Direct visualization of specifically modified extracellular glycans in living animals.

Modification patterns of heparan sulfate coordinate protein function in metazoans, yet in vivo imaging of such non-genetically encoded structures has been impossible. Here we report a transgenic method in Caenorhabditis elegans that allows direct live imaging of specific heparan sulfate modification patterns. This experimental approach reveals a dynamic and cell-specific heparan sulfate landscape and could in principle be adapted to visualize and analyze any extracellular molecule in vivo.

Fluorescent protein fusions [1] allow live imaging of virtually every protein. In contrast, in vivo visualization of non-genetically encoded molecules such as glycans or lipids is more challenging. Using ‘click’ chemistry some glycans have been visualized in live zebrafish and Caenorhabditis elegans [2,3], but modifications of the glycans, which are crucial for the function of many molecules, cannot be distinguished. For example, heparan sulfates are extracellular glycosaminoglycans with complex modification patterns (Fig. 1a) that are attached to heparan sulfate proteoglycan core proteins such as membrane associated syndecans, glypicans, or the secreted perlecan and agrin [4,5]. Heparan sulfate modification patterns play pivotal roles in metazoan development and physiology by interacting with, and modulating the function of a wide array of proteins [4,5]. Biochemical studies and the identification of specific neuroanatomical defects of mutants in heparan sulfate genes in vertebrates and invertebrates have led to the “Heparan sulfate code” hypothesis which posits that distinct heparan sulfate modification patterns are localized in a tissue- or cell-specific pattern, particularly in the nervous system [6].

Figure 1

A technique to visualize specifically modified heparan sulfate in vivo

(a) Heparan sulfate biosynthesis(NS: N-sulfated, NA: N-acetylated; 2S, 3S, 6S: 2-O-, 3-O-, 6-O-sulfate; C. elegans gene names in parenthesis) and putative protein interacting regions (FGF: fibroblast growth factor). (b) Schematic of a transgenic (scFv) antibody-GFP fusion comprising an N-terminal secretion signal and a C-terminal green fluorescent protein, with the variable heavy chain and light chains in blue and orange, and the complementarity determining regions (CDR) in magenta.. Confocal (c, d, e) and differential interference contrast (f, DIC) images showing colocalization of a cytoplasmic pan-neuronal red fluorescent marker (otIs173) with green HS4C3::GFP. AIY interneurons are labeled by a cytoplasmic GFP reporter transgene. (g)-(o) Heads of L1 animals transgenically expressing the HS4C3::GFP fusion under control of the coelomocyte-specific unc-122 promoter in different mutant backgrounds as indicated. Anterior and dorsal is to the left and top, respectively. An arrowhead indicates the nerve ring region and an asterisk autofluorescence. Scale bar: 25 μm. cc: coelomocyte. (p)-(q) Quantification of relative nerve ring fluorescence in mutants as indicated. Error bars denote the s. e. m. and statistical significance is indicated: ns, not significant; * P < 0.05; *** P < 0.0005.

We developed an in vivo technique to visualize and analyze complex heparan sulfate modification patterns to directly test this hypothesis. To this end, we made use of a subset of heparan sulfate-specific single chain variable fragment (scFv) antibodies [7] that recognize specifically modified forms of HS (Table 1). We fused a secretion signal and green fluorescent protein (GFP) to the N-terminus and C-terminus of heparan sulfate-specific scFv antibodies, respectively and expressed these constructs in six scavenger cells (coelomocytes) in C. elegans (Fig. 1b, Supplementary Fig. 1). Focusing initially on the HS4C3 scFv antibody [8] we found that transgenic expression of a secreted HS4C3::GFP fusion produced discrete staining in vivo, predominantly in the nervous system (Fig. 1c-o). We did not observe similar staining patterns in animals expressing a negative control antibody fusion (MPB49::GFP) that differs from the HS4C3 antibody in merely five amino acids and has no known epitope (Fig. 1h, Table 1)[9]. Staining patterns appeared independent of the promoter used to express the scFv-antibody fusion (Supplementary Fig. 2), the fluorescent protein (Supplementary Fig. 3) or the level of transgene expression (Supplementary Fig. 4). Lastly, the stained structures in the nervous system are distant from the site of antibody expression and secretion from the coelomocytes. The in vivo staining patterns of transgenic antibodies were overlapping but not identical to immunohistochemical stains of fixed animals which in addition to the nervous system and pharynx included the intestine (Supplementary Fig. 5). This suggests that some heparan sulfate may be located in cellular compartments that were inaccessible to the secreted transgenic scFv antibody GFP fusions. We conclude that the observed staining patterns are specific for the HS4C3::GFP antibody fusion.

