Co-opting the unfolded protein response to elicit olfactory receptor feedback

Abstract

Olfactory receptor (OR) expression requires the transcriptional activation of 1 out of 1,000s of OR alleles and a feedback signal that preserves this transcriptional choice. The mechanism by which olfactory sensory neurons (OSNs) detect ORs to signal to the nucleus remains elusive. Here, we show that OR proteins generate this feedback by activating the unfolded protein response (UPR). OR expression induces Perk-mediated phosphorylation of the translation initiation factor eif2α causing selective translation of activating transcription factor 5 (ATF5). ATF5 induces the transcription of adenylyl cyclase 3 (Adcy3), which relieves the UPR. Our data provide a role for the UPR in defining neuronal identity and cell fate commitment and support a two-step model for the feedback signal: (1) OR protein, as a stress stimulus, alters the translational landscape of the OSN and induces Adcy3 expression; (2), Adcy3 relieves that stress, restores global translation, and makes OR choice permanent.

Figure 1

(A) mRNA RPKM values for Atf5, Atf4, and developmental markers from cell populations isolated by fluorescence-activated cell sorting (FACS). HBCs: horizontal basal cells; GBCs: globose basal cells; mOSNs: mature olfactory sensory neurons. (B) Atf5 immunofluorescence (IF, red), Adcy3 IF (green). DAPI nuclear stain (blue). Sections are from P40 animals. Specificity of Atf5 antibody is shown in Supplemental Figure S1A,B. ATF5 mRNA expression values shown in Supplemental Figure S1C. (C) Atf5 IF in Foxg1-Cre; Lsd fl/+ at embryonic day 17 (E17) in (D) Atf5 IF in Foxg1-Cre; Lsd fl/fl (E) Atf5 IF in Foxg1-Cre; Lsd fl/fl and transgenic OR rescue (Gng8tta; OMPtta; tetO-MOR28itlacZ) (Lyons et al., 2013).

Figure 2

(A–F) Sections from P40 Atf5+/− (left panels) and P40 −/− (right panels) stained with antibodies against Adcy3 (A–B), Lsd1 (C–D), or MOR28 (E–F). Shown with or without DAPI merge. Arrows in (F) point to ER regions with MOR28 aggregates only seen in ATF5 KO MOEs. Quantitation of the IF signal intensities for LSD1 and MOR28, as well as quantification of the number of MOR28+ OSNs are shown in Supplemental Figure S2A–C. (G) Boxplot summary of expression of refseq ORs from mRNA-seq on P40 Atf5 +/+ (orange, 1041 ORs detected) and Atf5 −/− (green, 939 ORs detected). Pseudogene ORs excluded. (H) RPKM values normalized to wild-type for developmental markers. Atf5 +/+ shown in orange and Atf5 −/− in green. See Supplemental Figure S2D for expression levels of additional developmental markers.

Figure 3

(A) A genetic strategy to assay the stability of OR expression. One copy of the MOR28 gene also drives expression of Cre recombinase (MOR28-IRES-Cre), excising a stop signal from a Rosa lox-stop-lox-Tomato allele, permanently labeling the OSN with Tomato fluorescent protein. Cells with stable MOR28-IRES-Cre expression (left, yellow) are positive for Tomato and MOR28 as assayed by antibody staining, while cells with unstable MOR28-IRES-Cre expression (right, red), are positive for Tomato only. Cells that choose the wild type MOR28 allele are only green. (B) Sections from P40 MOR28-IRES-Cre; lox-stop-lox-Tomato; Atf5 +/− (left panels) or Atf5 −/− (right panels). MOR28 IF (green) alone (bottom) and with Tomato reporter (red) are shown (top). (C) Quantification of gene switching from animals shown in (B). Data represented as percentage Tomato+/MOR28− cells over percentage Tomato+/MOR28+ cells. p-value <2.2e–16 (Fisher's test).

