of the causes of death that we observed in +/− and −/− animals. Notably, the pattern of inflammatory cytokines chronically up-regulated in −/− animals included several members of the IL-23/IL-17 immune axis, which plays a pathogenic role in rheumatoid arthritis, psoriasis, Crohn disease, and multiple sclerosis (30).

Our findings led us to posit that C9orf72 serves an important function in the hematopoietic system. This hypothesis was substantiated by transplant studies in which mutant bone marrow transferred autoimmune phenotypes to wild-type recipients and wild-type bone marrow suppressed immune deregulation and significantly extended life span in mutant animals. However, although the phenotype of mutant animals that received wild-type marrow was improved relative to mutant mice receiving mutant bone marrow, their phenotype was still substantially worse than wild-type animals receiving wild-type bone marrow. Failure of wild-type bone marrow to completely rescue mutant mice leaves open the possibility that C9orf72 also functions in a radiation-resistant cell population, such as microglia, which are peripheral antigen-presenting cells, the dysfunction of which could lead to improper priming of adaptive immune responses. C9orf72 might also function in the thymic epithelium, which is critical for the establishment of peripheral tolerance. Alternatively, T cells themselves are well-known modulators of immunity, and the incomplete replacement of this compartment after transplantation provides another potential explanation for the incomplete rescue observed. Our findings, and those that indicate that T cell function becomes compromised in mice harboring mutations in additional ALS-implicated genes (19, 31), suggest that investigating whether C9orf72 functions in this cell type is warranted. Although we observed an increase in the abundance of CD25+ CD4+ cells, which are enriched for T regulatory cells, in the spleens of mutant animals, whether mutant T regulatory cells harbor functional deficits that contribute to the phenotypes that we observed remains to be determined. Elevation of T regulatory cells could represent a compensatory reaction to the reduced functionality of this cellular subset or a homeostatic response to persistent inflammation.

We note that our study does have certain limitations and leaves several questions unresolved. Although our studies demonstrate that bone marrow cells play a role in the phenotypes that we observed and suggest that C9orf72 functions in hematopoietic derivatives of the bone marrow, we cannot rule out the possibility that C9orf72 acts through marrow-resident mesenchymal cells. Additionally, we did not perform an exhaustive census of transplanted animals to monitor the extent of engraftment into solid tissues, including the brain. Therefore, we cannot address whether incomplete imposition or rescue of mutant phenotypes after transplant might be due to variable infiltration of cells into a given solid organ (32) rather than due to the function of C9orf72 in nonhematopoietic cell types. In the future, it will be important to resolve the identity of the specific cell types in which C9orf72 functions to preserve a normal immune response. We previously showed that C9orf72 was highly expressed in neuronal cell types that are most sensitive to cell death in ALS and FTD but is more modestly expressed in microglia. Despite the high level of C9orf72 expression in the neuronal lineage, the data we report here suggest that this gene acts through bone marrow–derived cells in which it is expressed. It remains to be determined whether C9orf72 serves some nonessential function in neurons that express it.

Recently, haploinsufficiency for TBK1, a kinase functioning in IFN signaling and selective autophagy (33), has been shown to cause ALS and FTD [reviewed in (34)]. Tbk1 −/− mice display an embryonic lethal phenotype; however, deletion of this gene on an outbred background leads to monocytosis, splenomegaly, and infiltration of immune cells into the skin, lung, liver, and kidneys (35). Cre-mediated ablation of Tbk1 in T cells resulted in their activation and overaccumulation in the spleen and lymph nodes (19). These phenotypic commonalities observed upon loss of Tbk1 and C9orf72 are striking. Whether these gene products act in a common pathway requires further investigation; however, it is notable that at least one report has suggested that C9ORF72 may also function in autophagy (11).

Whether malfunctions in the immune system directly contribute to degenerative phenotypes in ALS patients remains unresolved (36, 37). Studies have identified infiltrating T cells and hyperactivation of resident immune cells in the postmortem spinal cords of ALS patients (38, 39). Patients with a rapid course of disease were also found to have fewer circulating T regulatory cells than did individuals with slower disease progression (40), and levels of the proinflammatory cytokines IL-17 and IL-23 were found to be elevated in plasma and cerebral spinal fluid from at least one ALS cohort (41). Recently, a large epidemiologic study found substantially higher rates of autoimmune history in patients eventually developing ALS (42). However, studies of autoantibodies in ALS patients have been mixed, with some finding evidence of their presence, whereas others have not (36, 37). The ever-improving understanding of the genetic contributors to ALS suggests that revisiting immune phenotypes in patients who have been stratified by C9ORF72, TBK1, and other genotypes could now be warranted.

