A specific hybridisation internalisation probe (SHIP) enables precise live-cell and super-resolution imaging of internalized cargo

Facilitated by the advancements in microscopy, our understanding of the complexity of intracellular vesicle traffic has dramatically increased in recent years. However, distinguishing between plasma membrane-bound or internalised ligands remains a major challenge for the studies of cargo sorting to endosomal compartments, especially in small and round cells such as lymphocytes. The specific hybridization internalisation probe (SHIP) assay, developed for flow cytometry studies, employs a ssDNA fluorescence internalisation probe and a complementary ssDNA quenching probe to unambiguously detect the internalized receptors/cargo. Here, we adopted the SHIP assay to study the trafficking of receptor/ligand complexes using B lymphocytes and B cell receptor-mediated antigen internalization as a model system. Our study demonstrates the potential of the SHIP assay for improving the imaging of internalized receptor/ligand complexes and establishes the compatibility of this assay with multiple imaging modalities, including live-cell imaging and super-resolution microscopy.

A specific hybridisation internalisation probe (SHIP) enables precise live-cell and super-resolution imaging of internalized cargo Sara Hernández-Pérez 1,2,3* & Pieta K. Mattila 1,2,3* Facilitated by the advancements in microscopy, our understanding of the complexity of intracellular vesicle traffic has dramatically increased in recent years. However, distinguishing between plasma membrane-bound or internalised ligands remains a major challenge for the studies of cargo sorting to endosomal compartments, especially in small and round cells such as lymphocytes. The specific hybridization internalisation probe (SHIP) assay, developed for flow cytometry studies, employs a ssDNA fluorescence internalisation probe and a complementary ssDNA quenching probe to unambiguously detect the internalized receptors/cargo. Here, we adopted the SHIP assay to study the trafficking of receptor/ligand complexes using B lymphocytes and B cell receptor-mediated antigen internalization as a model system. Our study demonstrates the potential of the SHIP assay for improving the imaging of internalized receptor/ligand complexes and establishes the compatibility of this assay with multiple imaging modalities, including live-cell imaging and super-resolution microscopy.
Intracellular membrane trafficking has not only critical housekeeping functions but also controls various celltype-specific functions regulating, for instance, signalling cascades and cell fate. Over the years, a general view of the vesicle pathways has been relatively well established. In particular, in the last decade, improvements in fluorescence microscopy have aided the discovery of different vesicle subpopulations and transport machineries. These improvements have increased our understanding of the location and behaviour of different vesicle populations, not possible to decipher otherwise using biochemical techniques [1][2][3][4] . Nevertheless, despite the improvements, imaging of vesicular pathways is challenged by the dynamic and heterogeneous nature of the vesicles and vesicular networks in all cell types. A particular challenge is confronted in small and round cells, such as lymphocytes, where the reduced size of the cytoplasm and the resulting immediate vicinity of the plasma membrane generate major obstacles for imaging approaches.
B lymphocytes, or B cells, are a crucial part of the adaptive immune system, initiating the antibody responses against a vast repertoire of different antigens. Precise coordination of the endosomal machinery is essential for processing the recognised antigens and triggering antibody responses. Through their B cell receptor (BCR), B cells recognise specific antigens that are internalized and directed to a specialized vesicular pathway. This antigen processing pathway enables controlled digestion of antigens into peptides that are then loaded onto the major histocompatibility complex class II (MHCII) 5 . The peptide-MHCII (pMHCII) complex is transported back to the cell surface and recognized by cognate CD4 + lymphocytes (T helper cells, T H cells), providing a second activation signal to B cells. Reciprocally, pMHCII recognition also stimulates the cognate T H cells to orchestrate other branches of the immune system and generate CD4 + T cell memory 6 . Hence, antigen internalisation and processing is a critical process that involves specialized endosomal compartments, and it needs to be tightly regulated to trigger B cell differentiation and effector T cell responses.
