Type I interferon response impairs differentiation potential of pluripotent stem cells

lab-led

Forcing an interferon-stimulated gene program in human induced pluripotent stem cells with a constitutively active IRF7 produces lasting transcriptional change and impaired germ layer differentiation, supporting the proposal that the canonical type I interferon system and pluripotency are difficult to hold simultaneously.

Julie Eggenberger; Daniel Blanco-Melo; Maryline Panis; Kristen J. Brennand; Benjamin R. tenOever

2019 · Proceedings of the National Academy of Sciences · primary research

Senior authors

Benjamin R. tenOever

Correspondence

Benjamin R. tenOever

Research areas & themes

Citation

Eggenberger J, Blanco-Melo D, Panis M, Brennand KJ, tenOever BR. Type I interferon response impairs differentiation potential of pluripotent stem cells. Proceedings of the National Academy of Sciences. 2019. Volume 116, issue 4, pages 1384-1393.

DOI 10.1073/pnas.1812449116. PMID 30606801. PMCID PMC6347712.

One-sentence contribution

Forcing an interferon-stimulated gene program in human induced pluripotent stem cells with a constitutively active IRF7 produces lasting transcriptional change and impaired germ layer differentiation, supporting the proposal that the canonical type I interferon system and pluripotency are difficult to hold simultaneously.

Executive summary

Differentiated vertebrate cells detect viral replication intermediates through pattern recognition receptors and respond by inducing type I interferon and hundreds of interferon-stimulated genes. Pluripotent stem cells do neither, failing to induce interferon in response to viral RNA or infection and responding only weakly to interferon beta itself. Why cells of such developmental importance forgo this defence was unresolved. The authors built a matched primary cell system in which human foreskin fibroblasts were reprogrammed to induced pluripotent stem cells and then redifferentiated to fibroblasts, so pluripotent and differentiated states of one genetic background could be compared directly. Only the pluripotent state failed to respond to 5 prime triphosphate RNA, to interferon beta, or to influenza A virus lacking its interferon antagonist NS1, and the same behaviour was seen in mouse embryonic stem cells. Co-expression of individual reprogramming factors with a constitutively active IRF7 in cells competent to respond showed that KLF4, and to a lesser degree SOX2 and OCT4, suppress induction of interferon-stimulated genes, with KLF4 the most potent. To ask what happens when the defence is nonetheless engaged, the authors placed that active IRF7 under doxycycline control in pluripotent stem cells. A brief pulse induced interferon-stimulated genes and altered morphology, and after five days of rest roughly 2,000 genes remained differentially expressed even though IRF7 and its direct targets had returned to baseline. Differentiation after such a pulse was compromised for ectoderm and endoderm and dysregulated within the mesoderm lineage, with cardiomyocyte cultures showing more beating alongside roughly 5,000 differentially expressed genes.

Scientific context

Type I interferon and the interferon-stimulated genes it induces constitute a broadly conserved vertebrate defence capable of fully protecting cells from many viruses. Pluripotent stem cells stand apart. Earlier work by others had established that mouse embryonic stem cells and human pluripotent cells respond poorly to viral and bacterial pathogen-associated molecular patterns, that interferon beta treatment produces little transcriptional output in these cells, and that low pattern recognition receptor levels can account for only part of the picture. Despite this, pluripotent cells are not unusually susceptible to infection, an observation other groups had attributed variously to stem-cell-specific factors, to RNA interference, and to constitutively high baseline expression of a subset of interferon-stimulated genes. What remained unaddressed was why the system is unused rather than simply how it is blocked. The present study reframes the question as one about compatibility, asking what the consequences of switching the defence on would be for the pluripotent state itself.

Central question

Why do pluripotent stem cells not use the canonical type I interferon system, and specifically is the maintenance of pluripotency compatible with engagement of an interferon-stimulated gene program?