There are several other phenotypic studies of C9orf72-deficient mice (43, 44). These reports provide direct confirmation of our finding that C9orf72 mutant animals develop splenomegaly and lymphadenopathy. Furthermore, Atanasio and colleagues found, as we report here, that mutant animals exhibited elevated plasma cytokines and autoantibodies (43). However, there were a number of differences between our findings and these recent studies that warrant future resolution. Most notably, although both we and O’Rourke et al. generated C9orf72 mutant animals using the same targeting vector generated by the KOMP Consortium, heterozygous and homozygous mutants in our colony displayed an increased risk of mortality whereas mutants in their colony did not. This discrepancy suggests that either environmental factors or differences in genetic background may influence the longevity of mice lacking C9orf72, as is the case with other models of autoimmunity (29). Additionally, both Atanasio et al. and O’Rourke et al. found an increase in macrophage number in mutant animals and suggested that inflammatory phenotypes found in these mice may be due to dysfunction in this or other phagocytic cell types. Our conclusion that C9orf72 functions in bone marrow–derived cells is consistent with this idea. However, we also observed a much broader proliferative disorder affecting both myeloid and lymphoid lineages, leaving open the question of whether this gene may act in additional hematopoietic cell types. In the future, conditional ablation of C9orf72 in specific blood lineages should address where precisely this gene functions to suppress pathological inflammation.

Overall, our findings have implications for planned therapeutic interventions in ALS patients harboring the C9ORF72 repeat expansion. Regardless of whether loss of function in C9ORF72 plays a central role in the development of ALS, our experiments do strongly suggest that therapeutic efforts to reduce expression of the repeat expansion, such as with antisense oligonucleotides (9), should be designed and carried out with caution. If not, chronic depression of the limited quantity of normal gene product that still remains in these ALS patients could occur, potentially resulting in autoimmunity.


Study design

The goal of this study was to characterize the consequences of C9orf72 loss-of-function mutations in mice. The experimental design involved long-term survival studies paired with age-matched and genotype-matched histologic, cellular, and biofluidic analysis. Predefined end-point criteria included greater than 15% loss of body weight, signs of distressed breathing, enlarged abdomen, severe dermatitis, lymph nodes greater than 15 mm that inhibited mobility, inability of the mouse to right itself, or severe prolapse. Exploratory experiments were performed in excess of five mice per genotype. Sample sizes for BMT studies were calculated using G-power analysis based on previously defined effect sizes. Donor and recipient mice were randomized into treatment groups, and all outliers were included in data analysis. All experimental protocols and procedures were approved by the Animal Committee of Harvard University.

RNA extraction and quantitative PCR

Total RNA was isolated using TRIzol (Gibco). First-strand complementary DNA was synthesized using iScript (Bio-Rad). C9orf72 transcript abundance was analyzed using SYBR Green on CFX96 Real-Time System (ABI). C9orf72 exon 6 to 8 primers are as follows: 5′-GCAGTGCAGAGAAAGTAAATAAGATAG-3′ and 5′-ACTGCCTGTTGCATCCTTTAG-3′.

Fixation, sectioning, and tissue staining

Mice were anesthetized with Avertin and perfused with phosphate-buffered saline and then with 4% paraformaldehyde (PFA). Tissues were harvested and postfixed in 4% PFA overnight at 4°C. Tissue was dehydrated, embedded in paraffin, sectioned at 6 μm, and stained for hematoxylin and eosin. Sections were examined in a blinded manner and imaged using Zeiss AX10.

Western blotting

Total protein was extracted from a frozen brain with a reducing sample buffer (radioimmunoprecipitation assay) containing complete inhibitor cocktail (Roche). Protein (15 μg) was separated on 4 to 20% SDS–polyacrylamide gel, transferred to polyvinylidene difluoride membrane, blocked with 5% bovine serum albumin, incubated with primary antibodies [anti–C9ORF72–N-terminal (Proteintech), anti–C9ORF72–C-terminal (Abgent), and anti–α-tubulin (Abcam)] and detected by enhanced chemiluminescence (GE Amersham).

Blood and cytokine measures

Peripheral blood was collected via mandible puncture into EDTA-coated tubes. Blood counts were assessed using a Hemavet (Abaxis) and then centrifuged to pellet cells, and plasma was harvested from the supernatant. Luminex-based multiplexed fluorescence assay was used to assess 36 cytokines and chemokines.

Autoantibody profiling

Plasma was diluted 1:200 and assessed using a mouse α-dsDNA total-Ig kit (Alpha Diagnostic International). Antibodies against 124 autoantigens were measured on autoantigen arrays (26).