In our previous work, we characterised the antigen processing route in B cells by analysing different vesicle markers using spinning disk confocal microscopy (SDCM) 7 . However, unambiguous distinction between membrane BCRs and newly internalised BCRs remains a major obstacle in imaging approaches, as BCR is the most abundant receptor on the B cell surface (100,000-300,000 copies per cell) 8,9 and engaged BCRs often form clusters on the cell membrane. Inaccurate determination of the extracellular and intracellular signal can ultimately skew the analyses or misguide the interpretations of the data. Traditionally, methods such as stripping of membrane proteins or treatment with proteases (i.e. trypsin) have been used to address similar problems 10,11 . However, some proteins can resist the stripping procedure, and these methods can compromise cell viability. Anti-fluorochrome antibodies can also be used to quench non-internalised fluorochromes, but this approach is limited to a few fluorophores, such as Alexa Fluor 488, Alexa Fluor 594, or fluorescein isothiocyanate (FITC) 12,13 . Alternatively, the non-internalized pool can be labelled prior to permeabilization to distinguish it from the internalized pool (for example, Fig. S2A-C in Hernández-Pérez et al. 7 ). Nevertheless, quantification of the internalised antigen using this approach is time-consuming and not straightforward, and notably, none of these methods is compatible with live imaging. In an effort to address this problem, an elegant approach based on photoactivation (PA) of fluorescent proteins was recently used to visualize and quantify endocytic trafficking of T-cell surface receptors 14 . As a downside, the expression of an exogeneous proteins as well as a suitable high-end microscope setup is required in this method.
In 2013, Liu and Johnston developed an elegant method termed specific hybridization internalization probe (SHIP) assay to differentiate between internalised and non-internalised material in live cells 15 . This assay utilises a short (20-mer) single-stranded DNA (ssDNA) fluorescent internalization probe (FIP) coupled to the ligand of interest and a complementary ssDNA quenching probe (QP). Therefore, the SHIP assay can be used to investigate the kinetics of protein internalisation, and it has been employed in the past to analyse the nanoparticle internalization, antibody-dependent phagocytosis, and MHCII turnover in dendritic cells [16][17][18][19] . Notably, the SHIP system has been used almost exclusively in high-throughput flow cytometry studies, leaving its potential for imaging unexplored.
Here, we labelled antibodies recognising the IgM BCR expressed in A20 B cells with an Abberior® STAR 635Por ATTO 647N-conjugated FIP to test the applicability of the SHIP system to the study of antigen trafficking in B cells not only in conventional confocal imaging but also in live and super-resolution imaging. We report efficient abolishment of the membrane-bound signal by the quencher probe, clearly enhancing the detection of internalized antigen vesicles, and provide a general proof-of-concept of the power of the SHIP assay for imaging receptor-mediated internalization and vesicle traffic using different microscopy modalities, including live imaging and super-resolution microscopy.

Content-aware image restoration (CARE).
For 4D (xyzt) imaging, movies were recorded using a SDCM and a Photometrics Evolve camera as described before. After imaging, samples were fixed with 4% PFA, and pairs of low signal-to-noise ratio (20 ms exposure time, as used for 4D live imaging) and high signal-to-noise ratio (500 ms) images were acquired. These images were used for training as input and target images, respectively. The CARE 2D model was trained from scratch for 100 epochs on 700 paired image patches (image dimensions: (100,100), patch size: (16,16)) with a batch size of 16 and a Laplace loss function, using the CARE 2D ZeroCost-DL4Mic notebook (v1.12) 26,27 . Key python packages used include TensorFlow (v1.15), Keras (v2.3.1), CSBDeep (v0.6.2), NumPy (v1.19.5), CUDA (v11). The training was accelerated using a Tesla T4 GPU. Default Advanced Parameters were enabled, and no augmentation was used for training. The model was validated using unseen images, obtaining a structural similarity index measure (mSSIM) of 0.723.
