SARS-CoV-2 infection of the olfactory epithelium reorganizes the nuclear architecture of uninfected olfactory sensory neurons, dissipating the interchromosomal compartments that hold olfactory receptor genes and suppressing receptor and signal transduction transcription in both hamsters and human autopsy tissue.
Zazhytska M, Kodra A, Hoagland DA, Frere J, Fullard JF, Shayya H, McArthur NG, Moeller R, Uhl S, Omer AD, Gottesman ME, Firestein S, Gong Q, Canoll PD, Goldman JE, Roussos P, tenOever BR, Overdevest JB, Lomvardas S. Non-cell-autonomous disruption of nuclear architecture as a potential cause of COVID-19-induced anosmia. Cell. 2022. Volume 185, issue 6, pages 1052-1064.e12.
DOI 10.1016/j.cell.2022.01.024. PMID 35180380. PMCID PMC8808699.
SARS-CoV-2 infection of the olfactory epithelium reorganizes the nuclear architecture of uninfected olfactory sensory neurons, dissipating the interchromosomal compartments that hold olfactory receptor genes and suppressing receptor and signal transduction transcription in both hamsters and human autopsy tissue.
Loss of smell is among the most common symptoms of COVID-19 and one of the hardest to explain, because olfactory sensory neurons do not express the entry factors SARS-CoV-2 requires and are infected only rarely. The authors ask how a virus that does not enter these neurons nonetheless silences their defining function. They combine an infected golden hamster time course with human olfactory epithelium autopsies, using single-cell and bulk RNA sequencing to describe the transcriptional response and in situ Hi-C to describe nuclear organization. In hamsters the virus infects sustentacular cells, which are transiently depleted and then restored, while neuron representation stays constant. Despite escaping infection the neurons mount an antiviral response and then, with a delay, lose expression of olfactory receptor genes and of the signal transduction genes that make odor detection possible, a loss still present at ten days when other markers have recovered and the virus has been cleared. Preceding that loss, the long-range cis and trans contacts among olfactory receptor gene clusters, which normally converge into specialized genomic compartments, are reduced from one day after infection and remain reduced at ten days. Ultraviolet-inactivated serum from infected hamsters, applied intranasally to naive animals for 12.5 hours, reproduces the loss of contacts without transferring virus. Human autopsies show the same receptor downregulation and the same loss of interchromosomal contacts in sorted neuronal nuclei. The circulating factor responsible was not identified.
Olfactory dysfunction emerged early as a common neurological symptom of COVID-19, usually resolving within about six weeks but persisting in roughly ten percent of patients. Unlike smell loss in other upper respiratory infections, it is not attributable to congestion blocking odorants from reaching the epithelium. That made the mechanism puzzling, because published work had established that olfactory sensory neurons do not express ACE2 and TMPRSS2 and are not appreciably infected by the virus, while sustentacular cells are. Separately, work largely from the senior authors' field had established that olfactory receptor choice depends on an unusual nuclear organization. Receptor gene clusters from many chromosomes converge into a small number of specialized genomic compartments, and this convergence supports stable expression of a single receptor per neuron. Transcription factors including Lhx2 and Ebf family members are required for receptor and signaling gene expression. Prior work by other groups had also linked innate immune signaling in the olfactory epithelium to reduced odorant receptor levels. What had not been examined was whether an infection confined to neighboring cells could reach into the neuron and disassemble that architecture.
How does SARS-CoV-2 cause loss of smell when it does not infect olfactory sensory neurons, and specifically does infection of the surrounding epithelium act on those neurons through a non-cell-autonomous route that alters the nuclear organization on which olfactory receptor expression depends?
The design pairs a controllable animal time course with human tissue that carries the actual clinical condition, and reads each at two levels, transcription and genome organization. Golden hamsters are used because their ACE2 resembles the human protein and because pathogenesis and immune response in this species have been characterized. Single-cell RNA sequencing across mock and infected animals at one, three and ten days establishes which cells are infected, whether any population is lost, and what each cell type does transcriptionally. Because the single-cell chemistry used a 5-prime approach that is well suited to detecting viral transcripts but poorly suited to olfactory receptor messages given hamster genome annotation, bulk RNA sequencing at one, two, four and ten days is added specifically to quantify receptor expression. In situ Hi-C on the epithelium at matched time points asks whether the receptor compartments are intact. The causal step separating virus from effect is a serum transfer, in which serum drawn at three days from infected or mock animals is ultraviolet-irradiated to inactivate virus and inoculated intranasally into naive hamsters, with sequencing used to confirm no viral genome was carried over. On the human side, a specific region at the roof of the nasal cavity enriched for olfactory neurons is identified and sampled from control and COVID-19 autopsies, assessed by histology and in situ hybridization, profiled by bulk RNA sequencing with batch correction and outlier removal, and subjected to in situ Hi-C after fluorescence-activated sorting of neuronal nuclei.