Spleen and lymph node fluorescence-activated cell sorting

Spleens were mashed between two glass slides and passed repeatedly through a 25 5/8G needle to dissociate single cells. After RBC lysis, cells were quantified using a Countess (Invitrogen). Lymph nodes were processed as above without RBC lysis. Cells were stained using the following antibodies (BioLegend): B220 (RA3-6B2), CD19 (6D5), κ (RMK-45), λ (RML-42), CD11c (N418), CD8α (53-6.7), CD3ε (145-2C11), CD4 (GK1.5), CD62L (MEL-14), CD44 (IM7), CD25 (PC61), Pan-NK (DX5), CD11b (M1/70), Ly6G (1A8), CD115 (AFS98), F4/80 (BM8), and 4′,6-diamidino-2-phenylindole (Sigma).

Bone marrow transplantation

Female B6.SJL-PtprcaPepcb/BoyJ (The Jackson Laboratory) mice were used as age-matched wild-type bone marrow donors and recipients. Female KOMP C9orf72 −/− mice were used as age-matched bone marrow donors and recipients. At D110, marrow from femur and tibias were harvested from two wild-type and two −/− donors, flushed, single cell–dissociated, counted, pooled, and diluted to 5E6 cells/100 μl. Recipient animals received two rounds of 550 rad separated by 3 hours. 5E6 bone marrow cells were injected into the tail vein. The recipients were maintained on pH 3.0 water for 2 weeks. Reconstitution efficiency in RBC-lysed peripheral blood at 17 weeks after transplant was measured using the following antibodies (BioLegend): CD45.2 (104), CD3ε (145-2C11), B220 (RA3-6B2), Ly6G (1A8), CD11b (M1/70), and CD45.1 (A20).

CRISPR/Cas9 sequencing

Tail DNA was PCR-amplified with primers containing flanking adapters, followed by PCR barcoding with 24 unique barcodes. Samples were pooled, cleaned (Promega Wizard DNA clean-up kit), sequenced via MiSeq, and analyzed (45).


All statistical calculations were performed using GraphPad Prism. Tests between two groups used two-tailed Student’s t test. Tests between multiple groups used one-way analysis of variance (ANOVA) with Tukey multiple comparisons. Tests between multiple groups over time used two-way ANOVA with Dunnett’s multiple comparisons. Survival curves were evaluated by generalized Wilcoxon.



Fig. S1. Validation of C9orf72 loss-of-function allele.

Fig. S2. Analysis of spinal motor neurons.

Fig. S3. Motor cortex histology.

Fig. S4. Histology of thalamus, CA1, and cerebellum.

Fig. S5. GFAP expression in the central nervous system.

Fig. S6. Individual mouse weight curves after disease onset.

Fig. S7. Splenocyte-gating scheme.

Fig. S8. Bone marrow histology and cell counts.

Fig. S9. Analysis of B and T cells in cervical lymph node.

Fig. S10. Hematopoietic liver infiltrates occur in a subset of mutant animals.

Fig. S11. Analysis of cytokines and chemokines.

Fig. S12. Analysis of CD4+ CD25+ cells.

Fig. S13. Cellular and organismal phenotypes after bone marrow transplant.

Fig. S14. CRISPR/Cas9-targeted mutations in exon 4 of C9orf72.

Table S1. Analysis of autoantibodies in Neo-deleted mice.

Table S2. CRISPR/Cas9-induced mutations in C9orf72 exon 4.

Video 1. End-stage −/− marrow donor to −/− recipient.

Video 2. End-stage −/− marrow donor to WT recipient.

Video 3. End-stage WT marrow donor to −/− recipient.


Acknowledgments: We thank the Eggan laboratory, the HSCRB (Harvard Department of Stem Cell and Regenerative Biology) Histology Core, and M. Charlton, A. Wagers, D. Scadden, and K. Hochedlinger for advice, technical support, and manuscript review. Funding: K.E. was supported by the Howard Hughes Medical Institute, p2ALS, Target ALS, and NIH5R01NS089742. F.T.M. was supported by the NIH (5K99NS083713), the Wellcome Trust, the Academy of Medical Sciences, and the Medical Research Council (MR/P501967/1). Author contributions: K.E. conceived the project; K.E., A.B., and N.S. designed the experimental plan. A.B., N.S., J.-Y.W., A.S., K.K., and S.S.-U. supported animal experiments. R.M. supported DNA sequence analysis. D.A.M. evaluated staining. S.G. and Q.-Z.L. performed autoantibody array. A.S., M.H.S., J.J.T., D.J.R., L.Z., and L.D.N. performed hematological analysis. A.B. and F.T.M. performed CRISPR/Cas9 design. A.B., N.S., R.M., and K.E. wrote the manuscript. K.E. supervised the project. Competing interests: The authors declare that they have no competing financial interests. Data and materials availability: Whole-genome sequencing data for this study have been deposited in Sequence Read Archive SRP073407.
View Abstract

Navigate This Article