Image analysis. Deconvolved images (fixed samples, SDCM) were used for colocalisation analysis. Colocalisation analyses based on Mander's overlap coefficient 28 were performed in Fiji 29 using the Colocalisation Threshold tool. As negative control to assess random colocalisation, one of the channels was rotated 90° left, and Mander's coefficients were measured again. Mean fluorescence intensity (MFI) was measured from the raw images (sum intensity projection; 100 × 100 circular ROI) using ImageJ. MFI of the background was measured in 4 random positions per image and subtracted from the MFI of the cells. Particle tracking was done using Trackmate in Fiji 30 . To compare the resolution of the different techniques, the full-width-at-half-maximum (FWHM) value was used to determine the diameter of the vesicles. The line profile tool was used in ImageJ 31 and the fluorescence intensity profiles were fitted to a Gaussian curve using Curve Fitter in Fiji or Solver in Excel. The FWHM of this curve was calculated for each profile to obtain the final measurements (5-10 vesicles per condition).

Statistical analysis and illustrations.
The mean values of each experiment (n = 3) were used to calculate statistical significances in GraphPad Prism 6 using a paired Student's t-test. Statistical values are denoted as: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Graphs were created in GraphPad Prism 6 or SuperPlotsOfData 32,33 and illustrations were created with BioRender. Figure  www.nature.com/scientificreports/ Results FIP-IgM does not perturb the normal internalisation and traffic of BCR. Intracellular vesicular pathways have been primarily studied in cultivated fibroblasts or cancer cells that adhere firmly to a substrate and form a large spread area covering 600-1600 µm 234 , depending on the cell type and conditions. However, the detection of intracellular vesicles becomes particularly challenging in small (< 100 µm 2 ) or round cells. When studying the internalisation of membrane receptors or extracellular cargo, precise detection is further hindered by the remaining membrane-bound signal. Lymphocytes are round, non-adherent cells with a small cytoplasmic space, as the nucleus occupies 70% of the cell volume, and a high receptor concentration on their surface (Fig. 1A) [7][8][9]35 . With the aim to find a solution for these obstacles, we set up the SHIP assay to image internalized antigen in B cells. The SHIP assay requires two different oligos: a fluorescent ssDNA internalisation probe, that is conjugated to the desired ligand or antibody (FIP; 5′ fluorophore-TCA GTT CAG GAC CCT CGG CT-N3 3′), and a complementary ssDNA-containing quenching probe (QP; 5′ AGC CGA GGG TCC TGA ACT GA-Quencher 3′) (Fig. 1B). Previous studies have shown efficient labelling of antibodies and nanoparticles with Cy5-FIP and quenching using a Black Hole Quencher (BHQ) 2-labelled quenching probe (QP-BHQ2).
Here, based on our aim to evaluate the power of the SHIP assay for microscopy, we chose Abberior® STAR 635P-FIP (Abberior-FIP) and ATTO 647N-FIP (ATTO-FIP) as FIP fluorophores. These fluorophores have photophysical properties superior to those of Cy5, including better brightness and photostability, and they are also well suited for super-resolution microscopy. As quencher, the dark quencher BHQ3 (λ 620-730; λ max 672) was selected over BHQ2 (λ 550-650; λ max 579) due to its better spectral overlap with Abberior® STAR 635P (λ max 651) and ATTO 647N (λ max 664). As our model system, we chose A20 D1.3 B cells expressing an IgM BCR. Anti-IgM antibodies, commonly employed as surrogate antigens to activate B cells of any specificity, were labelled with Abberior-FIP and ATTO-FIP and used to engage the BCRs and trigger receptor-mediated endocytosis and antigen processing. A good degree of labelling was achieved with both probes, as cells stained with the FIP-labelled anti-IgM antibodies (FIP-IgM) displayed a similar intensity shift as those labelled with commercial Alexa Fluor® (AF) 647 anti-mouse IgM (Fig. S1A).
