Correlative light and electron microscopy reveals discrepancy … · 2017. 6. 2. · BM-DCs were...

Journal of Microscopy, Vol. 00, Issue 0 2017, pp. 1–9 doi: 10.1111/jmi.12567

Received 23 October 2016; accepted 27 March 2017

Correlative light and electron microscopy reveals discrepancybetween gold and fluorescence labelling

D . M . V A N E L S L A N D ∗,†, E . B O S‡, J . B . P A W L A K ∗,†, H . S . O V E R K L E E F T ∗,†, A . J . K O S T E R‡ & S . I . V A NK A S T E R E N ∗,†∗Division of Bio-organic Synthesis, Leiden Institute of Chemistry, Gorlaeus Laboratories, Leiden University, Leiden, The Netherlands

†Institute for Chemical Immunology, Gorlaeus Laboratories, Leiden University, Leiden, The Netherlands

‡Department of Molecular Cell Biology, Section Electron Microscopy, Leiden University Medical Center, Leiden, The Netherlands

Key words. Bioorthogonal labelling, cathepsins, CLEM, electron microscopy,fluorescence microscopy, immunogold, labelling, LAMP-1, thawedcryosections.

Summary

Electron microscopy (EM) is traditionally employed as a follow-up to fluorescence microscopy (FM) to resolve the cellular ul-trastructures wherein fluorescently labelled biomolecules re-side. In order to translate the information derived from FMstudies to EM analysis, biomolecules of interest must be iden-tified in a manner compatible with EM. Although fluorescentsignals can serve this purpose when FM is combined with EMin correlative light and electron microscopy (CLEM), the tradi-tional immunogold labelling remains commonly used in thiscontext. In order to investigate how much these two strategiesrelate, we have directly compared the subcellular localizationof on-section fluorescence labelling with on-section immuno-gold labelling. In addition to antibody labelling of LAMP-1,bioorthogonal click labelling was used to localize soluble cys-teine cathepsins or membrane-associated sialylated glycans.We reveal and characterize the existence of inherent discrep-ancies between the fluorescence signal and the distribution ofgold particles in particular in the case of membrane-associatedantigens.

Introduction

Fluorescence microscopy (FM) is a powerful imaging tool thatcan be employed to track, localize and monitor biomoleculeswithin cellular systems. However, with fluorescence mea-surements it is only the position of the labelled biomoleculesthat can be studied in relation to other fluorescently markedbiomolecules. The morphology of subcellular structureswherein they reside remains invisible. Conversely, electronmicroscopy (EM) enables observations at nanometre-scale res-

Correspondence to: Abraham Koster, Department of Molecular Cell Biology, Section

Electron Microscopy, Leiden University Medical Center, Leiden, The Netherlands.

Tel: +31-71-5269294; e-mail: [email protected]

olution and is therefore often employed to gain ultrastructuralcharacterization of the subcellular environment pertinent tothe biomolecules initially investigated with FM (Predescu et al.,2003; Robenek et al., 2006; Sarnataro et al., 2009; Choi et al.,2015; Boonen et al., 2016).

The location of biomolecules within EM imaged specimencan be visualized upon gold labelling. Colloidal gold is electrondense and, due to its punctate and precise labelling pattern, itis favourable for the localization of biomolecules within struc-tures studied at nanometre scale (Griffiths, 1993; Mayhew& Lucocq, 2008). Gold-labelling can be broadly applied, e.g.in combination with immunocytochemistry (through directantibody coupling or in combination with protein A), butalso with lectins and other binding proteins (Faulk & Taylor,1971; Griffiths, 1993). This allows recognition of fluorescentreporter proteins (GFP), fluorophores or direct epitopes studiedwith FM (Sirerol-Piquer et al., 2012). The preferred method forpreparing biological material for gold labelling is the Tokuyasucryosectioning technique. This technique is the only postem-bedding on-section immunolabelling approach that does notrequire dehydration by polar solvents before the application ofaffinity markers. It is therefore that thawed cryosections tendto give the highest accessibility of antigens to the antibodiesand are the best for most purpose on section labelling strategies(Tokuyasu, 1973; Slot & Geuze, 2007; Webster et al., 2008).

Besides colloidal gold markers, directly visible in the electronmicroscope, one can also use correlative light and electronmicroscopy (CLEM) to directly visualize fluorescent markerson EM imaged specimen. With CLEM, specific biomoleculesand cellular structures can be identified at low magnifica-tion with FM, followed by ultrastructural assessment of sub-cellular location and context with EM (Giepmans, 2008; deBoer et al., 2015). CLEM enables the detection of rare cellularevents and allows interpretation of fluorescent labelling on lowmagnification/ large overview images—in contrast to colloidal

C© 2017 The AuthorsJournal of Microscopy C© 2017 Royal Microscopical Society

2 D . M . V A N E L S L A N D E T A L .

gold markers, which are too small to detect at low magnifica-tion. A downside of fluorescence-based CLEM is that correla-tion of images requires enlargement of the FM image to coverthe high magnification EM image. This enlargement resultsin diffuse fluorescent signals with a diameter of a few hun-dred nanometres that cannot be related to the exact locationof proteins of interest, which can be around 4 nm in diame-ter and may be associated with organelles as small as 30 nm(Watanabe et al., 2011).

