Photoproximity Labeling from Single Catalyst Sites Allows Calibration and Increased Resolution for Carbene Labeling of Protein Partners In Vitro and on Cells

The cell surface proteome (surfaceome) plays a pivotal role in virtually all extracellular biology, and yet we are only beginning to understand the protein complexes formed in this crowded environment. Recently, a high-resolution approach (μMap) was described that utilizes multiple iridium-photocatalysts attached to a secondary antibody, directed to a primary antibody of a protein of interest, to identify proximal neighbors by light-activated conversion of a biotin–diazirine to a highly reactive carbene followed by LC/MS (Geri, J. B.; Oakley, J. V.; Reyes-Robles, T.; Wang, T.; McCarver, S. J.; White, C. H.; Rodriguez-Rivera, F. P.; Parker, D. L.; Hett, E. C.; Fadeyi, O. O.; Oslund, R. C.; MacMillan, D. W. C. Science2020, 367, 1091–1097). Here we calibrated the spatial resolution for carbene labeling using site-specific conjugation of a single photocatalyst to a primary antibody drug, trastuzumab (Traz), in complex with its structurally well-characterized oncogene target, HER2. We observed relatively uniform carbene labeling across all amino acids, and a maximum distance of ∼110 Å from the fixed photocatalyst. When targeting HER2 overexpression cells, we identified 20 highly enriched HER2 neighbors, compared to a nonspecific membrane tethered catalyst. These studies identify new HER2 interactors and calibrate the radius of carbene photoprobe labeling for the surfaceome.


SI Table
. Performance of the three controls chosen for comparison.3. Log2x transformed LFQ values for each replicate and emitter site.4. STRING analysis values.

Traz (Traz) Fab labeling protocol
Single site mutated Traz fab's were utilized from a previous study 32 .The fab mutants were grown following a previously established protocol 1 .DE3 cells were grown in supplemented Terrific Broth at 37° C for 6 hours, followed by 18 hours of growth at 30°C.After growth cells were lysed on a Qsonica sonicator in PBS and purified by protein A affinity chromatography using VWR Hitrap protein A 1mL columns.After purification proteins were verified by SDS-Page gel and concentrated for oxaziridine reactions.Reactions were performed following protocols published in Elledge.Et al, 2020 2 .Briefly, proteins were reacted at room temperature on the bench at 10µM antibody with 100µM oxaziridine in PBS.Labeling was allowed to occur for one hour before verifying labeling by Waters Xevo UPLC-MS.After oxaziridine labeling, proteins were buffer exchanged by Zeba spin column before loading with DBCO-Ir probe.Labeling was performed with 5-fold molar excess probe overnight at 4° C before buffer exchange by Zeba column and confirming labeling by Mass spectrometry.Estimates of Iridium probe size were performed by building the catalyst in Avogadro 3 and finding the maximum possible distance given stable geometries.

