IFITM3 protects the heart during influenza virus infection

Generation of the IFITM3 KO Mouse Model.

IFITM3 KO mice used in previous infection research are of a mixed genetic background and contain a fluorescent protein insertion at the IFITM3 locus (32). To generate a mouse model with a clean genetic deficiency on a pure C57BL/6 background, we used a CRISPR/Cas9-based deletion strategy involving 2 guide RNAs with unique sequence identity for a region of exon 1 of the Ifitm3 gene (Fig. 1A). Mutant pups with an Ifitm3 deletion were identified by PCR (Fig. 1B). In 1 targeted animal, sequencing confirmed deletion of a 53-bp region of exon 1. As a specificity control, RT-PCR for full-length IFITM1, IFITM2, and IFITM3 coding sequences was performed on mRNA from WT and mutant (termed IFITM3 KO) mouse embryonic fibroblasts (MEFs). The resulting PCR products showed a band shift in the KO cells for IFITM3 only (Fig. 1C), and DNA sequencing of the PCR products confirmed that IFITM1 and IFITM2 sequences were not mutated. Despite low probability scores, we DNA-sequenced the top 5 off-target sites for each guide RNA as predicted by Benchling software and did not observe any mutations (SI Appendix, Fig. S1). The Ifitm3 deletion introduces a nonsense mutation at codon 18 and a subsequent stop codon at position 37 (Fig. 1A). Importantly, motifs essential for IFITM3 antiviral activity, membrane association, localization, and dimerization are not contained within its N-terminal 17 amino acids (23, 27, 31, 33, 34). IFITM3 protein could not be detected by immunoblotting of tissue or cell lysates derived from the KO mice, even on IFNβ treatment (Fig. 1 D and E).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Generation and validation of our IFITM3 KO C57BL/6 mouse model. (A) C57BL/6 zygotes were injected with a mix of Cas9 mRNA and 2 guide RNAs to target exon 1 of the Ifitm3 gene. As a result of a 53-bp deletion and consequent frame shift, the mutant Δ53 IFITM3 KO gene bears a nonsense mutation at codon 18 and a subsequent stop codon at position 37. The protein depicted for the mutant is hypothetical. (B) Example genotyping PCR on genomic DNA from WT, heterozygous (het), and KO mice. Sequencing of the PCR products revealed a 53-bp deletion in the KO allele. (C) RT-PCR for IFITM1, IFITM2, and IFITM3 performed on mRNA extracted from IFNβ-treated MEFs obtained from mice of the indicated genotypes. Sequencing of the PCR products confirmed the 53-bp deletion in the IFITM3 mRNA in IFITM3 KO cells and no mutation of IFITM1 and IFITM2 sequences. (D) Western blot analysis of tissue lysates from WT and IFITM3 KO mice. (E) MEFs derived from embryos of the indicated genotypes were treated with IFNβ or mock-treated, and lysates were subjected to Western blot analysis. (F) MEFs treated as in E were infected with IAV or VSV for 24 h, and percent infection was determined by flow cytometry. Graphs depict normalized measurements from at least 3 separate experiments. Relevant comparisons were analyzed by unpaired t tests as indicated by lines. *P < 0.001. NS, not significant.

Given the well-characterized role of IFITM3 in blocking influenza virus infection of cells (35, 36), we examined the susceptibility of MEFs derived from WT and KO mice to infection with H1N1 IAV virus (PR8 strain). KO cells were significantly more susceptible to infection than WT cells even after IFN treatment, which caused decreased susceptibility to infection in WT cells (Fig. 1F and SI Appendix, Fig. S2). We observed similar trends on influenza virus infection of WT and IFITM3 KO bone marrow-derived macrophages (SI Appendix, Fig. S3). These results confirm previous experiments from our group and others indicating that IFITM3 is responsible for a significant portion of the anti-influenza response induced by type I IFN (35, 37, 38).

We next examined vesicular stomatitis virus (VSV) infection of WT and KO MEFs. VSV is highly susceptible to inhibition by IFN, and although it is reported to be mildly inhibited when IFITM3 is highly overexpressed, VSV has not been tested for susceptibility to inhibition by endogenous levels of mouse IFITM3 (39). Unlike influenza virus infections, WT and KO cells were infected at similar rates by VSV, and VSV infection was fully restricted by IFN treatment regardless of IFITM3 status (Fig. 1F and SI Appendix, Fig. S2). These results indicate that endogenous baseline IFITM3 does not inhibit VSV infection, and that IFITM3 is not a significant contributor to the type I IFN-induced inhibition of VSV in MEFs.

