Global Reprogramming of Host SUMOylation during Influenza Virus Infection

Summary Dynamic nuclear SUMO modifications play essential roles in orchestrating cellular responses to proteotoxic stress, DNA damage, and DNA virus infection. Here, we describe a non-canonical host SUMOylation response to the nuclear-replicating RNA pathogen, influenza virus, and identify viral RNA polymerase activity as a major contributor to SUMO proteome remodeling. Using quantitative proteomics to compare stress-induced SUMOylation responses, we reveal that influenza virus infection triggers unique re-targeting of SUMO to 63 host proteins involved in transcription, mRNA processing, RNA quality control, and DNA damage repair. This is paralleled by widespread host deSUMOylation. Depletion screening identified ten virus-induced SUMO targets as potential antiviral factors, including C18orf25 and the SMC5/6 and PAF1 complexes. Mechanistic studies further uncovered a role for SUMOylation of the PAF1 complex component, parafibromin (CDC73), in potentiating antiviral gene expression. Our global characterization of influenza virus-triggered SUMO redistribution provides a proteomic resource to understand host nuclear SUMOylation responses to infection.


TAP Procedure and Mass Spectrometry.
Lysates were diluted 25x in order to dilute out the denaturing 2% SDS and passed over IgG sepharose (GE Healthcare), which was followed by enzymatic removal of the Protein A portion of the TAP-tag (see Fig. S3A) using TEV protease (Promega). The resulting eluate was then affinity purified on calmodulin sepharose (GE Healthcare) followed by protein elution with buffer containing 10 mM EGTA, and protein recovery by precipitation with 100% TCA (w/v) and acetone washing. Purification resulted in ~25µg of protein sample that was resuspended in 30µl of 2x LDS sample buffer (Invitrogen).
Crude sample (~50µg) was also mixed 1:1 with 2x LDS sample buffer. Both purified and crude samples were resolved on NuPAGE Novex 10% Bis-Tris polyacrylamide gels using MOPS buffer (Invitrogen). Gel-fractionated proteins were stained with Coommassie blue and the gel was sliced into 13 sections as outlined in Fig. 3. Protein slices were subjected to in-gel digestion with trypsin (Promega) essentially as described previously (Shevchenko et al., 2006). The resulting peptide mixtures were vacuum-dried and resuspended in 30µl of 1% formic acid prior to analysis by LC-MS/MS on a Q Exactive mass spectrometer (Thermo Scientific) coupled to an EASY-nLC 1000 liquid chromatography system via an EASY-Spray ion source (Thermo Scientific) running at 75 µm x 500 mm at 45ºC on an EASY-Spray column. An elution gradient duration of 240 min was used. Data were acquired in the data-dependent mode. Full scan spectra (m/z 300-1800) were acquired with resolution R = 70,000 at m/z 400 (after accumulation to a target value of 1,000,000 with maximum injection time of 20 ms). The 10 most intense ions were fragmented by HCD and measured with a target value of 500,000, maximum injection time of 60 ms and intensity threshold of 1.7e3. A 40 second dynamic exclusion list was applied.

MaxQuant Analyses.
All raw files generated by MS analysis were processed with MaxQuant software (version 1.3.0.5)    First, all MaxQuant-defined unwanted hits (e.g. 'reverse' (peptide sequences that would match other sequences if reversed), 'contaminants', and 'identified by site' (only identified by modification site)) as well as any remaining suspected internal contaminants (e.g. keratins, immunoglobulins, non-human origin proteins) were removed. All such prefiltered content was copied into 2 identical tabs and named 'Crude' and 'Purified'. In the 'Crude' list, all hits with 0 unique crude peptides and/or no crude ratios reported were removed. In the 'Purified' list, all hits with 0 unique purified peptides and/or no purified ratios reported were removed. Although MaxQuant uses built-in normalization algorithms to account for variable isotope purity/incorporation or error in lysate mixing, it is only applicable to values distributed in a unimodal Gaussian manner. We therefore applied an alternative method of normalization for our 'purified' data, which due to their highly purified nature and substantial changes in SUMOylation means they are not unimodal.
Given that the majority of proteins from our crude samples were unaffected by any treatment (Figs. S3D & S3E), the median of the raw M/L, H/L, and H/M ratios for proteins was calculated and applied to normalize the raw ratios from crude samples as well as from corresponding TAP-purified samples. In order to calculate normalized values, raw ratios were divided by the normalization factors below:  (table above). All hits with ratios above a certain value (on the positive part of the axis) were defined as substrates with increased SUMOylation in response to treatment, while 'hits' below a certain value (on the negative part of the axis) were defined as substrates with decreased SUMOylation (values stated in table above).
A search for ubiquitylation sites (GlyGly) was also included in the processing of the raw mass spectrometry data from purified samples. However, very few (~20) ubiquitylation sites were identified (Tables S1 & S2), none of which were subsequently classified as SUMO substrates that change in abundance with IAV infection, and therefore were not followed-up in this study. Notably however, within our A549 SUMO-modified proteome, we identified ubiquitylation sites on ubiquitin itself at lysines 48 and 63. These sites may represent certain topologies of hybrid SUMO-ubiquitin chains (Praefcke et al., 2012).