In vitro DET labeling experiments
DET labeling experiments were performed in vitro by mixing 9µM Traz-catalyst with 100µM diazirinebiotin.Blue light illumination was performed on a PennPhD Photoreactor M2 for 10 minutes at full power.After illumination samples were flash frozen for later thawing and processing.Preparation for Mass Spectrometry was done using a Preomics Ist kit.Procedures were followed as described in the user manual with the modification of lysis by only a half volume of Trypsin/LysC for half an hour on the bench, to avoid excess protease activity on the samples.Samples were dried on a Genevac EZ-2 centrifugal evaporator before resuspension and normalizing to a 200ng/µL concentration.1µL of samples was run on a Bruker TimsTOF.Samples were run on a NanoViper system with a Bruker TEN 10cmx75µm nanoflow column.In vitro samples were run on a 45 minute gradient from 4-32% solvent B (Solvent A: 2% acetonitrile with 0.1% TFA Solvent B: Acetonitrile with 0.1% TFA).In vitro analysis was performed on the PEAKs 4 server using a 1% FDR and searching for a 616.25424 diazirine biotin label.A 20 PPM precursor mass error tolerance and 0.1 Da fragment mass error was used in analysis.
Labeled residues were collected and analyzed based on the number of instances of labeling.First analyzing by distance of labeling vs angstrom distance from the Ir-catalyst bearing residue.All distances were calculated from the PDB 1N8Z 5 and residue identities were taken from the PDB file.Analysis was performed in BioPython 6,7 .The second round of in vitro experiments was performed with equimolar concentrations of Her2 and catalyst-Traz.For consistency a 9µM antibody-catalyst was used along with 9µM Her2 and 100uM diazirine-biotin.Preparation for Mass Spectromety was then performed identically to the previous samples.For quantification the Mass Spectrometry .Mzml files were downloaded and reanalyzed in Skyline 8 .In Skyline analysis was performed using the same FDR and settings as Peaks.Skyline quantification was performed with a linear regression fit and normalizing to the total ion count.Results were exported from Skyline with the calculated quantification values, peptide identity, experiment, and isotope dot product.Quantified peptides were filtered by an isotope dot product of 0.85 or greater.In the "Her2 protection" analysis residues within 7.5 angstroms were identified from the 1N8Z crystal structure and collected.Residues within 10 angstroms were identified by collecting residues within 10 angstroms and removing any of the residues identified within 7.5 angstroms to find the unique residues.The same was done for 15 angstroms distance.Then quantification was performed by summing all the quantified labeling at any residue within each unique distance group.

On Cell labeling protocol
On cell labeling was performed with SKBR3 cells obtained from the UCSF cell culture facility.Cells were grown in T-175 flasks with RPMI media supplemented with 10% FBS and 1% penicillin/streptomycin.Cells were grown up to 70-80% confluency before harvesting.Cell harvesting was performed by first washing flasks with PBS before scraping cells into PBS.Cells were checked for count and viability, ensuring that they were at least ~80% live before aliquoting and storing on ice.In all on cell experiments 5 million cells were used.After scraping and washing, cells were collected and antibodies for labeling were added at 5µg and incubated on a rotator at 4°C for 30 minutes for binding.
Single site Fab labeling was performed by immediately incubating the cells with 5µg fab and allowing 30 minutes for binding on a rotator at 4° C. Following Fab binding 3 rounds of PBS washes were performed to wash the cells and avoid unbound antibody.After washing the cells were brought up in 1mL of 100µM diazirine biotin in PBS and irradiated with blue light for 10 minutes.After irradiation the cells were washed 3 final times with PBS to avoid residual diazirine in the samples then flash frozen for later processing.In the case of Secondary catalyst labeling the same procedure was followed, with the exception of two antibody binding sets.First a Traz-IgG (Obtained from Selleck chemicals catalog #A2007) antibody was bound for 30 minutes at 4° C on a rotator.Followed by washing then binding of the Ir catalyst conjugated secondary for another 30 minutes.The Ir secondary (ThermoFisher catalog # 12-4998-82) was prepared by the same NHS labeling protocol used in the micromap protocol for consistency 9 .After labeling, cells were washed identically to the Fab samples and stored.Biotinylated proteins were first prepared for mass spectrometry by resuspension of the frozen cells into 2X RIPA buffer with added Halt ThermoScientific protease inhibitor cocktail and 5mM EDTA.Cells were lysed on a by sonicator and after incubated in the RIPA buffer for 10 minutes on ice.After they were centrifuged for 10 minutes at 21,000g to pellet cellular debris.The supernatant was collected and batch bound to Thermo Scientific High capacity streptavidin-agarose beads for 2 hours at 4° C while on a rotator.After the beads were washed on Biorad Bio-spin small volume chromatography columns using vacuum to wash 3 solutions over the beads.5mL of 1X RIPA with 1mM EDTA was flowed over beads, taking care to disrupt the bead bed during washing.Followed by 5mL of PBS with 1M NaCl and finally 5mL 50mM ammonium bicarbonate with 2M urea again taking care to disrupt the beads in order to ensure full washing.These washed beads were then processed for mass spectrometry using the Preomics Ist kit.After preparation the peptides were dried on a SP Genevac before bringing up in 2% acetonitrile and .1% TFA in Mass Spectrometry grade water and diluting all samples to 200ng/uL.All samples were injected twice onto a Bruker TimsTOF.Samples were run on a NanoViper system with a Bruker TEN 10cmx75µm nanoflow column.In vitro samples were run on a 90 minute gradient from 4-32% solvent B (Solvent A: 2% acetonitrile with 0.1% TFA Solvent B: Acetonitrile with 0.1% TFA) This procedure was used for all cellular samples.Fatty acid catalyst experiments were performed by lipid starving the cells for 1 hour by removing media, washing once with PBS, then replacing media with RPMI with 1% Pen/strep but no supplemented FBS.After lipid starvation the media was removed and Azido fatty acid media was added.To prep the Azido caproyl acid a 100uM solution in DMSO was saponified with 20% molar equivalent potassium hydroxide and by heating to 65°C for 15 minutes.A 100uM solution of fatty acid in RPMI was warmed to 37°C before addition to the lipid starved cells.Cells were allowed an additional hour to grow in the lipid supplemented media.After the final incubation cells were washed twice in PBS and scraped by the same protocol as with Traz labeled cells.After scraping cells were washed once in PBS and then reacted for 1 hour at 37°C with 100uM DBCO-Ir probe.After the reaction cells were washed 3x with PBS to remove background and brought into PBS with 100uM Diazirine-biotin.Cells were blue light irradiated for 10 minutes before washing twice with PBS to remove unreacted diazirine and freezing.Cells were then processed identically to the previous samples.WGA-HRP CSC experiments were performed as described in publication 49 .The same T-175 flasks were used for collecting cells.Samples were processed identically to all previous cellular labeling samples.In the controls using a Ir-catalyst secondary with no primary the same procedure as catalyst-Fabs was used.Briefly after collecting cells they were incubated for 30 minutes at 4°C to allow for binding.After cells were washed twice before resuspending in 100µM diazirine biotin and labeling with blue light.After washing the samples post blue light they were frozen for preparation.