These infection experiments provide important controls demonstrating that cells derived from IFITM3 KO animals are not indiscriminately susceptible to all viruses, and that the antiviral IFN response is otherwise functional in IFITM3 KO cells.

Increased Morbidity and Mortality of IFITM3 KO Mice on Influenza A Virus Infection.

To examine the pathogenesis of IAV infection in IFITM3 KO versus WT mice, we infected mice intranasally with the PR8 strain at a dose of 10 TCID₅₀, the lowest dose that allowed reproducible infections in our experiments. This dose caused weight loss in WT mice, but the mice began to recover by day 12 postinfection (Fig. 2A). KO mice lost significantly more weight starting at day 4 postinfection and continuing throughout the remainder of infection (Fig. 2A). By day 12 postinfection, all KO mice had succumbed to infection or had lost more than 30% of their body weight and were euthanized (Fig. 2B). Virus titers in the lungs measured on days 5, 7, and 10 postinfection were found to be higher in KO versus WT mice at each time point (Fig. 2C). Relative to earlier time points in WT mice, day 10 virus titers were decreased, indicating adaptive immune clearance of the virus (Fig. 2C). In contrast, titers in the IFITM3 KO lungs on day 10 remained high (Fig. 2C). We also examined levels of the representative inflammatory cytokine IL-6 in lung homogenates at these postinfection time points and found higher levels of IL-6 in the KO lungs compared with WT lungs at all time points (Fig. 2D). Overall, relative to WT mice, IFITM3 KO mice were more severely infected by influenza virus strain PR8.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

IFITM3 KO mice experience increased morbidity and mortality on influenza virus infection. (A and B) WT (n = 29) and IFITM3 KO (n = 21) mice were intranasally infected with IAV strain PR8 (10 TCID₅₀) and followed daily for weight loss (A) and survival (B). Points in A depict mean values, and error bars represent SD of the mean. *P < 0.01, unpaired t test. (C) Infected mice were killed on the indicated days for TCID₅₀ measurement of virus titers in the lung. Data points collected for days 5 and 10 were from 2 independent experiments. (D) ELISA quantification of IL-6 in lungs collected on the indicated days postinfection. In C and D, each point represents an individual mouse, and bars represent mean values. Error bars represent SD of the mean. *P < 0.001, unpaired t test.

Influenza Virus Replicates to High Levels in the Heart in the Absence of IFITM3.

We next investigated whether in addition to the lung, virus disseminated to higher levels in other organs of IFITM3 KO mice, including the heart. We quantified virus titers in the hearts of infected animals on days 5, 7, and 10 postinfection. We observed a modest increase in virus titers in the hearts of KO mice compared with WT mice on day 5 postinfection, followed by markedly higher titers on days 7 and 10 (Fig. 3A). As seen in the lung, on day 10 postinfection, the virus titers remained high in hearts of KO mice but had declined substantially in WT mice (Figs. 2C and 3A). In fact, in WT mice, the low virus levels in the heart on days 5 and 7 were cleared to below the limit of detection by day 10, but remained at a titer similar to that seen on day 7 in KO mice (Fig. 3A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

IFITM3 KO mice show uncontrolled influenza virus replication in the heart. WT and IFITM3 KO mice were intranasally infected with IAV strain PR8 (10 TCID₅₀). (A) Infected mice were killed on day 5, 7, or 10 postinfection for TCID₅₀ measurement of virus titers in the heart. Data points collected for days 5 and 10 were from 2 independent experiments. Each point represents an individual mouse, and bars represent mean values. Error bars represent SD of the mean. *P < 0.05, ***P < 0.0001, unpaired t test. ND, not detected. (B) Representative images of heart sections from mice killed on day 10 postinfection. Green, anti-influenza virus nucleoprotein (NP) staining; blue, DAPI; gray, brightfield imaging. (Scale bar: 10 μm.) (C) Infected mice were killed on day 10 postinfection for TCID₅₀ measurement of virus titers in the spleen, liver, kidney, and brain. Each point represents an individual mouse, and bars represent mean values. Error bars represent SD of the mean. ***P < 0.0001, unpaired t test. ND, not detected.