Experimental strategy

Three design choices carry the argument. First, comparison is made within a single genetic background by reprogramming human primary foreskin fibroblasts to induced pluripotent stem cells and then redifferentiating them back to fibroblasts, which controls for clonal selection and loss of genetic material as explanations for unresponsiveness. Second, the block is probed from the direction of the reprogramming factors themselves, co-expressing OCT4, SOX2 and KLF4 individually with a constitutively active IRF7 in cells that normally respond, so that repression can be attributed to a defined factor rather than to the pluripotent state as a whole. The IRF7 construct, IRF7 delta, lacks residues 247 to 467 and therefore bypasses the requirement for phosphorylation by the IKK-related kinases, allowing the interferon-stimulated gene program to be driven without upstream signalling. Third, and centrally, the defence is forced on in pluripotent cells using a doxycycline-inducible lentiviral cassette expressing either IRF7 delta or GFP, which supplies a matched control for the induction procedure itself. Because IRF7 engages degenerate as well as perfect interferon-stimulated response elements, it can drive a wide subset of these genes independently of interferon signalling, which matters in cells where interferon signalling is itself inert. Consequences are then read at three depths. Immediate transcriptional output at forty-eight hours, persistence after a five day rest in stem-cell-maintaining conditions, and functional developmental potential measured by the hPSC ScoreCard panel, by undirected differentiation, by directed differentiation into all three germ layers analysed by sparse principal component analysis, and by a defined embryoid body protocol for cardiomyocytes as a mesoderm sublineage.

Key findings

1. Multidimensional scaling of RNA sequencing data placed fibroblasts and rederived fibroblasts close together and both far from the induced pluripotent state, while the pluripotent cells expressed NANOG, SOX2, TRA-1-60 and OCT4 and formed all three germ layers in directed differentiation (Figure 1A and SI Appendix Figure S1).

2. Only the pluripotent cells failed to induce RIG-I and IFIT1 protein after treatment with 5 prime triphosphate RNA or interferon beta, and failed to induce IFNB transcript after 5 prime triphosphate RNA, while both fibroblast populations responded (Figures 1B and 1C). Redifferentiation restored responsiveness, which excludes clonal selection or genetic loss. Small RNA profiles of resting and treated cells were unchanged, which the authors take as evidence that neither microRNAs nor virus-induced small RNAs underlie the phenotype (SI Appendix Table S2).

3. RNA sequencing after interferon beta showed strong enrichment for virus response and interferon signalling genes in fibroblasts and little in pluripotent cells, with IRF7 induced roughly 30-fold, MX1 roughly 300-fold and IFIT1 roughly 200-fold in fibroblasts against no induction, roughly 4-fold and roughly 5-fold respectively in pluripotent cells, corroborated by quantitative RT-PCR (Figure 2A and SI Appendix Figure S2A).

4. Infection with influenza A virus lacking NS1 produced a broad antiviral and non-antiviral transcriptional response in fibroblasts and only a small subset of interferon-stimulated genes in pluripotent cells (Figure 2A). Virus levels in pluripotent cells were nonetheless relatively contained, which the authors attribute to high baseline interferon-stimulated gene expression as reported by others, an attribution taken from the literature rather than established here.

5. Interferon beta pretreatment protected fibroblasts completely from influenza A virus lacking NS1, with STAT1 and IFIT1 induction, while pluripotent cells accumulated viral nucleoprotein and showed neither induction. Phosphorylation of IRF3 at S386 and STAT1 at Y701 was reduced in pluripotent cells, which the authors read as suggesting the presence of a dominant negative factor, a suggestion not tested directly (Figure 2B).

6. Mouse embryonic stem cells reproduced the phenotype, failing to induce IFIT1 or IFIT2 or to control viral nucleoprotein after murine interferon beta, in contrast to mouse embryonic fibroblasts (Figure 2C). This extends the observation across species and beyond cells produced by reprogramming.

7. In cells competent to respond, co-expression of OCT4, SOX2 or KLF4 with IRF7 delta repressed induction of endogenous IFIT1 and ISG15, with KLF4 the most potent (Figures 3A and 3B). RNA sequencing showed that KLF4 alone has its own transcriptional footprint that does not include the canonical interferon-stimulated genes, and that when co-expressed it retains its own signature while neutralising the IRF7 delta output (Figure 3C). The authors interpret this as KLF4 preventing access to interferon-stimulated response elements through chromatin compaction, citing chromatin accessibility work by others, and note in the discussion that since the binding motifs do not overlap the effect is presumably indirect.

8. Doxycycline-inducible IRF7 delta in pluripotent stem cells induced IFNA and IFIT1 within twelve to twenty-four hours together with distinct morphological change, and RNA sequencing at forty-eight hours showed induction of interferon-stimulated genes including IFIT1, IFITM3 and TRIM22 (Figures 4A to 4C). This is the engagement the authors state they could not achieve by any other means.

9. After a forty-eight hour pulse followed by five days of rest in stem-cell-maintaining conditions, roughly 2,000 genes remained differentially expressed relative to the GFP control, while IRF7 delta itself and the canonical interferon-stimulated genes had returned to baseline (Figure 4D and SI Appendix Figures S4D and S4E). A comparable pattern of transient interferon-stimulated gene induction followed by loss of pluripotency markers including KLF4 and changes in chromatin remodelling enzymes was seen in mouse embryonic stem cells using transient transfection of a different IRF-inducing transcript (SI Appendix Table S7).

