Site-Specifically-Labeled Antibodies for Super-Resolution Microscopy Reveal In Situ Linkage Errors

The precise spatial localization of proteins in situ by super-resolution microscopy (SRM) demands their targeted labeling. Positioning reporter molecules as close as possible to the target remains a challenge in primary cells or tissues from patients that cannot be easily genetically modified. Indirect immunolabeling introduces relatively large linkage errors, whereas site-specific and stoichiometric labeling of primary antibodies relies on elaborate chemistries. In this study, we developed a simple two-step protocol to site-specifically attach reporters such as fluorophores or DNA handles to several immunoglobulin G (IgG) antibodies from different animal species and benchmarked the performance of these conjugates for 3D STORM (stochastic optical reconstruction microscopy) and DNA-PAINT (point accumulation in nanoscale topography). Glutamine labeling was restricted to two sites per IgG and saturable by exploiting microbial transglutaminase after removal of N-linked glycans. Precision measurements of 3D microtubule labeling shell dimensions in cell lines and human platelets showed that linkage errors from primary and secondary antibodies did not add up. Monte Carlo simulations of a geometric microtubule-IgG model were in quantitative agreement with STORM results. The simulations revealed that the flexible hinge between Fab and Fc segments effectively randomized the direction of the secondary antibody, while the restricted binding orientation of the primary antibody’s Fab fragment accounted for most of the systematic offset between the reporter and α-tubulin. DNA-PAINT surprisingly yielded larger linkage errors than STORM, indicating unphysiological conformations of DNA-labeled IgGs. In summary, our cost-effective protocol for generating well-characterized primary IgG conjugates offers an easy route to precise SRM measurements in arbitrary fixed samples.


Experimental Section
LC-ESI-MS analysis. Spectra were measured on a Waters LCT Premier mass spectrometer.
10 μg of antibody was incubated with DTT (20 mM final concentration) at 37 °C for 30 min.
The eluent was ionized using an electrospray source (ESI+). Data were acquired with MassLynx 4.1 and deconvolved using MaxEnt1.
Microscopy setups for STORM and DNA-PAINT. SMLM imaging was performed at room temperature (21 °C) on two home-built setups of similar layout but located in different labs.
Setup 1 at EMBL, as previously described in Li et al., 1 was used for STORM and DNA-PAINT in U2OS cells (Fig. 2 b,c,d,f) while Setup 2 at RCSI was used for DNA-PAINT in platelets ( Fig. 2 g,h). Setup  Photonics Ltd, UK). The effective pixel size was 108 nm. In both setups, lasers are passed through a speckle reducer (LSR-3005-17S-VIS; Optotune, Dietikon, Switzerland) and coupled into a multimode fiber (M105L02S-A; Thorlabs, Newton, NJ, USA) whose output is magnified by an achromatic lens, filtered to remove fiber-generated fluorescence (390/482/563/640 HC Quad; AHF, Tübingen, Germany), and imaged into the sample. 2 Laser and camera triggering were controlled by an FPGA (Mojo; Embedded Micro, Denver, CO, USA). A closed-loop focus lock system was implemented, using total internal reflection of a near-infrared laser (iBeam smart 785, Toptica Photonics AG) at the coverslip-buffer boundary, its detection by a quadrant photodiode, and the output voltage to control the objective piezo (P-726.1CD; Physik Instrumente Ltd, Bedford, UK). The fluorescence emission was filtered by a bandpass filter (700/100, AHF) and the PSF engineered using an astigmatic lens. Microscopes were controlled using µManager 2.0gamma and EMU.   Table S1) is shown as an example after a) PNGaseF treatment and after b) modification with H2N-PEG3-N3 and DIBO-Alexa Fluor 647, which contributes to a difference in mass between non-glycosylated heavy chain and modified heavy chain of ~1370 Da. Shown are the radius r and thickness (full width at half maximum, fwhm) of the labeling shell as derived from cross-sectional fits. Compared to fits of pooled cross-sections (cf. Figure 2 in the main manuscript), the fitting tended to yield increased shell thicknesses, but overall confirmed the differences between direct (a+d) or indirect (b,c,e,f) labeling, as well as between STORM and DNA-PAINT.   Table 2 in the main text. Figure S6. Simulated reporter distribution of primary (left) or secondary (right) antibodies around microtubules depending on the location and orientation of the epitope on α-tubulin. The location/orientation is parameterized by the single parameter ϕ1 (see Figure 3a and Table 2 in the main text). Left panels: Contour plots of the reporter location density projected onto the plane of the microtubule cross-section. Right panels: radial distribution of the reporter. Variation of ϕ1 from straight a) 0° to increasingly oblique positions b) 15°, c) 30°, and d) 45° in the groves of the microtubule. Accordingly, the label moved closer to the microtubule while the spread remained largely unchanged. Figure S7. Simulated reporter distribution of primary (left) or secondary (right) antibodies around microtubules depending on the flexibility of the Fab-epitope binding orientation. The flexibility is parameterized by the parameter δϕ (see Figure 3a and Table 2 in the main text). Left panels: Contour plots of the reporter location density projected onto the plane of the microtubule cross-section. Right panels: radial distribution of the reporter. Variation of δϕ from rigid a) 5°, b) 22.5° to flexible c) 45°, d) 75°. With increasing flexibility of the Fab-epitope orientation, the primary antibody label moved closer to the microtubule; notably, no change was observed for peak of the secondary label distribution. In both cases, a minor widening of the radial distribution resulted from an increased Fab-epitope flexibility. Figure S8. Investigation of potential reasons for larger and thicker labeling shells obtained by DNA-PAINT. a) STORM of indirectly labeled microtubules using DNA-labeled primary antibodies yielded substantially thicker and larger labeling shells than with unlabeled primaries. b) Adjusting model parameters for straighter antibody labeling complexes to match the peak position of the experimentally determined label distribution leads to a drastically decreased labeling shell thickness, as opposed to a thicker labeling shell seen in DNA-PAINT. c) Circular dichroism (CD) spectrogram of unconjugated donkey anti-mouse IgG, deglycosylated IgG, and DNA-conjugated IgG. The latter was corrected for the CD spectra of an equivalent amount of DBCO-DNA along. A more negative ellipticity was observed in the region around 217 nm for the DNA-IgG conjugate, in line with reported changes for chemical denaturation of IgGs. 5