To examine whether cells in the heart were directly infected with influenza virus, we stained heart sections from WT and KO mice on day 10 postinfection with anti-influenza virus nucleoprotein antibody. We observed minimal staining of the WT hearts, but found staining of elongated cells consistent with a cardiac fibroblast morphology and localization in KO hearts (Fig. 3B). We next costained heart sections for virus antigen and vimentin, a marker of cardiac fibroblasts, and found that virus infection in KO hearts was particularly apparent in enlarged vimentin-positive cells (SI Appendix, Fig. S4). Interestingly, in WT hearts, IFITM3 is highly expressed in vimentin-positive cardiac fibroblasts (SI Appendix, Fig. S5A). Unexpectedly, we observed a small decrease in the number of cardiac fibroblasts in 4 out of 6 KO hearts at day 10 postinfection (SI Appendix, Fig. S5B). This modest depletion may be due to direct infection of the fibroblasts, which could promote cell death.

The high level of virus in the lungs and hearts of KO animals prompted us to examine whether virus disseminated to other organs. We observed a significant increase in spleen viral titer in the KO mice at day 10 postinfection compared with WT, but detected no virus in the liver, kidney, or brain of either WT or KO mice (Fig. 3C). The differences between WT and KO mice show that virus replication is inadequately controlled in each of the IFITM3 KO organs in which we detected influenza virus, including the heart.

Influenza Virus Infection Causes Cardiac Electrical Dysfunction in IFITM3 KO Mice.

Given that replication of influenza virus to high titers in the heart has been seen only rarely in mouse models (9, 10), we sought to examine the effect of the high virus loads in the hearts of IFITM3 KO mice on cardiac electrical function. First, we determined that IFITM3 KO mice exhibit cardiac function and activity at baseline comparable to that seen in WT mice as measured by both echocardiography and electrocardiography (ECG) (SI Appendix, Fig. S6). Next, we performed ECG measurements on WT and KO mice after infection with PR8 as previously established (Fig. 2). Both WT and KO mice showed depressed heart rates and increased RR interval (i.e., the interbeat time interval between the major R peaks of the ECG traces) averages at day 6 postinfection (Fig. 4A). WT mice returned to a normal heart rate and average RR interval duration by day 10, while the electrical activity of the KO mice continued to decline significantly (Fig. 4A). Importantly, these differing cardiac measurements correlate with high virus titers detected in the KO hearts and with virus clearance in the WT hearts (Fig. 3A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

IFITM3 KO mice suffer from cardiac electrical dysfunction following influenza virus infection. WT and IFITM3 KO mice were intranasally infected with IAV strain PR8 (10 TCID₅₀). (A) ECG measurements over the time course of infection. Except for day 25, data were collected over at least 3 independent experiments. Each point represents an individual mouse, and bars represent mean values. Error bars represent SD of the mean. *P < 0.05, unpaired t test. (B) Example ECG readings from WT and KO mice preinfection and postinfection with PR8. Selected RR intervals of the infected mice are highlighted by blue (WT) or green (KO) double arrows. (C) RR interval times plotted for a representative WT mouse and a representative IFITM3 KO mouse on day 10 postinfection. (D) RR ranges, defined as the difference between the longest and shortest RR intervals over an ECG measurement period of 5 min, were calculated for individual mice on day 10 postinfection. Each point represents an individual mouse, and bars represent mean values. Error bars represent SD of the mean. ***P < 0.0001, unpaired t test.

In addition to higher average RR interval values in the infected KO mice, we observed highly irregular RR intervals late in infection in KO mice with no discernible pattern, while RR intervals in WT mice remained regularly paced, comparable to those seen in mock-infected animals (Fig. 4B). Plotting of individual RR interval times revealed a higher level of variability in KO mice compared with WT mice (Fig. 4C). Indeed, RR interval ranges calculated for 5-min ECG readings in multiple mice confirmed a significant arrhythmia that was consistently observed in KO mice (Fig. 4D).

