The midnolin-proteasome pathway catches proteins for ubiquitination-independent degradation

Abstract

Cells use ubiquitin to mark proteins for proteasomal degradation. Although the proteasome also eliminates proteins that are not ubiquitinated, how this occurs mechanistically is unclear. Here, we found that midnolin promoted the destruction of many nuclear proteins, including transcription factors encoded by the immediate-early genes. Diverse stimuli induced midnolin, and its overexpression was sufficient to cause the degradation of its targets by a mechanism that did not require ubiquitination. Instead, midnolin associated with the proteasome via an α helix, used its Catch domain to bind a region within substrates that can form a β strand, and used a ubiquitin-like domain to promote substrate destruction. Thus, midnolin contains three regions that function in concert to target a large set of nuclear proteins to the proteasome for degradation.

Fig. 1.. Genetic screens reveal midnolin as a regulator of IEG protein degradation.

(A) Schematic showing the FACS-based genome-wide CRISPR-Cas9 screens using the Global Protein Stability (GPS) reporter of IEG proteins in HEK-293T cells (created with BioRender.com). (B and C) Results of the genetic screens revealed MIDN as the top hit for negatively regulating the stability of both EGR1 and FosB. The proteasomal components showed a weaker enrichment likely due to their essentiality. The MAGeCK score represents the negative log10 of the “pos|score” value generated from MAGeCK (56). (D and E) Losing midnolin stabilized, while overexpressing midnolin destabilized EGR1 and FosB. GPS EGR1 or FosB reporters were stably expressed in wild-type and two independent MIDN KO HEK-293T single cell clones. Vectors expressing BFP control alone (grey) or midnolin and BFP from a EF1α promoter (red) was transiently reintroduced by transfection before analyzing the GFP/DsRed ratio by flow cytometry.

Fig. 2.. Midnolin is induced and promotes the degradation of several IEG proteins in physiological settings.

(A) Loss of midnolin increased the expression of IEG proteins in NIH/3T3 cells. Immunoblotting was performed from NIH/3T3 cells stably expressing Cas9 and control or MIDN targeting single guide RNAs. This population-level mutagenesis of MIDN may show lower penetrance relative to an isogenic mutant since the knockout efficiency is dependent on the efficacy of the single guide RNA. The cells were starved of serum overnight before serum restimulation for the indicated time points. Asterisks mark non-specific cross-reactive proteins. (B) Overexpressing midnolin decreased the expression of IEG proteins in NIH/3T3 cells. Same assay as (A) but in NIH/3T3 cells stably overexpressing an N-terminally 2xFLAG tagged human midnolin using a CMV promoter. (C) Loss of midnolin increased the expression of IEG proteins in primary cortical neurons. Neurons were isolated from E16.5 mouse brains and cultured in a dish. On day 3 post-isolation, the neurons were infected with lentivirus encoding Cas9 with control or MIDN targeting single guide RNAs. Immunoblotting was performed on day 11 post dissection from neurons that were silenced overnight with tetrodotoxin (TTX, a sodium channel blocker) and D-AP5 (a NMDA receptor antagonist) and stimulated with KCl for the indicated time points to induce depolarization. (D) Overexpressing midnolin decreased the expression of IEG proteins in primary cortical neurons. Similar assay as (C) but using lentivirus to overexpress a BFP control or human midnolin co-expressing BFP using an EF1α promoter. (E and F) qPCR analysis for mRNA levels of the indicated genes from (E) primary mouse cortical neurons that were KCl stimulated or (F) from NIH/3T3 cells that were serum restimulated for the indicated time points. Error bars represent the standard deviation from three biological replicates. Data were analyzed using an ordinary one-way ANOVA followed by Tukey’s multiple comparisons test where **** represents p < 0.0001.

Fig. 3.. Midnolin can promote the degradation of numerous transcriptional regulators.