Table 1

HS binding characteristics of scFv antibodies in vitro and in vivo

scFv AB	heavy chain	CDR3 [a]	in vitro modifications [b]	in vivo modifications [c]
HS4C3 [9]	DP-38	GRRLKD	6S, 3S, (2S)	6S, 3S, (2S)
HS3A8 [14]	DP-38	GMRPRL	6S, 2S, IdoA [14]	6S, 2S
AO4B08 [14]	DP-47	SLRMNGWRAHQ	NS, 6S, 2S, IdoA [15]	6S, 2S
EW3G6 [16]	DP-38	GGTTRIRK	nd	nd
MPB49 [9]	DP-38	WRNDRQ	na	na

Abbreviations: CDR3: variable heavy chain complementarity determining region 3; 6S: HS 6-O-sulfate; 3S: HS 3-O-sulfate; 2S: HS 2-O-sulfate; NS: HS N-sulfate, IdoA: iduronic acid as a result of HS C-5 glucuronyl-epimerization of glucuronic acid; nd: not determined; na: not applicable.

HS modifications required for optimal competition in ELISAs with major and (minor) contributions.

HS modifications required for binding as determined in this study. Note that in all cases the recognized HS epitopes in vivo were attached to the HS core protein syndecan, at least within the nerve ring.

In vitro, the scFv HS4C3 recognizes an HS epitope comprising 3-O-sulfated and 6-O-sulfated glucosamine residues with a minor contribution of 2-O-sulfation on hexuronic acid moieties [8]. To determine whether the transgenically expressed HS4C3::GFP antibody fusion recognized heparan sulfate with similar specificity we removed genes required for heparan sulfate modifications and quantified relative fluorescence, using the prominent nerve ring staining as a proxy (Supplementary Fig. 6a). Genetic removal of the heparan sulfate 6-O- sulfotransferase hst-6 or both 3-O-sulfotransferases reduced nerve ring staining (Fig. 1i,l,p), whereas individual removal of the 3O-sulfotransferases hst-3.1 or hst-3.2 resulted in no reduction of nerve ring staining (Fig. 1j,k,p). This suggested that hst-3.1 and hst-3.2 can act redundantly in forming the heparan sulfate modification pattern recognized by the HS4C3::GFP antibody fusion. In contrast, genetic removal of the heparan sulfate C-5 epimerase hse-5 or 2-O-sulfotransferase hst-2 increased nerve ring staining (Fig. 1m,n,p). Since hst-2 restricts heparan sulfate 6-O-sulfation [10], its removal increased 6-O-sulfation which may be sufficient for increased binding or, alternatively, 2-O-sulfation may have an inhibitory effect on binding of the HS4C3::GFP antibody fusion. We found nerve ring staining in the hst-6 hst-2 double mutant significantly reduced compared to the hst-6 single mutant (P = 1.25 × 10−7, n = 34, Fig. 1i,o,p), demonstrating that (1) increased staining in hst-2 mutants was hst-6 dependent and (2) that in the absence of hst-6, nerve ring staining was dependent on, rather than being inhibited by, the 2-O-sulfotransferase hst-2 and by inference 2-O-sulfation..