The virus targets sustentacular cells and largely spares neurons. Across 68,951 hamster cells and thirteen annotated types, viral RNA was present in about 5 percent of cells at one and three days and gone by ten days. At one day roughly 47 percent of infected cells were sustentacular, about 40 percent of that population, while only about 6 percent of infected cells were neurons (Figure 1C-1E and Figure S1B). Spike protein colocalized with sustentacular and microglial markers, and colocalization with neuronal markers was rare and confined to regions of shedding and structural damage (Figures S1C and S1E).
Sustentacular cells are transiently depleted and neurons are not. Sustentacular representation fell from 20.6 percent in mock animals to 6 percent at three days, with a concurrent rise in microglia and other immune cells, and both returned to preinfection levels by ten days. Neuronal representation was stable throughout (Figures 1A and 1B).
Infected sustentacular cells show cell-autonomous changes. Splitting infected epithelia into virus-positive and virus-negative sustentacular cells revealed upregulated cytokines and chemokines and downregulated sustentacular markers in the positive fraction (Figure 2B).
Uninfected neurons respond anyway. Olfactory sensory neurons activated antiviral gene expression at three days, declining by ten days as virus cleared, and genes required for smell, notably Adcy3, were significantly reduced at three days (Figures 2C, 2D and S2). Reduction of Adcy3 message and protein was confirmed by in situ hybridization and immunofluorescence at four days, including in regions with little detectable virus (Figures 2E and 2F), which the authors take as evidence that the effect does not require local infection.
Receptor gene downregulation is the dominant transcriptional change and it persists. In bulk sequencing, viral load rose through four days and was eliminated by ten. Olfactory receptor downregulation began at two days, peaked at four and continued through ten days, when other neuronal markers had recovered and when the most variable genes in the epithelium had returned to baseline (Figures 3F, 3G and 3H). Adcy3, Gng13, Cnga2, Rtp1 and Gfy followed the same pattern (Figure 3I).
Cell-type responses are staggered. Sustentacular and immediate neuronal precursor changes came early and resolved, with precursor transcription factors Lhx2, Ebf1 and Ebf2 reduced mainly at two days, while neuronal and globose basal cell changes were delayed, and globose basal markers peaked at ten days, which the authors read as progenitor activation toward replacement (Figures 3D and 3E).
Receptor gene compartments come apart. In situ Hi-C showed the expected strong long-range cis contacts and widespread trans contacts among hamster receptor clusters in controls. Contacts were reduced from one day, most strongly at three days, and remained low at ten days (Figures 4A-4C). Hidden Markov model compartment scores showed widespread genome-wide compartment changes by three days, later than the receptor-specific disruption, and still disrupted at ten days after viral clearance (Figure 4D).
Serum from infected animals reproduces the nuclear effect without virus. Ultraviolet-inactivated serum collected at three days and given intranasally to naive hamsters for 12.5 hours significantly reduced trans contacts among receptor clusters and changed genome-wide compartment scores, with sequencing confirming no viral genome in the recipients (Figures 5B-5E). At this early point receptor transcription had not yet changed significantly, though a downward trend across most receptor genes was present and more pronounced than at one day of actual infection (Figure S4A). Sustentacular markers did not respond to infected serum (Figure S4C), consistent with those changes being cell autonomous.
Human autopsies show the same transcriptional signature. Across eighteen infected and a small number of control olfactory epithelium samples, viral RNA was detected in every infected specimen at variable levels while epithelial and respiratory marker representation did not change with viral load. Controls had higher receptor message levels, principal component analysis separated infected from control samples when restricted to receptor genes but not when using the whole transcriptome, and sensory perception of smell was the most significantly enriched gene ontology term among downregulated genes (Figures 6A, 6E-6H, 6J and 6K). One nominal control carried human coronavirus OC43 and clustered with the infected samples, and was excluded from the control pool.
Human neurons lose the same contacts. In situ Hi-C on nuclei sorted from two control and four infected autopsies showed conserved receptor-specific long-range cis and trans contacts in humans and their reduction in infected samples, including in specimens where genome-wide compartmentalization was not comparably altered (Figures 7A-7E), and interchromosomal compartments containing Adcy3 and other olfaction genes also dissipated (Figure S6B).
Compartment disruption can precede transcription factor loss. In hamsters the loss of receptor compartments preceded downregulation of Lhx2 and Ebf and persisted after their restoration, and in two infected human specimens receptor transcription and compartmentalization were disrupted while Lhx2 and Ebf were near control levels. The authors use this ordering to argue that compartment disruption is upstream rather than a consequence of transcription factor loss, which is an argument from temporal and correlative evidence rather than from perturbation.
The study does not establish a definitive mechanism, and the authors say so in their own limitations. What the data support is the following chain. SARS-CoV-2 infects sustentacular cells and elicits a local and systemic inflammatory response. Something in circulation, present in serum by three days and surviving ultraviolet inactivation, reaches olfactory sensory neurons that the virus itself does not enter. In those neurons the interchromosomal compartments that gather olfactory receptor gene clusters lose their contacts, and receptor and signal transduction transcription falls and stays low beyond viral clearance. Reduced expression of receptors, of the chaperones Rtp1 and Rtp2, of the signaling components Adcy3 and Gng13 and of the channel Cnga2 would be expected to impair odor detection, an inference the authors draw explicitly from the phenotypes of knockout mice rather than from measurement of smell in this study.