Next, we investigated whether the SHIP assay could be used to study antigen traffic in B cells without interfering with normal BCR internalisation. Using flow cytometry, we analysed the IgM internalisation rate in cells activated with Abberior-or ATTO-FIP-IgM at different time points. As a control, cells activated with a commercially available biotinylated anti-IgM were used. We found no differences in BCR internalization induced by the FIP-IgM probes or the biotinylated anti-IgM (Fig. S1B-D). These results were consistent with the literature, as the Cy5-FIP (7 kDa) has been described not to interfere with large molecules such as antibodies 36 . We observed, though, a slight yet notable decrease in the intensity of the total FIP-IgM signal over time in the absence of QP (Fig. S1C-D; red histograms). We hypothesized that this decrease could be due to (1) sensitivity of the fluorophores to the low pH endosomal environment, (2) antibody-mediated receptor shedding, or (3) fast antibody off-rates. To test this, we compared the FIP-IgM antibodies to other commercial and in-house labelled anti-IgM antibodies over time. B cells were incubated with ATTO-FIP-IgM, Abberior-FIP-IgM, and the same polyclonal anti-IgM antibody batch conjugated in-house to ATTO 647, allowing direct comparison of the antibody binding. Additionally, commercial anti-IgM antibodies labelled with AF 633 (pH-insensitive) and FITC (pH-sensitive) were used as controls for pH sensitivity. The signal of the AF 633 and ATTO 647 antibodies remained unchanged over time, as expected for pH-insensitive dyes, but also demonstrating that the used anti-IgM antibodies, labelled in-house or commercially, did not induce receptor shedding and bound stably to the target (Fig. S1E). The results implied that the IgM-FIP probes possessed some intrinsic level of instability in B cells which could be linked to the acidity of the antigen processing endosomal compartments or other specific features of our model system. However, this finding was not expected to pose major problems for imaging applications. In addition, we also experimentally verified that the addition of the QP did not affect cell viability, unlike the acid wash treatment classically used to strip cell surface-bound molecules (Fig. S1F) and that the FIP-IgM probes, but not the FIP ssDNA oligo alone, triggered normal B cell activation, using CD19 and Syk phosphorylation as readout (Fig. S1G).
Next, to evaluate the performance of the SHIP assay in microscopy, we engaged the BCRs with ATTO-FIP-IgM, Abberior-FIP-IgM, or AF 488-labelled anti-IgM as control. B cells were activated for different time points (5, 10, and 30 min), followed or not by quenching with the QP. Samples were fixed and imaged with a spinning disk confocal microscope, and images were deconvolved (Fig. S2A). As expected, an extremely high degree of colocalisation, compatible with non-disturbed BCR internalisation and correct endosomal targeting, was found between the FIP-IgM and the AF 488 anti-IgM antibodies before treatment with the QP (Fig. 1C,E; Abberior-FIP-IgM data not shown). After the addition of the quenching probe, a dramatic decrease in the FIP-IgM intensity was detected, especially at early time points, consistent with the flow cytometry analysis (Fig. S2B-D). Since only a small percentage of the antigen was internalised at 5 min (< 10%; Fig. S1B), few vesicles were visible in the quenched samples at that time (Fig. 1D). The number of antigen vesicles increased over time, and after 30 min most of the FIP signal remained detectable after the addition of the quencher, in accordance with the majority of the antigen (around 60%; Fig. S1B) being internalised at that point (Fig. 1D,E). As in the flow cytometry experiments, we also observed a decay in the total signal at later time points before the addition of the quencher (Fig. S2C). Nevertheless, the weakening of the signal did not seem to interfere with the imaging, and we often even noticed brighter intracellular signal using FIP-IgM compared to AF 488 anti-IgM, especially notable later in live imaging (see Fig. 3B or Supplementary Movie 2).