Proteomics sample processing
Proteomics samples were processed in Maxquant 2.0.3.0 10,11 .Peptide search was performed with the SwissProt GOCC Plasma Membrane database.A FDR rate of 0.01 was used for searches, LFQ values were obtained from Maxquant using their classic normalization.After peptide identification additional filtering was performed in Perseus 12,13 .First peptides were excluded which were identified by a single site, identified as reverse peptides, or peptides identified by Maxquant as likely contaminants.Next proteins were eliminated which were identified by only one unique peptide.Experiments were grouped together by labeling site.Following this, proteins were Log2(x) transformed and proteins were eliminated which were not identified by LFQ in 3 replicates of at least one experimental set.Imputation was performed with a down shift of 2.2 and sampling a width of 0.3 using the Perseus "Replace missing values from normal distribution" command.Following this Volcano plots were created between groups with significance tested by a t-test, using 250 randomizations with a FDR of 0.05 and S0 of 0.1.In presented plots IGK was removed as it originated from the antibody used in the experiment.Hierarchical clustering was performed by averaging groups by geometric mean in Perseus.Clustering was performed in Python using the Seaborn statistical data visualization package 14 , an unweighted pair group method with arithmetic mean was used with a Euclidian distance metric.Proteins highlighted for further literature search were chosen by taking a collection of the most enriched proteins (Over 4x log2x enriched) from the 8 emitter sites.All of the proteins were indicated to interact with ERBB2's network by STRING analysis.The full list of proteins meeting this level is: CAN1, MPP6, ENOA, MGST1, ANXA1, BCAM, TGFR1, PTPRF, GTR1, ADT1, EPHB3, MPZL1, ATD3B, ST14, UBR4, LSR, PLXB2, KHDR1, ERBB2, and IGK.A literature search revealed interesting roles for each of these proteins but the final 5 chosen for presentation in the main text were those with a diverse set of possible roles.
All protein visualizations were performed in Pymol.