Several elements are proposals. The identity of the circulating molecule or molecules is unknown and the neuronal signaling pathway that receives the signal was not determined. The nuclear memory idea, that compartments may form only during differentiation so that their disruption in mature neurons could be effectively irreversible and recovery would require neuronal replacement, is offered as a hypothesis with a stated reason rather than as a finding. The suggestion that restored contacts might reassemble differently and produce odor misrepresentation is a reasonable extrapolation the authors raise and do not test. The causal link from receptor downregulation to anosmia in patients is an inference, since olfactory function was not measured in the autopsy cohort beyond one self-reported case. The direction of the compartment to transcription factor relationship rests on timing and on two human specimens, not on manipulation.
The work supplies a concrete cellular route by which an infection confined to a small fraction of cells can change the identity and function of cells it never enters, which reframes a general puzzle about SARS-CoV-2 pathology in terms of systemic signals acting on nuclear organization. It also brings three-dimensional genome organization into the interpretation of an acute infectious symptom, and it demonstrates that in situ Hi-C can be performed on neuronal nuclei sorted from human autopsy tissue, which makes nuclear architecture an accessible readout in post-mortem material generally. The serum transfer establishes an assay in which a systemic consequence of infection can be separated from the virus and delivered to a naive animal, and the observation that the receptor compartments are the earliest and most durable casualty, lost before genome-wide compartmentalization changes and before the relevant transcription factors fall, makes the fragility of these particular interchromosomal contacts a testable property.
Within the tenOever laboratory's line of work the paper connects to the group's characterization of the host transcriptional response to SARS-CoV-2 and to the use of the golden hamster as a model for that response, both of which supplied the infection system and the antiviral gene framework used here. The recurring question the paper shares with that work is how much of the damage attributed to a virus is done by the host response rather than by infection of the damaged cells themselves, a question that also runs through the group's earlier interferon signaling studies from the opposite direction. Reading the antiviral program in uninfected neurons here alongside the group's work on interferon-driven transcriptional programs as a single thread about the costs of the response is category 3 synthesis, offered as such, and it is not a claim this paper makes.
The authors state their own limitations plainly and those are reproduced here. The circulating molecule or molecules responsible were not identified, nor was the neuronal signaling pathway that transduces the signal, so generalization to other neuronal populations is speculation they say they have not explored. They did not establish that receptor and signaling gene downregulation causes the anosmia, inferring this instead from knockout mouse phenotypes. In humans they can only deduce that infection caused the downregulation, because expression before infection cannot be measured, and although the hamster experiments support the interpretation they cannot exclude rodent-specific mechanisms. Beyond those, the human cohort is small and unbalanced, with eighteen infected samples against a handful of controls, one of which was removed as an outlier and one of which carried a different coronavirus, and olfactory status is known for only one patient, with the prevalence of deficits in the remainder estimated from the literature. Autopsy material carries variable post-mortem intervals, treatments and disease durations, and batch correction was applied. The human Hi-C comparison rests on two control and four infected autopsies. The hamster time course extends only to ten days, so persistence beyond that, and any recovery, is not observed, and the proposed irreversibility of compartment loss is untested. The serum transfer used a single 12.5 hour exposure to serum from a single time point and shows sufficiency of something in that serum, not necessity of any identified factor, and no smell testing was performed in hamsters or humans.
A virus that does not enter smell neurons still silences them, unraveling the chromosome contacts that hold odorant receptor genes together and shutting down odor detection machinery.
SARS-CoV-2 infects support cells in the nose but rarely the sensory neurons themselves, which made COVID-19 smell loss hard to explain. In hamsters and in human autopsy tissue, infection collapsed the unusual contacts that gather odorant receptor genes from many chromosomes into shared nuclear compartments, and receptor and signaling gene expression fell and stayed low after the virus was gone. Serum from infected hamsters, with virus inactivated, reproduced the nuclear change in naive animals.
Olfactory sensory neurons express one receptor each, supported by interchromosomal compartments that gather receptor gene clusters. The authors asked how SARS-CoV-2 causes anosmia without infecting those neurons. Single-cell sequencing of infected hamster olfactory epithelium showed the virus in sustentacular cells, which were transiently depleted, while neuron numbers held steady. Those uninfected neurons nonetheless mounted an antiviral response and then lost olfactory receptor and signal transduction transcripts, a loss beginning at two days, peaking at four and persisting at ten days after viral clearance. In situ Hi-C showed receptor cluster contacts falling from one day and staying low, before genome-wide compartment changes and before Lhx2 and Ebf downregulation. Ultraviolet-inactivated serum from infected animals reproduced the contact loss in naive hamsters without transferring virus. Human autopsies showed matching receptor downregulation and reduced contacts in sorted neuronal nuclei. The circulating mediator and the receiving pathway were not identified.
co-led. Frere 2022 and Serafini 2023 are co-corresponding with the Zachariou laboratory, Zazhytska 2022 was led with the Lomvardas and Overdevest groups, and Heaton 2014 is co-led with Peter Palese and tenOever as joint senior and corresponding authors