The SHIP assay improves the analysis of internalised antigen at early time points. In conventional immunofluorescence studies, it is impossible to unambiguously distinguish the internalised BCR-antigen complexes from the surface-bound receptors. Inaccurate determination of the signal can skew, for instance, the colocalisation analysis with vesicle markers, lowering the colocalisation coefficients and potentially leading to www.nature.com/scientificreports/ www.nature.com/scientificreports/ misguided biological interpretations. Thus, we proceeded to test the performance of the SHIP assay for antigen/endosome colocalization analysis. Typically, two different stages can be observed when studying antigen internalisation. At early timepoints (5-20 min) vesicles appear disperse and at late timepoints (15-45 min) the vesicles start clustering in the nuclear groove. Here, we selected two different activation times, one for early activation (10 min) and one for late activation (30 min), and prepared immunofluorescence samples using ATTO-FIP-IgM and AF 488 anti-IgM as a control. After this time, samples were quenched, fixed, and stained for the late endosome (LE) marker Rab7 or the early endosome (EE) marker EEA1 to compare the colocalisation of the total antigen (AF 488 anti-IgM) and the internalised antigen (ATTO-FIP-IgM) with the vesicle markers. As observed in the images, internalised vesicles (FIP-IgM + /AF 488 + , white circles) were commonly found together with Rab7 or EEA1, unlike the non-internalised antigen (FIP-IgM − /AF 488 + , yellow circles) ( Fig. 2A,B). Moreover, in the later time point, the internalised FIP-IgM was found to colocalise with Rab7 + endosomes close to the perinuclear compartment, as previously reported 7,37,38 , indicating once more that the FIP-IgM probe did not perturb normal antigen trafficking. In order to measure the improvement provided by the SHIP assay, we compared the colocalization between the antigen (ATTO-FIP-IgM) and the vesicle markers in samples treated or not with the QP (Fig. 2C; Fig. S3). The conditions without QP resemble the conventional setup where the extracellular and internalized ligands are not distinguished. Colocalisation of antigen with EEA1 and Rab7, measured by Mander's overlap coefficients, showed a significant increase, from 25 to 45%, at early time points (10 min) when using the SHIP assay (+QP; Fig. 2C). On the other hand, when the quencher was added at later time points (30 min), no significant differences were observed in the colocalisation, as the majority of the antigen was already internalised at that time, and most of the vesicles had moved towards the perinuclear area. These results demonstrate that the SHIP system can provide a valuable benefit to the colocalisation analysis of cellular cargo, especially shortly after internalisation.

The SHIP assay allows for specific imaging of internalised antigen in living cells. To date, there
are no reports showing the use of the SHIP assay in living cells. Thus, we next evaluated the potential of the SHIP assay for live imaging. B cells were attached to a fibronectin-coated microscopy dish and engaged with ATTO-FIP-IgM or AF 488 anti-IgM as a control. The time points after receptor triggering/FIP addition when the quencher was added were defined as early (< 15 min; Fig. 3A and Supplementary Movie 1) or late (> 15 min; Fig. 3B and Supplementary Movie 2). Images were recorded for at least 30 s before the addition of the quencher without stopping the acquisition. Only a few seconds after QP addition, the extracellular FIP signal was dramatically quenched, while the control AF 488 signal remained unmodified. Notably, after quenching the non-internalised signal, the internalised vesicles became more visible and distinct, improving vesicle tracking ( Fig. 3C and Supplementary Movie 3).
In our previous work, we found a considerable overlap of the internalised antigen with both early and late endosomal markers, as well as with Lysotracker, a fluorescent dye that stains acidic compartments, suggesting that antigen is directed to specialised processing vesicles immediately after internalization 7 . However, challenged by the surface-derived antigen signal, we were able to clearly distinguish a few vesicle internalization events because of the overlapping surface-derived signal. Now, to tackle this issue, we utilized the SHIP to follow the internalized vesicles together with endosomal markers or LysoTracker. Consistently with our previous report, when activating the cells with ATTO-FIP-IgM, we found that some antigen vesicles colocalised with GFP-Rab5 (EE marker) and RFP-Rab7 (LE marker) (Fig. 4A,B; Fig. S4A; Supplementary Movie 4A-B and 5), and LysoTracker ( Fig. S4B and Supplementary Movie 6). After QP addition, the vesicles were clearly distinguishable and showed colocalisation with the vesicle markers, demonstrating that the internalised antigen was indeed often moving close to or together with acidic vesicles positive for early or late endosomal markers.