We next tested which HS core protein(s) bears the heparan sulfate epitope recognized by the HS4C3::GFP antibody fusion. Removal of the single C. elegans sdn-1/syndecan, but not the two glypicans (gpn-1, lon-2), perlecan (unc-52), or agrin eliminated staining of the nerve ring as well as nerve cords, indicating that the majority of heparan sulfate recognized by the HS4C3::GFP antibody fusion is associated with the nervous system and attached to the syndecan core protein (Fig. 1q, Supplementary Fig. 6b-g). Increased nerve ring staining upon loss of gpn-1/glypican was sdn-1/syndecan-dependent (Fig. 1q, Supplementary Fig. 6b,c,f), possibly due to compensatory mechanisms. Interestingly, staining of structures outside of the nerve ring was reduced in strains lacking sdn-1/syndecan (Supplementary Fig. 6c,f), consistent with the interpretation that HS4C3::GFP recognized heparan sulfate epitopes attached to sdn-1/syndecan with possibly lower affinity. The differences in staining in heparan sulfate mutants were neither due to changes in transgene expression (Supplementary Fig. 7), nor the permeability of basement membranes as a result of an altered charge distribution [11]. First, the distribution of the MPB49::GFP control antibody fusion was unaffected in heparan sulfate mutant backgrounds (Supplementary Fig. 8) and second, different heparan sulfate-specific antibody GFP fusions displayed distinct specificities compared to the HS4C3::GFP antibody fusion (Table 1, Supplementary Fig. 9). Thus, the HS4C3::GFP fusion recognizes a heparan sulfate modification pattern in vivo that is borne by the syndecan core protein, and consistent with in vitro binding studies of HS4C3 [8], involves 6-O-, 3-O- and 2-O-sulfation for optimal binding.

To study the dynamics and diversity of heparan sulfate modification patterns in vivo, we analyzed animals expressing HS4C3::GFP and several additional heparan sulfate-specific scFv-antibody fusions with different binding characteristics, including HS3A8::GFP, AO4B08::GFP, and EW3G6::GFP (Table 1, Supplementary Fig. 9). The HS4C3::GFP antibody fusion was associated specifically with the nervous system from the embryonic 3-fold stage through larval into adult stages (Fig. 2a-c,g, Supplementary Fig. 10). The nerve ring which forms the major neuropil in the worm appeared diffusely stained, whereas staining of the dorsal and ventral nerve cords appeared more punctate and of approximately equal intensity (Fig. 2g). Thus, heparan sulfate may be associated with motor neuron processes which are present in roughly equal numbers in the dorsal and ventral cords. During later larval stages, we observed discrete staining of the vulval epithelium and uterine tissues (Fig. 2f). By late larval and adult stages, we detected additional staining of the basement membranes that surround the body wall muscles (Fig. 2c,e,k) and the anterior and posterior bulbs of the pharynx (Fig. 2g). We do not know whether as a secreted molecule heparan sulfate is expressed by the stained tissues or produced non-autonomously by surrounding tissues.

Figure 2

Distinct subcellular HS modification patterns

(a) Embryo at 3-fold stage expressing the HS4C3::GFP fusion construct. Anterior is to the left in all panels and white arrowheads point to the nerve ring (a, b, g-i) or the vulval epithelium and uterine cells (f). Arrows indicate the junction between the pharynx and the intestine (i) or the neurites of possibly an individual pair of cells (j). A black asterisk denotes gut autofluorescence and a white asterisk the vulva. Scale bar: 25 μm. bwm: body wall muscles, cc: coelomocytes, dnc: dorsal nerve cord, pag: preanal ganglion, and vnc: ventral nerve cord. (b) L1 larval stage and (c) L4 larval stage animals expressing the HS4C3::GFP fusion construct. (d)-(e) Medial (d) and a left, more superficial focal plane (e) of a tail of a young adult male displaying staining associated with the preanal ganglion and the intestinal basement membranes (d) and body wall muscles (e). (f) Ventral view showing staining of the vulval epithelium, uterine and possibly cells of the somatic gonad in L4 larval animals expressing the HS4C3::GFP fusion. (g)-(j) Maximum intensity projection of the head of a wild type L4 animal transgenically expressing scFv antibody-GFP fusions as indicated.