10. The hPSC ScoreCard panel after pulse and rest showed that pluripotency was not formally lost but that NANOG and IDO1 were dysregulated, along with markers of mesendoderm, ectoderm, mesoderm and endoderm (Figure 5A). Undirected differentiation produced a general tendency toward mesoderm in both populations, with ectoderm and endoderm significantly dysregulated after the IRF7 delta pulse (Figure 5B).

11. Directed trilineage differentiation analysed by sparse principal component analysis showed overlapping positions for pulsed and control cells at rest, tight clustering for mesoderm, partial overlap for ectoderm, and a large departure for endoderm after the IRF7 delta pulse (Figure 5C).

12. Within the mesoderm lineage, pulsed cells showed elevated baseline expression of several cardiomyocyte markers before any directed differentiation, and after embryoid body differentiation with activin A and BMP4 they showed higher TNNT2, SIRPA, MYH6 and MYL7 while MYL2 was reduced relative to control despite high baseline induction (Figure 6A and SI Appendix Figure S6B). Beating was significantly increased at day 20 (Figure 6B), yet RNA sequencing of the resulting cardiomyocytes showed roughly 5,000 differentially expressed genes (Figure 6C). The authors read this as successful but aberrant differentiation.

Mechanistic model

The study does not establish a single definitive mechanism, and the authors are explicit that the basis for the incompatibility is likely multifactorial. Two partially separable layers are supported. On the repression side, reprogramming factors and above all KLF4 prevent transcriptional output at interferon-stimulated response elements, and because the KLF4 binding motif does not overlap the response element the authors infer an indirect route through chromatin remodelling, resting their case on published chromatin accessibility data rather than on measurements made here. They also point out that this account does not explain the reduced phosphorylation of IRF3, NF-kappaB and STAT1 that they and others observe, and note that RIG-I protein is low in pluripotent cells despite comparable DDX58 read counts, which they suggest may reflect stem-cell-specific proteases or microRNAs, a suggestion offered as speculation. On the consequence side, the data show that when the program is driven directly by IRF7 delta, bypassing all of this, the pluripotent cell is durably altered. Direct IRF7 targets revert when the stimulus is withdrawn while thousands of other transcripts do not, and subsequent differentiation is skewed and internally inconsistent. The authors frame this as IRF7 priming cells to exit pluripotency, with a bias toward mesoderm, and state that whether the mesoderm bias is specific or reflects the baseline tendency of the particular cell line remains to be determined. The broader proposal, that pluripotency and the interferon system are incompatible, is offered as a hypothesis the data support rather than as a demonstrated mechanism.

Conceptual or technical advance

The work supplies a way to switch on an interferon-stimulated gene program inside cells that cannot be induced by interferon, viral RNA, or infection, which converts a descriptive observation about unresponsiveness into a testable manipulation. It identifies KLF4 as a specific and potent repressor of IRF7-driven transcription and shows that a reprogramming factor alone can confer the pluripotent cell's unresponsiveness on a differentiated cell. It also shifts the question from why the defence is absent to what the defence would cost, and shows that the cost is measurable as lasting transcriptional change and compromised germ layer potential after a stimulus as brief as forty-eight hours. That result makes the developmental consequences of interferon exposure in early embryonic cells an experimentally approachable problem and raises a caution for protocols in which pluripotent cultures encounter interferon or infection.

Relationship to the broader research program

The paper draws directly on earlier work from the laboratory on the transcriptional architecture of the antiviral response, including redundancy among the transcription factors that induce the antiviral state and mitogen-activated protein kinase-mediated licensing of IRF3 and IRF7, which supply the rationale for using a constitutively active IRF7 as a sufficient driver of interferon-stimulated genes. It also connects to the laboratory's interest in the evolution of antiviral defence systems and in how defence and developmental programs borrow from one another, a theme the discussion develops with examples including Toll-like receptors, RNA interference and microRNA, and RNase III nucleases. Marked as category 3 synthesis, the corpus-level thread here is a recurring use of influenza A virus lacking NS1 as a probe that reports on host sensing capacity, appearing in this stem cell context as a tool rather than a subject. A second synthesis thread is the treatment of the interferon response as a program with costs to the cell beyond its antiviral benefit, an idea that recurs where the laboratory examines how the same response shapes differentiation, tissue state, or disease outcome rather than only virus replication.