We envisaged that the centre of the fluorescent signalscould be found by combining gold labelling and fluorescence-based CLEM in a single experiment. Surprisingly, this combi-nation showed for the lysosomal membrane protein LAMP-1a discrepancy between the distribution of gold and fluo-rescent label. To understand the scope and implications ofthis phenomenon, we devised a series of experiments in-volving bioorthogonal labelling of cryosections (van Elslandet al., 2016). Bioorthogonal labelling is a chemical labellingstrategy whereby firstly a small, physiologically inert chem-ical group is incorporated into a biomolecule of interest,which is subsequently reacted with a detection group usinga highly selective chemical reaction. With this approach awide range of biomolecules can be labelled, with high speci-ficity and minimal perturbations (Grammel & Hang, 2013;van Elsland et al., 2015). Upon the use of bioorthogonal la-belling, we were able to compare fluorescence and immuno-gold labelling of model membrane-associated or soluble epi-topes using the same labelling procedures. We discoveredthat in particular membrane-associated antigens are stronglylabelled with gold particles, whereas the signal of fluores-cence labelling of the same target molecules is weak or notdetectable.

Materials and methods

Specimen preparation

Mouse BM-DCs (bone marrow dendritic cells) and Jurkat cellswere cultured as described elsewhere (van Elsland et al., 2016).BM-DCs were incubated for 2 h with DCG-04-azide (Hoogen-doorn et al., 2014) (final concentration of 10 µM) after whichthe cells were washed with PBS (phosphate buffered saline)and kept for 2 h in fresh medium. Jurkat cells were incubatedfor 3 days with 50 µM of N-azidoacetylmannosamine (Man-nAz: Invitrogen, Cat. 88904) from stock solutions in DMSO(dimethylsulfoxide).

Samples were prepared for cryosectioning as described else-where (Peters et al., 2006; van Elsland et al., 2016). Briefly,BM-DCs and Jurkat cells were fixed for 24 h in freshly pre-pared 2% PFA in 0.1 M phosphate buffer. Fixed cells wereembedded in 12% gelatin (type A, bloom 300, Sigma) and cutwith a razor blade into 0.5 mm3 cubes. The sample blockswere infiltrated in phosphate buffer containing 2.3 M sucrosefor 3 h. Sucrose-infiltrated sample blocks were mounted on

aluminium pins and plunged in liquid nitrogen. The frozensamples were stored under liquid nitrogen.

Ultrathin cell sections of 75 nm were obtained essentiallyas described elsewhere (van Elsland et al., 2016). Briefly, thefrozen sample was mounted in a cryo-ultramicrotome (Leica).The sample was trimmed to yield a squared block with a frontface of about 300 × 250 µm (Diatome trimming tool). Usinga diamond knife (Diatome) and antistatic devise (Leica), a rib-bon of 75 nm thick sections was produced that was retrievedfrom the cryo-chamber with the lift-up hinge method (Boset al., 2011). A droplet of 1.15 M sucrose was used for sectionretrieval.

Obtained sections were transferred to a specimen grid pre-viously coated with formvar and carbon. Grids were addition-ally coated with 100 nm FluoroSpheres carboxylate-modified(350/440) (Life Technologies, Cat. F8797).

On section labelling

Sections that were click-labelled with AlexaFluor-488/TAMRA Alkyne and immunogold-labelled were labelled asfollows; thawed cryosections on an EM grid were left for30 min on the surface of 2% gelatin in phosphate bufferat 37°C. Subsequently, grids were incubated on drops ofPBS/glycine and PBS/glycine containing 1% BSA. Gridswere then incubated on top of the ccHc- cocktail (0.1 MHEPES, pH 7.3, 1 mM CuSO4, 10 mM sodium ascor-bate, 1 mM THPTA ligand, 10 mM amino-guanidine,5 µM AlexaFluor-488/TAMRA Alkyne (Invitrogen, Cat.A10267/Cat. T10183, respectively)) for 1 h and washedsix times with PBS. Sections were then blocked again withPBS/glycine containing 1% BSA after which the grids wereincubated for 1 h with 0.1 M PBS/glycine, 1% BSA (block-ing solution) supplemented with a rabbit IgG anti-Alexa488 antibody (Invitrogen, Cat. A11094). After washing withPBS/glycine and blocking with PBS/glycine 0.1% BSA, gridswere incubated for 20 min on PBS/glycine 1% BSA supple-mented with protein A coated 10 or 15 nm gold particles(CMC, Utrecht University). Grids were then washed with PBS,labelled with DAPI (1:5000 in PBS for 5 min), and addition-ally washed with PBS and aquadest. In case of protein A Alex-aFluor labelling, protein A coated gold particles were replacedfor protein A Alexa 488 or protein A Alexa 647 (20 µg/mL,Invitrogen, Cat. P11047/Cat. P21462, respectively).