We found staining by the HS3A8::GFP fusion to be very similar, yet stronger than HS4C3::GFP (Fig. 2h). Since in vivo binding characteristics of HS3A8 and HS4C3 were distinct (Table 1, Supplementary Fig. 9) and both antibody fusions showed incomplete overlap in colocalization studies (Supplementary Fig. 11), we suggest that HS4C3 recognizes a heparan sulfate modification pattern that partially overlaps with the epitope recognized by HS3A8. Similarly, the AO4B08::GFP fusion labeled the nerve ring and the nerve cords in a both overlapping and distinct fashion. The nerve ring displayed more discrete, sheath-like staining in close apposition to the pharynx, rather than the diffuse staining seen with the HS4C3::GFP fusion (Fig. 2i). This staining could be associated with the sheath-like extensions of the GLR cells or the muscle plate that wraps around the pharynx [12]. Intriguingly, the EW3G6::GFP fusion did not label body wall muscles at all. Rather, EW3G6::GFP appeared to stain exclusively part of the neurites but not cell bodies of a single pair of neurons. This suggests the existence of unique and cell-specific heparan sulfate modification patterns that are localized to defined subcellular structures (Fig. 2j). We also detected differences in staining outside of the nervous system (Fig. 2i, Supplementary Figure 12). We conclude that individual cells can express cell-specific heparan sulfate modification patterns that are associated with defined subcellular-specific structures and dynamically remodeled during development.

Whereas the C. elegans system can take advantage of coelomocytes to remove unbound antibody fusions, transgenic approaches in other genetic models could use inducible systems to limit antibody production. The transgenic expression of scFv antibody GFP fusions shares several unique features. First, while shown here for heparan sulfate, this approach should in principle be applicable to live imaging of any molecule such as glycans or lipids, as well as proteins undergoing posttranslational modification through phosphorylation, sulfation, acetylation or glycosylation as long as specific scFv antibodies exist. Second, it renders non-genetically encoded molecules amenable to genetic analysis to probe their biosynthesis and, possibly, function. While we do not observe obvious defects in transgenic worms, our method could conceivably combine visualization with functional interference if the antibody fragment had the ability to functionally neutralize its epitope, as has been shown in cell culture experiments [13]. Last, it should allow to study the dynamics of such molecules in vivo, including live 4D imaging of remodeling and, possibly, turnover.. Our technique potentially allows development of in vivo biosensors for essentially any molecule and would expand the toolbox for in vivo visualization and analysis of molecules in living animals.

Online Methods

Strains

Worms were cultured as previously described [17]. The following strains and alleles were used: N2 Bristol wild-type, hse-5(tm472) (OH1487), hst-2(ok595) (OH1876), hst-6(ok273) (OH1421), hst-3.1(tm734) (EB649), hst-3.2(tm3006) (EB650), sdn-1(zh20), gpn-1(ok377), lon-2(e678), unc-52(e998), agr-1(tm2051), unc-104(rh43), otIs76 mgIs18 (OH904), dzEx195 [Punc-122::HS4C3::GFP(S65T); rol-6(su1006)] (EB313), dzEx341 [Punc-122::HS3A8::GFP(S65T); rol-6(su1006)] (EB569), dzEx369 [Punc-122::AO4B08::GFP(S65T); rol-6(su1006)] (EB637), dzEx475 [Punc-122::EW3G6::2xsfGFP; rol-6(su1006)] (EB917), dzEx206 [Punc-122::MPB49::GFP(S65T); rol-6(su1006)] (EB324), dzIs10V [Punc-122::HS4C3::GFP(S65T); rol-6(su1006)] (EB502), dzIs13 [Punc-122::HS3A8::GFP(S65T); rol-6(su1006)] (EB667), dzIs14 [Punc-122::AO4B08::GFP(S65T); rol-6(su1006)] (EB668), dzIs16 [Punc::122::MPB49::GFP(S65T)] (EB703), otIs173III [Prgef-1::DsRed2; Pttx-3::GFP] [18]. Doubles between heparan sulfate mutant strains and antibody expressing strains were constructed using standard procedures. Details for genotyping are available on request.