We next performed infections with 2 less pathogenic strains of IAV compared with PR8 to examine whether these strains could also cause cardiac electrical dysfunction. These included X31, a PR8 reassortant virus containing the HA and NA segments from A/Hong Kong/1/1968 (H3N2) and A/WSN/1933 (H1N1) known as WSN. We chose a dose of X31, 1,000 TCID₅₀, that caused measurable weight loss in WT mice (SI Appendix, Fig. S7A). This dose resulted in a greater decline in weight in KO mice and delayed recovery (SI Appendix, Fig. S7A). However, infection with this less pathogenic strain did not result in significant cardiac electrical changes (SI Appendix, Fig. S7 B and C), and thus we did not investigate X31 further. For WSN infections, we used a dose of 200 TCID₅₀, which caused weight loss in WT mice and was more virulent in KO mice, but did not cause lethality (SI Appendix, Fig. S8A). WSN infection caused moderate electrical dysfunction, including decreased heart rate and irregular RR intervals, by day 8 postinfection that abated by day 10 postinfection as the mice began to recover (SI Appendix, Fig. S8 B–E). These WSN infection data, along with data from WT mice infected with PR8 (Fig. 4A and SI Appendix, Fig. S8B), indicate that mild or moderate cardiac dysfunction during influenza virus infection is reversible. Furthermore, these results show that the inherent virulence of individual virus strains, along with the increased susceptibility caused by IFITM3 KO, determine the severity of cardiac dysfunction during influenza virus infection.

IFITM3 Protects Mice from Excess Collagen Deposition in the Heart during Infection.

Cardiac fibrosis results from excess fibroblast secretion of extracellular matrix components, particularly collagen, and unresolved collagen accumulation has been shown to reduce the efficiency of electrical conductivity of the heart (40, 41). Thus, we hypothesized that fibrotic collagen deposition may be linked to the severe cardiac electrical defects observed in PR8-infected IFITM3 KO mice. We examined heart sections from PR8-infected WT and KO mice at day 10 postinfection after Masson’s trichrome staining, which allows visualization of collagen via blue staining. We observed significantly more blue staining in the hearts of KO animals, indicating an increased fibrotic response (Fig. 5 A and B). Importantly, we did not observe baseline fibrosis in mock-infected WT or IFITM3 KO hearts (Fig. 5 A and B). Given that we observed direct infection and depletion of cardiac fibroblasts in infected IFITM3 KO mice (SI Appendix, Figs. S4 and S5), this fibrotic response cannot be explained by an increase in fibroblast numbers, suggesting that activation, rather than overall numbers of fibroblasts, may underlie this phenotype.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Cardiac fibrosis in influenza virus infection of IFITM3 KO mice. WT and IFITM3 KO mice were intranasally infected with influenza virus strain PR8 (10 TCID₅₀) or were mock-infected. (A) Hearts were collected on day 10 postinfection, and sections were stained with Masson’s trichrome stain, in which blue staining is indicative of fibrotic collagen deposition. A representative infected heart and a representative mock-infected sample are shown for each genotype. Boxed areas are regions magnified in the far-right images. (B) Blue pixel percentages from images as in A were quantified for mock-infected and infected WT and KO hearts using ImageJ software. (C and D) qRT-PCR on mRNA extracted from hearts collected at day 10 postinfection was performed to assess expression of Col1A2 (C) and Tgfb1 (D). (E) IL-6 ELISA was performed on heart homogenates collected on day 10 postinfection. In B–E, each point represents a heart from an individual mouse, and bars represent mean values. Error bars represent SD of the mean. *P < 0.05, ***P < 0.0001, unpaired t test.

We also confirmed via qRT-PCR analysis that collagen gene (Col1A2) expression was increased in infected KO hearts at day 10 postinfection (Fig. 5C). Since collagen up-regulation is often associated with induction by TGFβ signaling, we also examined Tgfb1 gene expression in the hearts of infected animals and found that it was also up-regulated in KO hearts compared with WT hearts (Fig. 5D). In addition, we measured IL-6 levels in heart tissue after infection, because we observed infection of cardiac cells via microscopy analysis (Fig. 3B), and IL-6 is induced directly by influenza virus infection via activation of the RIG-I signaling pathway (42), and because IL-6 is associated with cardiac fibrosis via TGFβ induction in cardiac fibroblasts (43). We detected significantly higher amounts of IL-6 protein as measured by enzyme-linked immunosorbent assay (ELISA) in heart homogenates from infected KO mice compared with WT mice (Fig. 5E). In sum, these data show that fibrotic pathways are activated at higher levels in the hearts of infected KO mice, resulting in excess collagen deposition. These experiments showing an enhanced fibrotic response further highlight the critical cardiac protection afforded by IFITM3 during IAV infection, thus identifying a new essential role for this innate immunity protein.