Related publications

• Schmid et al. 2010 (Journal of Biological Chemistry), methodological foundation. Established transcription factor redundancy in induction of the antiviral state, from the same laboratory, and underpins the use of IRF7 as a sufficient driver.
• Schmid, Sachs and tenOever 2014 (Journal of Biological Chemistry), methodological foundation. Licensing of IRF3 and IRF7 by mitogen-activated protein kinase signalling, cited here for the properties of IRF7 that make IRF7 delta effective.
• tenOever 2016 (Cell Host and Microbe), review or synthesis. The author's review of the evolution of antiviral defence systems, cited in the opening framing of defence diversity.
• Aguado et al. 2017 (Nature), conceptual extension. Work from the same laboratory on RNase III nucleases as antiviral effectors across kingdoms, cited in the discussion of repurposing between developmental and defence systems.
• Blanco-Melo, Venkatesh and Bieniasz 2016 (Cell Host and Microbe), conceptual extension. Origins and evolution of tetherin, cited by a co-author of the present paper as an example of gene duplication serving both systems.
• Brennand et al. 2011 (Nature), methodological foundation. Source of the induced pluripotent stem cell modelling approach contributed by the co-author's laboratory.

Limitations and boundaries

The central manipulation is artificial. IRF7 delta drives a subset of interferon-stimulated genes directly and bypasses upstream signalling entirely, so what is tested is the consequence of an interferon-like transcriptional program rather than of a physiological infection or of interferon exposure, which pluripotent cells cannot mount in any case. The authors describe the induced response as partial. Dosage and duration are fixed at a single forty-eight hour pulse followed by five days of rest, so the shape of the dose and time relationship is unknown. The main human work rests on a small number of induced pluripotent stem cell clones derived from foreskin fibroblasts and from healthy patient fibroblasts by two different reprogramming routes, and the authors state explicitly that whether the mesoderm bias is a property of the response or of the baseline tendency of their particular line is unresolved. RNA sequencing conditions used two biological replicates, and differentiation readouts are transcriptional and morphological rather than functional in most cases, the beating cardiomyocytes being the exception. The repression mechanism attributed to KLF4 is inferred from published chromatin accessibility data rather than measured here, and the authors note it does not account for the reduced phosphorylation of IRF3, NF-kappaB and STAT1 that they observe. All differentiation is in vitro, with no embryo or animal work, so no claim is made about development in a living organism. Finally, the framing that pluripotency and the interferon system are incompatible is presented by the authors as a hypothesis their data support and as speculation where it extends to evolutionary crosstalk.

Audience summaries

25 words

Stem cells ignore interferon signals. Forcing the antiviral gene program on anyway leaves thousands of genes altered days later and damages their ability to form tissues.

75 words

Pluripotent stem cells neither make nor respond to type I interferon. Reprogramming factors, KLF4 most strongly, block the transcription factor that drives interferon-stimulated genes. Using a constitutively active IRF7 to switch the program on regardless changed stem cell morphology and gene expression, and left roughly 2,000 genes altered after five days of recovery. Differentiation afterwards was impaired for ectoderm and endoderm and distorted within mesoderm, with cardiomyocytes beating more but transcriptionally abnormal.

150 words

Eggenberger and colleagues compared human foreskin fibroblasts, induced pluripotent stem cells derived from them, and fibroblasts rederived from those stem cells. Only the pluripotent state failed to respond to 5 prime triphosphate RNA, to interferon beta, or to influenza A virus lacking NS1, and mouse embryonic stem cells behaved the same way. Co-expression experiments showed that KLF4, and less potently SOX2 and OCT4, suppress transcription driven by a constitutively active IRF7, apparently by restricting access to interferon-stimulated response elements. A doxycycline-inducible IRF7 delta then allowed the program to be engaged in pluripotent cells directly. A forty-eight hour pulse induced interferon-stimulated genes and altered morphology, and after five days of rest roughly 2,000 genes remained differentially expressed although the direct targets had reset. Subsequent differentiation compromised ectoderm and endoderm and distorted mesoderm sublineage output, with cardiomyocytes showing increased beating alongside roughly 5,000 differentially expressed genes.

Discoveries supported by this paper

Documented publication relationships

• Transcription Factor Redundancy Ensures Induction of the Antiviral State — methodological foundation. Direction: this paper → linked paper. stated in the Related publications section of 2019-eggenberger-type-i-interferon-response-impairs

Pathogens

Technologies