SI Figure 6 .SI Figure 7 .SI Figure 8 .
Biological details and overlap of high confidence identified neighbors.Volcano plots for each of the eight Traz catalysts used in the study.Comparing the LFQ values from CSC, FA azide, and unbound Ir-secondary controls.

Figure 1 .SI Figure 2 .SI Figure 4 .SI Figure 5 .SI Figure 6 .
Methionine sites and conjugation chemistry used for step-wise addition of Ir-catalyst to the Fab of Traz. A. Ribbon diagram showing front and back view of the complex between the extracellular domain of HER2 bound with Traz.The eight sites across light and heavy chains in Traz that were mutated to methionine in order to perform site-specific oxaziridine conjugation are shown in red dots distributed.The HER2 structure is shown in Teal with the Traz light chain in gray and heavy chain in light cyan B. The two reaction steps involved to create the DET-probe conjugated to Traz are shown (i) oxaziridine conjugation chemistry onto the Met to attach the azide and (ii) attachment of the Ir catalyst using DBCO click chemistry.The protein is minimized in scale in order to highlight the structure of the probe.Intramolecular labeling of Traz by the Ir-catalyst from eight different Met sites.A. MS traces of the Ir-catalyst conjugated Fabs.100% labeling was not achieved in all cases but was not found to be essential to experiments.B. Venn diagram of the labeling of for six of the eight catalyst sites on the Traz light and heavy chain (Light cyan and grey ribbons, respectively).Full reports of labeled sites are available in the supplemental tables.Most sites overlap but each catalyst has a unique detailed labeling pattern.C. Sites of carbene insertion (red dots) from each of the eight catalyst sites (shown by star).D. Individual plots of the intramolecular labeling of each of the 8 emitter conjugated Trazs.Labels are identified from experiments without Her2 binding.Distance dependence for the labeling of the same residue on HER2 from each of the eight catalyst sites on Traz. A. Structure of Traz (light cyan and grey ribbons) bound to HER2 (teal ribbons).Traz catalyst sites are indicated in red dots and example common proximal or distal residues on HER2 labeled by the carbene indicated in orange dots.B. Distance dependence for carbene labeling at five common residues on HER2 from the eight different catalysts on Traz.Note that proximal residues show relatively flat even profiles across distance while those distal show a loss at greater distance dependence and both terminate near 110Å.Comparing the labeling of R66M L alone and the secondary labeling strategy relative to the most enriched proteins from all eight Traz catalysts.The vast majority of the proteins identified by R66M L alone and most of the secondary emitter significantly enriched proteins are the same as those seen in the larger dataset.Biological details and overlap of high confidence identified neighbors.A. A Venn diagram showing the overlap for all high confidence identified proteins identified in the FA-emitter and the Fabemitters.Virtually all of the high confidence proteins identified in the 8 Fab-emitters were detected in the FA-emitter experiment.B. A String analysis for eachof the highly enriched proteins from the Fabemitter volcano plots.There was a strong association with HER2 as 71 of 88 proteins have a known link.C. A Venn diagram showing there is considerable overlap among highly enriched proteins from the Fabemitters and a list of proteins present in at least four of the Fab-emitters.D. The cellular location GO annotation of the highly enriched proteins demonstrating that hits are predominantly identified as membrane or extracellular, with only 16 annotated as cytoplasmic E. A breakdown of the strategy used to narrow identified proteins down to high confidence potential targets.SI Figure 8. Comparing the LFQ values from CSC, FA azide, and unbound Ir secondary controls.A. The mean LFQ of the FA azide and WGA-HRP controls.The line is set to represent even LFQ values.B. The mean LFQ of the FA azide and Ir secondary without a primary control.The line is set to represent identical LFQ values.

Table 1 .
Performance of the three controls chosen for comparison.

Table 2 .
Mean LFQ values from experiments.Fab-catalyst sites are grouped together with means calculated for the 3 biological replicates and 2 technical replicates.The fatty acid catalyst experimental group is shortened to FA.The secondary-catalyst experiment is shortened as 2nd.