(A) Schematic showing the midnolin GPS ORFeome screen. The GPS ORFeome library (~12,000 barcoded human ORFs tagged to GFP) was introduced into MIDN KO HEK-293T and the library-expressing cells were transfected with BFP control or midnolin co-expressing BFP before FACS sorting the library into populations based on the GFP/DsRed ratio (created with BioRender.com). (B) Analysis of the GPS ORFeome screen showing the change in protein stability (ΔPSI) between midnolin and BFP, which was calculated based on the change in read distribution of the barcoded ORFs. Approximately 5% of the library showed significant destabilization with ΔPSI values less than −0.5. Several validated hits from the screen are shown in the boxed table. (C) Gene set enrichment analysis (GSEA) based on the GPS ORFeome screen for molecular function. (D) Validation of screen hits indicates their potent regulation by midnolin. GPS reporters for the indicated genes were stably expressed in MIDN KO HEK-293T cells and a control BFP or midnolin co-expressing BFP were transiently transfected before analyzing the GFP/DsRed ratio by flow cytometry. (E) Endogenous proteins of numerous screen hits are regulated by midnolin. Immunoblotting was performed from wild-type, MIDN KO, and MIDN KO HEK-293T cells where midnolin expression was stably induced with doxycycline (100 ng/mL) for 2 days using a TRE promoter. Shown are putative midnolin targets (red) based on the GPS ORFeome screen and negative controls (black). (F) Validation of midnolin-mediated degradation of endogenous IRF4 in Ramos B cells. Immunoblotting was performed from Ramos B cells expressing Cas9 and control or MIDN targeting single guide RNAs, or stably overexpressing midnolin using an EF1α promoter.

Fig. 4.. Midnolin associates with the proteasome to promote ubiquitination-independent degradation of bound substrates.

(A) A 3xHA-tag was introduced at the N-terminus of the endogenous midnolin locus in HEK-293T cells using CRISPR-Cas9 initiated recombination. Cells were treated with MG132 for 6 hours before immunoprecipitation of 3xHA-midnolin followed by mass spectrometry. The results revealed a large enrichment of the 26S proteasome (Data S3) and shown is a STRING analysis of the top co-immunoprecipitated proteins identified from the mass spectrometry. (B) Midnolin co-immunoprecipitates the proteasome and IEG proteins endogenously. Immunoblotting was performed from anti-HA immunoprecipitants of endogenous 3xHA-midnolin from the knock-in HEK-293T cells treated with the indicated drugs for 6 hours. PMA was used to induce the transcription of IEGs. (C) Endogenous midnolin protein levels are strongly increased by proteasomal inhibition but not by ubiquitin E1 inhibition. Immunoblotting was performed from wild-type and 3xHA-midnolin knock-in HEK-293T cells treated with 10 μM MG132 or 500 nM TAK-243 for 6 hours. (D) Lysine-dependent ubiquitination on substrates is not necessary for midnolin interaction. Immunoblotting was performed from anti-HA immunoprecipitants of HEK-293T cells that were transfected with the indicated constructs, either wild-type or all lysine residues mutated to arginine residues (K to R). Cells were treated with 10 μM MG132 for 6 hours. CBX8 serves as a negative control as it is not targeted by midnolin. (E) Midnolin does not require lysine residues on substrates to promote degradation. Wild-type and K to R mutant substrates were stably introduced into MIDN KO HEK-293T cells using a CMV promoter. Then, midnolin expression was induced using doxycycline (100 ng/mL) for 2 days using a TRE promoter before lysis and immunoblotting.

Fig. 5.. Midnolin contains three regions that function in concert to promote proteasomal degradation of bound substrates.