In case of immunofluorescence and immunogold labellingagainst LAMP-1, thawed cryosections on an EM grid wereleft for 30 min on the surface of 2% gelatin in phosphatebuffer at 37°C. Grids were washed five times with PBS/glycineand blocked with PBS/glycine containing 1% BSA after whichthe grids were incubated for 1 h with PBS/glycine 1% BSAsupplemented with a rat anti-mouse LAMP-1 AlexaFluor647 (Biolegend, Cat. 121609). Grids were then washed fivetimes with PBS/glycine and blocked again with PBS/glycinecontaining 1% BSA after which the grids were incubated for

C© 2017 The AuthorsJournal of Microscopy C© 2017 Royal Microscopical Society, 00, 1–9

C L E M R E V E A L S D I S C R E P A N C Y B E T W E E N G O L D A N D F L U O R E S C E N C E L A B E L L I N G 3

1 h with PBS/glycine 1% BSA supplemented with a rabbit IgGanti-rat (Nordic Immunology, Cat. RARa/IgG(H+L)/7S). Af-ter washing with PBS/glycine and blocking with PBS/glycine0.1% BSA, grids were incubated for 20 min on PBS/glycine1% BSA supplemented with protein A coated 10/15 nm goldparticles (CMC, Utrecht University). Grids were then washedwith PBS, labelled with DAPI (1:5000 in PBS for 5 min), andadditionally washed with PBS and aquadest. In case of proteinA Alexa labelling, protein A coated 10/15 nm gold particleswere replaced for protein A Alexa 488 or protein A Alexa 647(20 µ g/mL).

In case click labelling and LAMP-1 labelling were performedon the same section, the click labelling was preformed prior tothe LAMP-1 labelling steps.

Microscopy and correlation

The CLEM approach used was adapted from Vicidomini et al.and has been described elsewhere (van Elsland et al., 2016).Briefly, grids containing the sample sections were washed with50% glycerol and placed on glass slides (precleaned with 100%ethanol). Grids were then covered with a small drop of 50%glycerol after which a cover slip was mounted over the grid.Cover slips were fixed using Scotch Pressure Sensitive Tape.Samples were imaged with a Leica TCS SP8 confocal micro-scope (63× oil lens, NA = 1.4). After confocal microscopy, theEM grid with the sections was removed from the glass slide,rinsed in distilled water and incubated for 5 min on droplets ofan aqueous solution containing 2% methylcellulose and 6%uranyl acetate. Excess of methylcellulose/uranyl solution wasblotted away and grids were air-dried. EM imaging was per-formed with an Tecnai 20 transmission electron microscope(FEI) operated at 120 kV acceleration voltage.

Correlation of confocal and EM images was performed inAdobe Photoshop CS6. In Adobe Photoshop, the FM imagewas copied as a layer into the EM image and made 50% trans-parent. Transformation of the FM image was necessary tomatch it to the larger scale of the EM image. This was per-formed via isotropic scaling and rotation using interpolationsettings; bicubic smoother. Alignment at low magnificationwas carried out with the aid of nuclear DAPI staining in com-bination with the shape of the fluorescently labelled cells. Athigh magnification alignment was performed using the fidu-cial beads (Kuipers et al., 2015).

Colocalization analysis and quantification of label distributions

Upon the morphological appearance of a visible membrane,profile organelles were depicted either positive for gold or forfluorescence. For the determination of structures positive forgold, only structures positive for two or more particles weredepicted.

Colocalization of gold and fluorescence was calculated asa percentage upon quantification of the overlap of gold andfluorescence using the count tool in Photoshop. With this

strategy gold particles were counted that either colocalizedwith fluorescence or did not colocalize with fluorescence.

The Pearson’s coefficient was determined on magnified con-focal images of protein A double labelled samples using Coloc2function in ImageJ after background was corrected to elim-inate nonspecificity. These images were obtained in AdobePhotoshop CS6 upon CLEM correlation.

Results

Defining the problem

In order to determine the position of fluorophore-tagged anti-gens within cellular ultrastructures, we applied a previouslydescribed CLEM strategy for detection and localization of fluo-rescent tags (van Elsland et al., 2016) and labelled fluorophore-antigen complexes with immunogold particles. This strategyinvolves cryosectioning according to the Tokuyasu techniqueand thus allows fluorescence-based CLEM and immunogoldlabelling on the same specimen sections. Labelling strategiesemployed in this study are schematically displayed in all thefigures and the corresponding symbol legends are shown inTable 1. The fluorescence-gold labelling strategies employedin this study all rely on indirect fluorophore to gold conjuga-tion upon use of antibodies. This indirect fluorophore to goldconjugation has been shown to minimize the possible quench-ing of the fluorescence signals and is therefore the method ofchoice to combine fluorescence-based CLEM and immunogoldlabelling (Kandela & Albrecht, 2007).