Next, we assessed the capabilities of the SHIP assay in 3D live imaging (i.e. 4D: XYZT). B cells were engaged with ATTO-FIP-IgM as above and imaged before and after the addition of the quencher. An exposure time of 500 ms was used, and every time point was composed of 49 slices, allowing an average imaging speed of 27 s per volume. After quenching the surface signal, the separation of the internalised vesicles from the plasma membrane became drastically better ( Fig. 5A; stack image, sum intensity projection). In order to push the speed limits of image acquisition, we turned to deep learning-based image restoration. Now, an exposure time of 20 ms was used, allowing an average imaging speed of 3-4 s per volume (49 slices). Images were then restored using content-aware image restoration (CARE 27 ) to obtain the final 3D movie (Fig. 5B and Supplementary Movie 7). The results showed that the ATTO-FIP-IgM signal was sufficiently stable over time and bright enough to perform 4D live imaging for at least up to 20 min. Longer acquisition times were not tested. In this way, we could image and analyse a considerably higher number of antigen vesicles than earlier possible.
The SHIP assay is compatible with super-resolution imaging. Due to the diffraction limit of light, traditional confocal microscopy cannot resolve structures below 200-250 nm. Since intracellular vesicles often fall close to or below this resolution limit, we tested whether the SHIP assay could be combined with superresolution imaging techniques. First, stimulated emission depletion (STED) microscopy was tested 39 . B cells were labelled with Abberior-FIP-IgM and AF 488 anti-IgM F(ab)' 2 , activated for 10 and 30 min, quenched, fixed, and imaged using an Abberior STED system. Conventional scanning confocal images of both channels were acquired, and in addition, the Abberior-FIP-IgM channel was also acquired in STED mode (Fig. 6A). The apparent diameter of the vesicles was determined by measuring the full width at half maximum intensity (FWHM) of the Gaussian curve of the FIP-IgM fluorescence. Indeed, STED imaging considerably reduced the apparent vesicle size (from 280-330 to 165-210 nm). As STED requires a relatively strong fluorescence signal, detecting the FIP signal was, however, challenging after quenching at early time points when only a limited amount of antigen had been internalised (Fig. S2C-D). Next, we tested two other high-resolution imaging systems that are often www.nature.com/scientificreports/ . Images were deconvoluted with Huygens and colocalisation was measured using thresholded Mander's overlap coefficient (tM1 = % of antigen colocalising with EEA1 or Rab7). Different colours represent independent experiments: the mean of each experiments is shown with large symbols, and small symbols show the measured cells. As a control for random colocalisation, one channel was rotated 90° and colocalisation was measured. **P < 0.01, paired t-test (n = 3). www.nature.com/scientificreports/ more easily accessible to researchers: an Airyscan confocal microscope 40,41 and super-resolution radial fluctuations (SRRF) on a spinning disk confocal microscope 24 . B cells were stained with ATTO-FIP-IgM, activated for 10 or 30 min, quenched, fixed, and imaged. An LSM880 microscope with Airyscan was used to image the cells in 3D (z-stack; Fig. 6B). For SRRF, 100 frames per image (1 slice) were acquired using a SDCM and then processed in Fiji (Fig. 6C). Airyscan images showed a modest decrease in the apparent vesicle size (from 380-400 to 330-370 nm), similar to those images acquired with a SDCM and deconvolved using Huygens (from 450 to 320 nm) (Fig. 6D, Fig. S2A). SRRF, on the other hand, showed a good improvement (from 450 to 260 nm), but this was limited to one optical section. Although Airyscan and SRRF are compatible with live-cell imaging, these methods require a bright signal to obtain high-quality data. Resulting from the small number of internalised vesicles, the low brightness of the samples at early time points hampered the acquisition of live super-resolution images (data not shown) in our model system. Nevertheless, we demonstrated that the SHIP assay is compatible with several super-resolution approaches using fixed samples. The resolution increase, especially in STED and SRRF, clearly improved the separation of the vesicles, also unveiling small vesicles located close to each other. Hence, the SHIP assay could be applied in future research to facilitate the study of internalised cargo and receptors in different cell types.