To create transgenic worms, single chain variable fragment (scFv) antibody constructs were injected with pRF4 (rol-6(su1006)) as a dominant injection marker at 25 ng/μl and 100 ng/μl, respectively. Several transgenic lines giving comparable results were obtained for each construct. The transgenic lines dzEx195, dzEx341, dzEx369, and dzEx206 were integrated to yield dzIs10V, dzIs13, dzIs14, and dzIs16, respectively. All mutants and integrated transgenes were backcrossed at least four times prior to analysis.

scFv::GFP antibody fusion constructs

A panel of scFv antibodies that bind different heparan sulfate preparations were originally isolated from a phage display library [19]. The antibodies comprise different variable heavy chains as indicated (Table 1) and an invariant variable light chain [19]. To assemble the transgenic antibody expression constructs, we used the following general scheme. Sequences of the invariable light chain and each published variable heavy chain were optimized for C. elegans codon usage and synthesized de novo by Entelechon (Regensburg, Germany). The scFv heavy chains were then assembled with the invariable light chain using a set of unique restriction sites (Supplementary Fig. 1). Subsequently, the scFv antibodies were cloned SalI/AgeI into pJF33 [Pmyo-3::secGFP(S65T)] [20] to yield a fusion protein in which the scFv antibody is N-terminally fused in frame to the SEL-1 signal sequence [21] and C-terminally to the S65T green fluorescent protein variant (Supplementary Fig. 1). scFv constructs sharing the same heavy chain but differing in the CDR3 region (EW3G6 and HS3A8) were built by PCR mediated mutagenesis using a Quickchange kit (Stratagene) using the following primers TATTATTGTGCCAGAGGAGGAACCACCCGCATCCGCAAGTGGGGACAGGGAACA and TGTTCCCTGTCCCCACTTGCGGATGCGGGTGGTTCCTCCTCTGGCACAATAATA (for EW3G6) and, TATTATTGTGCCAGAGGAATGCGTCCACGCCTTTGGGGACAGGGAACA and TGTTCCCTGTCCCCAAAGGCGTGGACGCATTCCTCTGGCACAATAATA (for HS3A8). The completed scFv antibody cassettes were subcloned under the control of the coelomocyte-specific unc-122 promoter [22] or glo-1 promoter [23]. The EW3G6 scFv was fused with two consecutive sfGFP (superfolder GFP) molecules (2xsfGFP) using an established PCR fusion technique [24] and cloned under control of the unc-122 promoter. All sequences that were synthesized de novo or underwent PCR amplification were verified by sequencing and are available upon request.

Fluorometric and microscopic analyses

For quantitative epifluorescence microscopy, worms were synchronized by hypochlorite bleaching and plated on regular nematode plates with OP50 E. coli. Approximately 24h later, animals were washed off in M9, anaesthetized with 1mM levamisole, mounted on a 5% agarose pad and analyzed on an AxioImagerZ1 compound microscope (Zeiss) or a Leica SP2 AOBS confocal microscope. All images used for quantification were taken with an exposure time of 250 ms at 400× magnification and fluorimetrically analyzed using Zeiss AxioVision software version 4.8 as follows. Relative pixel intensity in a region of “specific fluorescence” around the nerve ring was measured and divided by the relative pixel intensity measured in a region of “total fluorescence” comprising the whole head to the posterior end of the pharynx (Supplementary Fig. 6) to yield relative fluorescence. Care was taken to exclude fluorescing coelomocytes from this quantification, although this was not possible in a few select cases due to coelomocyte migration defects (e.g. some sdn-1 mutants animals). Since this did not obviously affect total head fluorescence (Supplementary Fig. 7) which serves as the denominator, we included these values in the calculations of relative fluorescence. Statistical significance was calculated using the two-tailed Student’s t-test with the Bonferroni correction where applicable.

Microscopic images were acquired with an AxioCam MRm camera mounted on an AxioImager Z1 compound microscope using Axiovision 4.8 Zeiss proprietary software. For preparation of figures we exported images from Axiovision 4.8 by applying a 5% bestfit gamma curve. Some animals were optically sectioned using the Zeiss Apotome on an AxioImager Z1 or a Leica SP2 AOBS confocal microscope. Optical sections were processed as 12-bit grayscale images using Zeiss Axiovision software 4.8 (if obtained with the Apotome) or as 8-bit grayscale images using ImageJ (if obtained with the Leica confocal) to produce maximum intensity projections. All images were assembled into figures using Adobe Photoshop and Adobe Illustrator.