Table 1. Symbols legends labelling strategies.

Component Symbol

Protein A conjugated gold

Protein A conjugated AlexaFluor

IgG antibody

AlexaFluor conjugated IgG antibody

Antigen

Azide modified antigen

AlexaFluor alkyne



We chose to apply this strategy to the lysosomal mem-brane glycoprotein LAMP-1, a major protein component ofthe lysosomal membrane (Eskelinen, 2006). LAMP-1 is elabo-rately used as a marker for the identification of late endosomesand lysosomes with FM and EM (van Meel & Klumperman,2008; Appelqvist et al., 2013). BM-DCs were cultured, pro-cessed for cryosectioning (Peters, 2006), and sectioned into75 nm thick sections, followed by primary labelling with a flu-orescent (AlexaFluor 647) rat anti-mouse LAMP-1 antibodyand secondary labelling with 15 nm protein A-coated goldparticles after a rabbit anti-rat IgG binding step (Fig. 1A, Strat-egy). Sections were then imaged with a confocal microscope,after which sections were stained with uranyl acetate and im-

aged using an electron microscope. The FM and EM imageswere superimposed (Fig. 1A, CLEM), and the correspondinglabel distributions were analysed (Fig. 1A, label distribution).To facilitate interpretation of the gold labelling in relationto the fluorescent signals, all gold particles are highlightedwith green dots. Upon the morphological appearance of a vis-ible membrane profile, several organelles were depicted whitewhen fluorescent label is associated, and yellow when goldlabel is associated. We found that the distribution of LAMP-1gold only partially overlapped with the distribution of LAMP-1 fluorescence, although LAMP-1 fluorescence was alwaysassociated with gold label (Fig. 1A, Label distribution). Notethat gold can be localized on fluorescent label, but not on

Fig. 1. Labelling of LAMP-1 with gold results in different label patterns compared to the labelling of LAMP-1 with fluorescence. (A) BM-DC sections werelabelled with a fluorescent AlexaFluor 647 (red) rat anti-mouse LAMP-1 antibody, and with protein A coated gold particles. Gold particles labelling wasperformed using a rabbit anti-rat binding step (Strategy). Representative CLEM image (CLEM). Upon the morphological appearance of a visible membraneprofile several organelles were depicted white when fluorescent label is associated, and yellow when gold label is associated (Label distribution). (B) BM-DCswere incubated with DCG-04-azide and were labelled with AlexaFluor 488 alkyne (green) and with protein A coated gold directed against AlexaFluor488 using a rabbit anti-AlexaFluor 448 binding step (Strategy). Representative CLEM image (CLEM). Upon the morphological appearance of a visiblemembrane profile, several organelles were depicted white when fluorescent label is associated, and yellow when gold label is associated (Label distribution).Scale bars, 500 nm.



organelle structures. Moreover, areas with high fluorescencewere not highly gold labelled. Quantification of overlap re-vealed that only 66% of the gold particles colocalized withfluorescence.

We deemed the above results remarkable, considering thatthe gold label was directed towards the fluorescent LAMP-1antibody. To explore whether such label discrepancy repre-sents a general phenomenon associated with combining flu-orescence and gold techniques, we next tested the CLEM andimmunogold approach for a nonmembrane-bound molecule,namely, cathepsins–papain like cysteine proteases known toreside within lysosomes. In order to label the cathepsins withfluorescence and immunogold, BM-DCs were incubated with aDCG-04 probe, which binds to active cathepsins and contains abioorthogonal azide-group (Greenbaum et al., 2000; Punger-car et al., 2009). After cryosectioning of BM-DCs, the DCG-04-azide was labelled with an AlexaFluor 488 fluorophore usingan on-section copper catalysed Huisgen cycloaddition (ccHc)reaction (van Elsland et al., 2016). This was followed by an im-munogold labelling step directed against the AlexaFluor 488(Fig. 1B, Strategy). The distribution of cathepsin gold label cor-responded well with the distribution of the cathepsin fluores-cence label (Fig. 1B, Label distribution). Quantification showedthat 87% of the gold particles colocalized with the fluorescencesignal. These results suggest that the degree of correspondencebetween gold and fluorescence label distribution may varywith the location of the antigen (i.e. membrane-associated vs.soluble).