(A) Midnolin structure prediction by AlphaFold (Q504T8-F1) reveals three regions with defined structure (26). (B) Schematic representation of mutations (57) or truncations introduced into the midnolin cDNA. See methods for the truncation boundaries and regions used for sufficiency experiments. (C) Regions with defined structure are necessary for a functional midnolin. The GPS IRF4 reporter was stably expressed in MIDN KO HEK-293T cells and a control BFP or wild-type and mutant versions of midnolin co-expressing BFP were transiently transfected before analyzing the GFP/DsRed ratio by flow cytometry. (D) The midnolin Catch domain is necessary for binding substrates and the C-terminal α helix is necessary for proteasomal association. Immunoblotting was performed from anti-FLAG immunoprecipitants of HEK-293T cells stably expressing 2xFLAG-tagged midnolin using a CMV promoter. Cells were treated with 10 μM MG132 for 6 hours. (E) The midnolin αHelix-C is sufficient to interact with the proteasome. Immunoblotting was performed from anti-FLAG immunoprecipitants of MIDN KO HEK-293T cells transfected with the indicated 2xFLAG-tagged proteins. (F) The midnolin Catch domain is sufficient to bind substrates. Immunoblotting was performed from anti-FLAG immunoprecipitants of MIDN KO HEK-293T cells transfected with the indicated 2xFLAG-tagged proteins. The 111 amino acid sequence between Catch1 and Catch2 was shortened to 10 amino acids (ΔLoop1), 16 amino acids (ΔLoop2), or 28 amino acids (ΔLoop3). Cells were treated with 10 μM MG132 and 20 ng/mL of PMA for 6 hours. (G) The Catch1 and Catch2 regions of midnolin interact when expressed as independent proteins. Immunoblotting was performed from anti-HA immunoprecipitants of MIDN KO HEK-293T cells co-transfected with 2xHA-GFP-Catch1 and 2xFLAG-MBP-Catch2 constructs, where “e” signifies empty 2xHA-GFP or 2xFLAG-MBP.

Fig. 6.. Midnolin catches regions within its substrates that constitute a β strand degron.

(A) AlphaFold structure prediction of midnolin bound to its substrate IRF4 reveals an adopted β strand capture model. (B) Midnolin requires the predicted β strand within IRF4 to promote degradation. The GPS IRF4 reporters were stably expressed in MIDN KO HEK-293T cells and a control BFP or midnolin co-expressing BFP were transiently transfected before analyzing the GFP/DsRed ratio by flow cytometry. (C) Predicted β strands are necessary for interaction with midnolin. Immunoblotting was performed from anti-FLAG immunoprecipitants of 3xHA-midnolin knock-in HEK-293Ts transfected with 2xFLAG-tagged substrates. For FosB, the comparison is between the full-length protein and ΔFosB. Cells were treated with 10 μM MG132 for 6 hours. See methods for the truncation boundaries. (D) Amino acid frequency of midnolin substrate β strands predicted by AlphaFold reveals a strong preference for hydrophobic residues. ‘Inward’ is defined by the residues buried within the Catch domain, while ‘outward’ is defined by the solvent-exposed residues. (E) The hydrophobicity of residues within the β strand was determined by a mean hydrophobicity index at pH 7 (58, 59) of residues immediately preceding, within, or immediately following the β strand. (F) AlphaFold structure prediction of the midnolin Catch domain bound to IRF4. (G) Hydrophobic β strand residues buried within the Catch domain are required for midnolin interaction. Similar assay as (C) from cells transfected with the 2xFLAG-tagged IRF4 constructs. (H) Midnolin requires the hydrophobic β strand residues buried within the Catch domain to promote degradation. Similar assay as (B) (I) Regions encompassing predicted β strand(s) are sufficient for conferring an interaction with midnolin. Similar assay as (C) from cells transfected with the indicated 2xFLAG-GFP-peptide fusions.

Fig. 7.. Model for how midnolin functions to promote ubiquitination-independent proteasomal degradation.

Midnolin is induced by growth factors and neurological stimuli and its overexpression is sufficient to cause the degradation of its targets including transcription factors such as c-Fos, FosB, ERG1, NR4A1, IRF4, and potentially many other proteins within the nucleus, where midnolin primarily resides. The degradation of its substrates does not require ubiquitination. Instead, midnolin utilizes its Catch domain to bind unstructured hydrophobic regions within substrates that have the potential to form a β strand that functions as a midnolin degron. Midnolin associates with the proteasome using its long C-terminal α helix and promotes the destruction of Catch-bound substrates via its N-terminal ubiquitin-like domain. Structures of the midnolin domains are derived from AlphaFold predictions. How the C-terminal α helix of midnolin binds the proteasome, whether a conformational change occurs after substrate binding, and how the ubiquitin-like domain confers degradative activity require further investigation (created with BioRender.com).