Figure 4

(A) IF for Atf5 (red) with DAPI (left panel), IF for Adcy3 (green) with DAPI (middle panel) and IF for Atf5 and Adcy3 (right panel) in a section from a P0 Eif2 S51A/+ animal. (B) IF for the same markers in a littermate Eif2 S51A/S51A animal. ATF5 mRNA levels in control and mutant MOEs shown in Supplemental Figure S3A. (C) IF for the same markers in an Eif2 S51A/S51A animal with transgenic Atf5 rescue (Gng8-tta; tetO-Atf5). (D) Genetic strategy for Atf5 transgenic rescue. Endogenous Atf5 protein is expressed just prior to Adcy3 expression (left). Eif2 phosphomutants (S51A/S51A) fail to express Atf5 or Adcy3 (middle). Transient expression of the Atf5 coding sequence under the control of Gng8-tta results in a pattern of Atf5 expression slightly expanded towards the basal MOE, and partially rescues Adcy3 expression (see Supplemental Figure 3C for quantification).

Figure 5

(A) IF for Atf5 (red) with DAPI (left panel), IF for Adcy3 (green) with DAPI (middle panel) and IF for Atf5 and Adcy3 (right panel) in a section from a P0 Perk +/− animal. (B) IF for the same markers in a Perk −/− littermate. Quantification of the numbers of ATF5-expressing cells and the intensity of IF signal shown in Supplemental Figure S4A,B. (C) A female Lsd fl/fl mated to a male Lsd fl/+; Foxg1-Cre was given a single IP injection of tunicamycin at E16.5. At E17.5 pups were collected and sectioned. Shown is IF for Atf5 (red) and DAPI (left panel) and IF for Adcy3 (green) and Atf5 (red). For comparison with Foxg1-Cre; Lsd fl/fl or Lsd fl/+, see Figure 1 and for quantification of the numbers of ATF5-expressing cells see Supplemental Figure S4A.

Figure 6

(A) IF for Atf5 (red) with and without DAPI merge in Adcy3 +/− (left panels) and Adcy3 −/− (right panels). (B) Quantification of Atf5 fluorescence intensity in a section from Adcy3 +/− or Adcy3 −/− animals. Shown as % basal to apical position vs. % maximum intensity (see methods). Raw data shown as scatterplot in background and locally weighted scatterplot smoothing shown in orange (Adcy3+/−) or green (Adcy3 −/−) (C) RPKM values for Atf5 mRNA in Adcy3 +/− and Adcy3 −/−

Figure 7

(A) A model for the generation of the OR feedback signal: Lsd1 transcriptionally activates an OR, which is co-translationally detected by Perk in the endoplasmic reticulum. OR-Perk interaction activates Perk, which then phosphorylates eif2α, resulting in a global pause in translation initiation and a selective increase in nuclear Atf5 translation. Atf5 activity initiates Adcy3 transcription, and according to our RNAseq analysis also activates transcription of OR chaperones RTP1 and RTP2 (Supplemental Figures S2D). (B) nATF5-dependent upregulation of Adcy3 and OR-specific chaperones relieves the ER stress and restores global translation in the OSN. Although this leads to an increase of OR and Adcy3 protein levels, it stops the translation of nATF5 isoform clearing this protein from the nucleus. Increased OR and Adcy3 levels also cause downregulation of LSD1 (Lyons et al., 2013) preventing OR switching and stabilizing OR choice. (C) A two-step model of feedback explains its specificity for ORs by providing two independent tests: 1) induction of UPR, and 2) relief of the UPR. Intact ORs (left) that pass both tests are stably expressed for the life of the neuron; non olfactory GPCRs and/or pseudogene ORs may fail to activate Perk allowing the process of OR choice to continue until an intact OR is expressed. At a second level of specificity a GPCR or pseudogene OR that passes the first test may not be recognized by OR-specific chaperones causing prolonged ER stress and sustained LSD1 expression, eventually allowing activation of an OR allele and/or OR gene switching.