Analysing the problem

In contrast to LAMP-1, the distribution of gold label corre-sponded well with the distribution of the fluorophore label forcathepsins. Since LAMP-1 and cathepsins are associated withthe same organelles, we anticipated to see predominantly over-lap of the LAMP-1 and cathepsin label distributions (Bromme,2001). To test this, we defined LAMP-1 associated organellesas being positive for cathepsin label and analysed the distri-butions of LAMP-1 gold (Fig. 2A, Strategy) and LAMP-1 flu-orescence (Fig. 2B, Strategy) in relation to these cathepsinpositive organelles. As expected, association of LAMP-1 withcathepsin-positive organelles was confirmed with immuno-gold labelling (Fig. 2A, CLEM and label distribution). How-ever, immunofluorescence labelling of LAMP-1 showed thatLAMP-1 also associated with cathepsin-negative organelles,and that cathepsin positive vesicles are not always associatedwith LAMP-1 (Fig. 2B). To exclude the possibility that thesecondary antibody in the gold label strategy represented inFigure 2(A) would influence the labelling, we substituted theprotein A gold for protein A AlexaFluor 647 (Fig. 2C, strat-egy) and compared the result of this strategy (Fig. 2C, labeldistribution) with the results shown in Figure 2(B). Statisti-cal colocalization analysis (Fig. 2C, colocalization) showed thatsubstitution of protein A gold by protein A AlexaFluor 647 re-

sulted in a label distribution similar to the distribution shownin Figure 2(B). Collectively, these results indicate that the ap-plication of gold particles within a label strategy influences theimmuno detection of antigens on sections.

Testing the hypothesis

The results presented so far show that the distribution of im-munogold label for LAMP-1 does not completely correlate withimmunofluorescence labelling of the same antigen. As thiswas prominently observed for the LAMP-1 epitope, but notfor the cathepsin epitope, we postulated that there could bean effect of the immediate surrounding of the epitope on theability of the gold conjugates to bind to cognate epitopes. SinceLAMP-1 is a membrane-bound molecule whereas cathepsin issoluble, we hypothesized that immunogold labels membrane-bound molecules differently from soluble epitopes. We thusset out to install the same bioorthogonal group used to labelcathepsins via the DCG-04 probe on a membrane-associatedepitope, creating the same covalent fluorophore introductionin a membrane-associated context. This was achieved by in-stalling the bioorthogonal group on sialylated glycans (Saxon& Bertozzi, 2000), as these constitute parts of membrane-bound molecules.

Jurkat cells were incubated with Ac4-N-azidoacetylmannosamine (Ac4-ManNAz), to add a bioorthogonal function-ality to sialylated glycans under conditions reported previ-ously (van Elsland et al., 2016). Cryosections from these cellswere subjected to the same labelling strategy as used for thecathepsins above. To eliminate the possibility that a differentcell type and/or rearrangement of the bioorthogonal groupwould influence either the binding of AlexaFluor 488 alkyneor recognition of AlexaFluor 488 by anti AlexaFluor 488 IgG,we first double-labelled the modified epitopes with AlexaFluoralkyne and protein A AlexaFluor (Figs. 3A, B, strategy). Similaroverlap of the two fluorophores was observed for both epitopes(Figs. 3A, B, label distribution) with no significant difference be-tween the two azide-modified epitopes (Fig. 3B, colocalization),meaning that the affinity of the two labels were not alteredupon the epitope or cell type change. Next, we double-labelledthe sialylated glycans and substituted protein A AlexaFluorfor protein A gold (Fig. 3C, strategy). We found that the goldlabel localized primarily on the plasma membrane, whereasthe fluorescent signal predominantly labelled intracellularstructures (Fig. 3C, CLEM). Quantification of colocalization(Fig. 3C, label distribution) revealed that again only 64% ofthe gold label colocalized with the fluorescent foci, lendingcredence to the hypothesis that the cellular location of anepitope influences the degree of gold labelling. For solubleepitopes the degree of gold label is relatively low comparedto the fluorescent signal, whereas the membrane-bound epi-topes get highly labelled with gold, even when the fluo-rescence label is low or not detectable at all with confocalmicroscopy.



Fig. 2. Labelling with gold particles shows a different antigen distribution than with fluorescent label. (A) BM-DCs were incubated with DCG-04-azideand were labelled with AlexaFluor 488 alkyne (green). Sections were subsequently labelled with anti-LAMP-1 and protein A coated gold directedagainst the LAMP-1 antibody using a rabbit anti-rat binding step (Strategy). CLEM images obtained (CLEM) were analysed on the label distribution andgold distribution was marked with red dots. Structures that were positive for LAMP-1 gold were depicted white and structures positive for cathepsinfluorescence were depicted yellow. (B) BM-DCs were incubated with DCG-04-azide and were labelled on section with AlexaFluor 488 alkyne (green).Sections were subsequently labelled with fluorescent AlexaFluor 647 anti-LAMP-1 (red) (Strategy). CLEM images obtained (CLEM) were analysed on thelabel distribution. Structures positive for LAMP-1 fluorescence were depicted with white and structures positive for cathepsin fluorescence with yellow.(C) BM-DCs were incubated for with DCG-04-azide and were labelled on section with TAMRA alkyne (green). Sections were subsequently labelled withanti-LAMP-1 and protein A AlexaFluor 488 fluorophore (red) directed against the LAMP-1 antibody using a rabbit anti rat binding step (strategy). Arepresentative confocal image of this sample is shown in Fluorescence. Colocalization analysis of the labelling sequences in (B) and (C) (Colocalization).Scale bars, 500 nm.