Manual Data Processing of 'Slice-by-Slice' Analyses to Confirm SUMO Modification.
For selected putative SUMO substrates, predicted molecular weight (preMW) was compared to their observed electrophoretic mobility (obsEM) in both 'crude' and 'purified' datasets. Such a method has been developed previously to confirm ubiquitin conjugation to target proteins (Peng and Cheng, 2005). We assumed that a protein should run according mostly to its mass by SDS-PAGE, and should thus be detected within the gel slice (1-13) covering its approximate MW. For almost all proteins in crude lysates, the preMW was found to correspond to its obsEM when individual protein abundance in each gel slice was analyzed. However, bona fide covalent SUMO substrates would have an increased MW, and should therefore be detected in slices corresponding to larger masses.
The more highly modified the substrate, the larger the obsEM/preMW ratio should be.
This would be most apparent in TAP-SUMO-purified samples given that SUMO is usually only attached to a small proportion of the total population of a protein. Thus, for selected substrates, protein intensity values in individual gel slices from TAP-SUMOpurified material were compared between SILAC conditions. Raw intensity values for M and H data were normalized with the M/L and H/L normalization factors, respectively (table above). All three SILAC ratios were normalized as described for 'global' analyses.
Among all tested proteins with sufficient data, obsEM/preMW ratios were close to 1 for >95% of crude and contaminant proteins (no SUMO modification), while ratios for putative SUMO substrates were usually infinitely high (no protein observed at its preMW) for >95% of putative SUMO substrates. Protein intensities (abundance) and SILAC ratios within each slice were taken into account when creating graphical representations of these data as shown in Figs. 4B and S5.

Data Comparisons with Other Studies.
To compare SUMOylation responses between stresses, H/M ratios (i.e. treatment vs nontreatment) for SUMO substrates common to two different datasets were visualized on tsMAPs with Pearson's coefficient values indicated (Fig. 5). All datasets were generated using the same or very similar experimental and data processing methodology (Fritah et al., 2014;Golebiowski et al., 2009;Tatham et al., 2011;Yin et al., 2012). Comparison analyses were made with IPA (Ingenuity Pathway Analysis) software by matching respective gene names and ratios. To assess the reliability of our data, we also compared the substrates identified from the combined SUMO1 and SUMO2 IAV experiments in this study with either the respective crude lysate proteomes (this study) or lists of identified SUMO substrates from other studies (we first generated a database of SUMO substrates defined by publically-available proteomic studies either from our own laboratories or from independent groups (Barysch et al., 2014;Blomster et al., 2009;Fritah et al., 2014;Ganesan et al., 2007;Golebiowski et al., 2009;Hendriks et al., 2014;Matafora et al., 2009;Pungaliya et al., 2007;Rosas-Acosta et al., 2005;Schimmel et al., 2008;Tammsalu et al., 2014;Tatham et al., 2011;Vertegaal et al., 2006)). Comparisons were performed and output collated from IPA software prior to results being depicted as Venn diagrams (Figure S3H-I).

RT-qPCR.
Total RNA was isolated using the RNAeasy Plus Kit (Qiagen). RT-qPCRs were performed as a two-step process, and each sample was normalized to an endogenous control (18S rRNA or GAPDH (ii) were repeated 40 times. Data were analyzed using 7500 Fast system software (v2.0.5; Life Technologies).

Xbp1 Splicing Assay.
XBP1 mRNA levels were determined using the One-Step AccessQuick RT-PCR system (Promega). Briefly, 1μg of total RNA was reverse-transcribed with AMV Reverse Transcriptase (AMV RT) and the cDNA amplified with Tfl DNA polymerase using XBP1 specific primers (5-AGTGGCCGGGTCTGCTGAGT and 5-GGCTTCCAGCTTGGCTGATGACG). Following separation on a 1% agarose gel, the RT-PCR products corresponding to the unspliced and spliced (26 nucleotide deletion) forms of XBP1 mRNA were visualized by ethidium bromide staining.

Statistical Analyses.
Statistical analyses were performed using an unpaired two-tailed Student's t-test. The p values for significance are stated in the figure legends.