Discussion
In the last two decades, microscopy has become an increasingly powerful tool to study dynamic cellular events, increasing our understanding of the complexity of intracellular vesicle traffic. However, distinguishing between internalised and non-internalised ligands, especially in small cells such as lymphocytes, remains a major challenge even with modern high-resolution techniques. A tool for analysing internalisation, termed Specific Hybridization  www.nature.com/scientificreports/ www.nature.com/scientificreports/ Internalisation Probe (SHIP), uses a fluorescence internalisation probe (FIP) and a complementary quenching probe (QP) to unambiguously separate the internalised receptors or ligands from those on the plasma membrane 15 . While the assay has been used in a handful of studies to measure internalisation using flow cytometry, the potential of this system for advanced microscopy studies has remained entirely unexplored. Our study demonstrates the high potential of the SHIP assay for improving the imaging of receptor/ligand internalization and the associated vesicle compartments and shows that the SHIP assay can be adapted to multiple imaging modalities, including live imaging and super-resolution microscopy.
Here, we investigated the functionality of the SHIP system using B cells as a model system. We validated and employed ATTO/Abberior-FIP-labelled anti-IgM antibodies as a surrogate antigen to trigger BCR activation and antigen/BCR internalisation and trafficking ( Fig. 1; Fig S1). As expected, the addition of the QP efficiently quenched the extracellular signal enabling crisp detection of the internalised receptors (Fig. S1C-D and S2B). When evaluating the advantages of the SHIP system for colocalisation studies in immunofluorescence samples, we detected a significant increase in the colocalisation coefficients when analysing the internalised IgM (FIP-IgM + QP) together with vesicular markers soon after internalization as compared to the conventional labelling ( Fig. 2). Upon BCR engagement, fast vesicle sorting and trafficking events occur within minutes but the imaging of these internalised receptors is strongly hindered by the bright signal of the yet non-internalized receptors. A better understanding of these intricate vesicle trafficking events would answer critical open questions about the mechanisms enabling processing of the internalized antigen into the peptides for MHCII presentation and, ultimately, mounting the life-saving high-affinity antibody responses. Using SHIP, the dramatic improvement in the definition of internalized antigen/BCR enabled precise vesicle detection and tracking (Figs. 2 and 3). In the later time points, when most of the antigen was already internalized, the improvement offered by SHIP remained modest.
As live imaging is a critical tool to understand the vesicle dynamics and maturation events, we proceeded to investigate the performance of the SHIP system in live cells. A remarkable improvement was observed after adding the quencher at both early and late time points (Fig. 3), as eliminating the bright fluorescence signal from the plasma membrane unveiled numerous endosomes close to the plasma membrane. We also tested the system in cells co-transfected with GFP-Rab5 and RFP-Rab7 or loaded with LysoTracker to visualize antigen colocalisation with these vesicle markers ( Fig. 4 and Fig. S4). Using SHIP, we were able to unambiguously detect the vesicles where antigen colocalized with endosomal markers even when they were located at the close vicinity of the plasma membrane. www.nature.com/scientificreports/ www.nature.com/scientificreports/ Since vesicles are highly dynamic structures that travel in three dimensions, live imaging in 2D can provide only a limited amount of information. The widely available spinning disk confocal microscopes equipped with a sensitive camera, such as EMCCD, can be used for 3D live imaging of rapid cellular processes, as opposed to classical confocal microscopes 42 . Using a SDCM, 3D volumes of B cells were acquired with a temporal resolution of 27 s without compromising signal quality (Fig. 5A). This speed, however, does not guarantee the capture of fast endosomal vesicle movements. Thus, in an attempt to gain higher temporal and spatial resolution, we combined the SDCM acquisition with an image restoration approach. In the last decade, machine learning and deep learning approaches have become popular tools for image analysis and image restoration. Here, contentaware image restoration (CARE), a deep learning-based image restoration algorithm 27 , was used to denoise 4D data sets. To train this network, pairs of images with a low and a high signal-to-noise ratio (SNR) were acquired after the fixation of the structures with PFA. We were able to improve the temporal resolution to 3.7 s per volume with this approach (Fig. 5B), which is comparable to the timeframes of 2.5-4.4 s achieved using lattice light-sheet microscopy, the fastest commercially available modality for 3D live imaging, to study endosomal trafficking in T lymphocytes [43][44][45] .