Fig. 3. The cellular location of an epitope influences gold labelling, resulting in a discrepancy between fluorescence and gold label distribution. (A) Jurkatcells were incubated for with N-azidoacetylmannosamine and were labelled on section with AlexaFluor 488 alkyne (green). Sections were subsequentlylabelled with a AlexaFluor 647 protein A fluorophore directed against the AlexaFluor488 using a rabbit anti-rat binding step (strategy). (B) BM-DCsincubated with DCG-04-azide were labelled on section with AlexaFluor 488 alkyne (green) and protein A AlexaFluor 647 (red) directed toward AlexaFluor488 using a rabbit anti-AlexaFluor 488 binding step. Colocalization analysis of sialylated glycans and protein A compared to cathepsins and proteinA (Colocalization). (C) Jurkat cells were incubated N-azidoacetylmannosamine and were labelled on section with AlexaFluor 488 alkyne (green). Goldparticles labelling was performed using a rabbit anti-AlexaFluor 448 binding step (Strategy). Representative CLEM image (CLEM). Organelles positive forsialylated glycan-fluorescence were depicted with yellow and organelles positive for sialylated glycan-gold were depicted with white. The distribution ofthese spheres was displayed in the context of the label distribution (Label distribution). Scale bar, 500 nm.



Discussion

In order to translate information of FM studies to EM analysis,biomolecules initially observed with the former need to belabelled in a manner compatible with EM. Traditionally, thisis performed using immunogold labelling of fluorescentlydetected biomolecules. Here we identify and explore the dis-crepancies arising between distributions of fluorescence andgold labelling. We found that a subset of cathepsin-positiveorganelles is negative for LAMP-1 when labelling is doneby fluorescence, but positive for LAMP-1 when labelling isdone with gold particles. Variations in label distributionscan be explained by marker heterogeneity in phagosomal,endosomal and lysosomal structures (Henry et al., 2004;Luzio et al., 2007; Huotari & Helenius, 2011). However,such heterogeneity cannot explain our observation that thelabel distribution pattern changes upon changing the type oflabel (gold vs. fluorescent label). We reason that the antigendetection level could change upon changing the type of label,but that the label distribution pattern should not.

In a comparison of fluorescent and gold label it can beargued that the localization of diffuse enlarged fluorescentsignals may result in a less distinct mapping of positivestructures, and that fluorescence can be difficult to detectwhen label density is weak, which is especially the casefor fluorescence labelling on non-dense structures such asmembranes. However, since the fluorescence intensity isrelated to the concentration of the epitope, there should still bean overlap of fluorescence and gold label (Tohda et al., 2001),and this is not expressed by the discrepancy we observe.

Also, gold label may show different antigen detection levels.Variations in labelling of conjugated gold particles have beenpreviously reported by Griffiths and Hoppeler, describing vari-ations in immunogold labelling efficiencies for the ER, Golgiapparatus and viral membranes (Griffiths & Hoppeler, 1986).In their study the ER membranes were found to exhibit thehighest labelling efficiency compared to the Golgi apparatusor viral membranes. These observations are supported by oth-ers noting profound variations in immunogold labelling effi-ciency between different cellular compartments and structures(Posthuma et al., 1984; Tooze et al., 1987; Lucocq, 1994). It isgenerally considered that cross-linking of target molecules byfixatives, as well as steric hindrance and valency of both targetand ligand molecules constitute factors that influence labellingefficiency (White et al., 1999). Other investigators report thatthe density of cellular material affects the labelling efficiencyas well. In dense secretory granules, there is almost only sur-face labelling whereas the endoplasmic reticulum gets moreprominently labelled (Catherine Rabouille, 1999). We postu-late that these epitope availability effects more strongly affectgold labelling compared to fluorescence labelling, resulting inthe discrepancies observed between these labelling strategies.

If epitopes of interest are found in various structures withinthe same specimen, gold labelling may result in a biased

detection of target molecules. As a consequence, it may be ar-gued that in this context fluorescence labelling is preferable togold labelling. However, to reach the punctuality of the goldparticles, the detection of these fluorescence labels needs tobe at a higher resolution. Super-resolution microscopy tech-niques, such as STORM and PALM, may serve this purposeas these techniques allow to precisely localize molecules ata resolution of a few nanometres. Moreover, examples oftheir applications to CLEM have yet been explored (Watan-abe et al., 2011; de Souza, 2015). Another approach resultingin high resolution labelling, includes labels capable of poly-merizing diaminobenzidine (DAB). It has been shown thatphoto-oxidation-based polymerization of DAB occurs withina few nanometres away of the epitope (Sosinsky et al., 2003;Cortese et al., 2009; Gaietta et al., 2011; Ngo et al., 2016).Application of these high resolution methods could pave theway to provide alternative labels for accurate mapping in cor-relative FM an EM studies.

Acknowledgement

We would like to thank Prof. Gareth Griffiths, Dr. Urska Repnikand Dr. I Berlin for critical reading of the manuscript.

References

Appelqvist, H., Waster, P., Kagedal, K. & Ollinger, K. (2013) The lysosome:from waste bag to potential therapeutic target. J. Mol. Cell Biol. 5, 214–226.

Boonen, M., Staudt, C., Gilis, F., Oorschot, V., Klumperman, J. & Jadot, M.(2016) Cathepsin D and its newly identified transport receptor SEZ6L2can modulate neurite outgrowth. J. Cell. Sci. 129, 557–568.

Bos, E., SantAnna, C., Gnaegi, H., Pinto, R.F., Ravelli, R.B., Koster, A.J., deSouza, W. & Peters, P.J. (2011) A new approach to improve the qualityof ultrathin cryo-sections; its use for immunogold EM and correlativeelectron cryo-tomography. J. Struct. Biol. 175, 62–72.

Bromme, D. (2001) Papain-like cysteine proteases. Current protocols inprotein science/editorial board, John E. Coligan ... [et al.], Chapter 21, Unit21 22.

Catherine Rabouille, E.H. & Nasser, M.A. (1999) Quantitative aspectsof immunogold labeling in embedded and nonembedded sections, inelectron microscopy methods and protocols. Humana Press 117, 125–144.

Choi, S.I., Maeng, Y.S., Kim, T.I., Lee, Y., Kim, Y.S. & Kim, E.K. (2015)Lysosomal trafficking of TGFBIp via caveolae-mediated endocytosis.PLoS One 10, e0119561.

Cortese, K., Diaspro, A. & Tacchetti, C. (2009) Advanced correlativelight/electron microscopy: current methods and new developments us-ing Tokuyasu cryosections. J. Histochem. Cytochem. 57, 1103–1112.

de Boer, P., Hoogenboom, J.P. & Giepmans, B.N. (2015) Correlated lightand electron microscopy: ultrastructure lights up! Nat. Methods 12,503–513.

de Souza, N. (2015) Super-resolution CLEM. Nat. Methods 12, 37–37.Eskelinen, E.L. (2006) Roles of LAMP-1 and LAMP-2 in lysosome biogen-

esis and autophagy. Mol. Aspects Med. 27, 495–502.Faulk, W.P. & Taylor, G.M. (1971) An immunocolloid method for the

electron microscope. Immunochemistry 8, 1081–1083.



Gaietta, G.M., Deerinck, T.J. & Ellisman, M.H. (2011) Correlated live celllight and electron microscopy using tetracysteine tags and biarsenicals.Cold Spring Harbor Protocols 2011, pdb top94.

Giepmans, B.N. (2008) Bridging fluorescence microscopy and electronmicroscopy. Histochem. Cell. Biol. 130, 211–217.

Grammel, M. & Hang, H.C. (2013) Chemical reporters for biological dis-covery. Nat. Chem. Biol. 9, 475–484.

Greenbaum, D., Medzihradszky, K.F., Burlingame, A. & Bogyo, M. (2000)Epoxide electrophiles as activity-dependent cysteine protease profilingand discovery tools. Chem. Biol. 7, 569–581.

Griffiths, G. (1993) Fine-structure preservation. Fine Structure Immunocy-tochemistry. Springer Berlin Heidelberg, Berlin, Heidelberg.

Griffiths, G. & Hoppeler, H. (1986) Quantitation in immunocytochemistry:correlation of immunogold labeling to absolute number of membraneantigens. J. Histochem. Cytochem. 34, 1389–1398.

Henry, R.M., Hoppe, A.D., Joshi, N. & Swanson, J.A. (2004) The unifor-mity of phagosome maturation in macrophages. J. Cell Biol. 164, 185–194.

Hoogendoorn, S., van Puijvelde, G.H., Kuiper, J., van der Marel, G.A.& Overkleeft, H.S. (2014) A multivalent ligand for the mannose-6-phosphate receptor for endolysosomal targeting of an activity-basedprobe. Angew. Chem. Int. Ed. Engl. 53, 10975–10978.

Huotari, J. & Helenius, A. (2011) Endosome maturation. EMBO J. 30,3481–3500.

Kandela, I.K. & Albrecht, R.M. (2007) Fluorescence quenching bycolloidal heavy metals nanoparticles: implications for correlativefluorescence and electron microscopy studies. Scanning 29, 152–161.

Kuipers, J., Ham, T.J., Kalicharan, R.D., Veenstra-Algra, A., Sjollema, K.A., Dijk, F., Schnell, U. & Giepmans, B.N.G. (2015) FLIPPER, a combina-torial probe for correlated live imaging and electron microscopy, allowsidentification and quantitative analysis of various cells and organelles.Cell Tissue Res. 360, 61–70.

Lucocq, J. (1994) Quantitation of gold labelling and antigens in immuno-labelled ultrathin sections. J. Anatomy 184 (Pt 1), 1–13.

Luzio, J.P., Pryor, P.R. & Bright, N.A. (2007) Lysosomes: fusion and func-tion. Nat. Rev. Mol. Cell. Biol. 8, 622–632.