Super-resolution imaging has become increasingly popular in cell biology in the last two decades 46 . Superresolution techniques include stimulated emission depletion (STED), single-molecule localisation microscopy (e.g. PALM and (d)STORM), structure illumination microscopy (SIM), expansion microscopy, pixel reassignment super-resolution techniques such as Zeiss Airyscan, and fluctuation-based super-resolution techniques, such as SRRF. These techniques have different strengths and weaknesses that need to be considered depending on the sample type, application, desired resolution, and, more importantly, availability to the researchers. We first tested the SHIP system using STED (Fig. 6A), as a theoretical resolution of 20-60 nm can be achieved with this system. The average size of the antigen-containing endosomes in B cells, measured in electron microscopy (EM) images, is around 100-200 nm for unilamellar vesicles and 200-400 nm for multilamellar vesicles 7 . Using SHIP in conjunction with STED imaging, we achieved an apparent vesicle size of 160 nm in good agreement with the previous EM data. Although traditional super-resolution modalities such as STED and single-molecule localisation microscopy provide the best resolution, these techniques require special microscopes and knowledge, and they are not yet widely applicable to 3D and live imaging. Hence, we next tested Airyscan, a Zeiss confocal microscope system well-suited for live imaging that has gained notable popularity in recent years, and SRRF, an algorithm compatible with most microscopes without hardware modifications. While Airyscan provided only a modest improvement in our setup (Fig. 6B,D), it allowed 3D imaging and improved the signal-to-noise ratio. SRRF, on the other hand, notably improved the imaging, but this algorithm can only be applied in 2D (Fig. 6C,D). Although both SRFF and Airyscan are compatible with live imaging, a bright signal is required to obtain high-quality images. In our study, following the published protocols 16 , we used a 2-fold molar excess of FIP to label the antibodies. Although a good brightness level was achieved, as measured by flow cytometry and microscopy ( Fig. S1A and Fig. S2B-C), the limited amount of internalised antigen at early time points also limited the FIP-IgM signal remaining in the cells after quenching (Fig. 1B). As a result, we did not achieve good super-resolution live imaging data (data not shown), indicating that a higher fluorophore:antibody ratio should be used for such challenging live imaging super-resolution approaches. According to the literature, an optimal fluorophore:antibody ratio can be found between 2 and 8 47 , and therefore, it would be interesting to test in the future if a higher degree-of-labelling would further improve the imaging.
At the moment, the SHIP assay has only been used to track the internalisation of one ligand at a time. Therefore, in the future, it will be interesting to investigate the possibility to image two different ligands simultaneously using two pairs of complementary probes, such as ATTO647N-FIP/QP-BHQ3 and ATTO488-FIP/QP-BHQ1. A disadvantage of the SHIP system is, however, the high cost associated with the synthesis of the labelled oligos ($700-900 for every FIP and $200-300 for the QP). In addition, the FIP labelling of each ligand/antibody needs to be optimised. In order to omit the labelling step and utilize the same FIP for detection of different subjects, fluorescent FIP oligos could also be conjugated with biotin and attached to different streptavidin-labelled ligands or antibodies 18 .
In conclusion, this study explored for the first time the capabilities of the SHIP system for microscopy. SHIP can significantly improve the analysis of the internalized receptor/ligand, especially at locations close to the plasma membrane soon after internalization when the accurate detection of the vesicles is hampered by the membrane signal. We also found SHIP fully applicable for live and super-resolution imaging of internalised ligands and receptors. Together with its already described potential for high-throughput internalisation and recycling assays using flow cytometry, these properties confer the SHIP system broad applicability in different research fields, such as cell biology, immunology, and cancer biology.