Mayhew, T.M. & Lucocq, J.M. (2008) Developments in cell biology forquantitative immunoelectron microscopy based on thin sections: a re-view. Histochem. Cell. Biol. 130, 299–313.

Ngo, J.T., Adams, S.R., Deerinck, T.J. et al. (2016) Click-EM for imagingmetabolically tagged nonprotein biomolecules. Nat. Chem. Biol. 12,459–465.

Peters, P.J., Bos, E. & Griekspoor, A. (2006) Cryo-immunogold electronmicroscopy. Current protocols in cell biology/editorial board, Juan S. Boni-facino ... [et al.], Chapter 4, Unit 4 7.

Posthuma, G., Slot, J.W. & Geuze, H.J. (1984) Immunocytochemical as-says of amylase and chymotrypsinogen in rat pancreas secretory gran-ules. Efficacy of using immunogold-labeled ultrathin cryosections toestimate relative protein concentrations. J. Histochem. Cytochem. 32,1028–1034.

Predescu, S.A., Predescu, D.N., Timblin, B.K., Stan, R.V. & Malik, A.B.(2003) Intersectin regulates fission and internalization of caveolae inendothelial cells. Mol. Biol. Cell 14, 4997–5010.

Pungercar, J.R., Caglic, D., Sajid, M. et al. (2009) Autocatalytic processingof procathepsin B is triggered by proenzyme activity. FEBS J. 276, 660–668.

Robenek, H., Hofnagel, O., Buers, I., Robenek, M.J., Troyer, D. & Severs,N.J. (2006) Adipophilin-enriched domains in the ER membrane aresites of lipid droplet biogenesis. J. Cell Sci., 119, 4215–4224.

Sarnataro, D., Caputo, A., Casanova, P. et al. (2009) Lipid rafts and clathrincooperate in the internalization of PrP in epithelial FRT cells. PLoS One4, e5829.

Saxon, E. & Bertozzi, C.R. (2000) Cell surface engineering by a modifiedStaudinger reaction. Science 287, 2007–2010.

Sirerol-Piquer, M.S., Cebrian-Silla, A., Alfaro-Cervello, C., Gomez-Pinedo,U., Soriano-Navarro, M. & Verdugo, J.M. (2012) GFP immunogoldstaining, from light to electron microscopy, in mammalian cells. Micron43, 589–599.

Slot, J.W. & Geuze, H.J. (2007) Cryosectioning and immunolabeling. Nat.Protoc. 2, 2480–2491.

Sosinsky, G.E., Gaietta, G.M., Hand, G. et al. (2003) Tetracysteine genetictags complexed with biarsenical ligands as a tool for investigating gapjunction structure and dynamics. Cell Commun. Adhesion 10, 181–186.

Tohda, K., Lu, H., Umezawa, Y. & Gratzl, M. (2001) Optical detection in mi-croscopic domains. 2. Inner filter effects for monitoring nonfluorescentmolecules with fluorescence. Anal. Chem. 73, 2070–2077.

Tokuyasu, K.T. (1973) A technique for ultracryotomy of cell suspensionsand tissues. J. Cell Biol. 57, 551–565.

Tooze, J., Hollinshead, M., Frank, R. & Burke, B. (1987) An antibodyspecific for an endoproteolytic cleavage site provides evidence that pro-opiomelanocortin is packaged into secretory granules in AtT20 cellsbefore its cleavage. J. Cell Biol. 105, 155–162.

van Elsland, D.M., Bos, E., de Boer, W., Overkleeft, H.S., Koster, A.J. & vanKasteren, S.I. (2016) Detection of bioorthogonal groups by correlativelight and electron microscopy allows imaging of degraded bacteria inphagocytes. Chem. Sci. 7, 752–758.

van Elsland, D.M., Bos, E., Overkleeft, H.S., Koster, A.J. & van Kasteren, S.I.(2015) The potential of bioorthogonal chemistry for correlative lightand electron microscopy: a call to arms. J. Chem. Biol. 8, 153–157.

van Meel, E. & Klumperman, J. (2008) Imaging and imagination: under-standing the endo-lysosomal system. Histochem. Cell. Biol. 129, 253–266.

Watanabe, S., Punge, A., Hollopeter, G., Willig, K.I., Hobson, R.J., Davis,M.W., Hell, S.W. & Jorgensen, E.M. (2011) Protein localization in elec-tron micrographs using fluorescence nanoscopy. Nat. Methods 8, 80–84.

Webster, P., Schwarz, H. & Griffiths, G. (2008) Preparation of cells andtissues for immuno EM. Methods Cell. Biol. 88, 45–58.

White, J.G., Krumwiede, M.D. & Escolar, G. (1999) Glycoprotein Ib ishomogeneously distributed on external and internal membranes ofresting platelets. Am. J. Pathol. 155, 2127–2134.


Correlative light and electron microscopy reveals discrepancy … · 2017. 6. 2. · BM-DCs were...

Documents

Transcript of Correlative light and electron microscopy reveals discrepancy … · 2017. 6. 2. · BM-DCs were...