Expression of Antibody Fragments with a Controlled N -Glycosylation Pattern and Induction of Endoplasmic Reticulum-Derived Vesicles in Seeds of Arabidopsis

Intracellular trafﬁcking and subcellular deposition are critical factors inﬂuencing the accumulation and posttranslational modiﬁcations of proteins. In seeds, these processes are not yet fully understood. In this study, we set out to investigate the intracellular transport, ﬁnal destination, N -glycosylation status, and stability of the fusion of recombinant single-chain variable fragments to the crystallizing fragment of an antibody (scFv-Fc) of two antiviral monoclonal antibodies (2G12 and HA78). The scFv-Fcs were expressed in Arabidopsis ( Arabidopsis thaliana ) seeds and leaves both as secretory molecules and tagged with an endoplasmic reticulum (ER) retention signal. We demonstrate differential proteolytic degradation of scFv-Fcs in leaves versus seeds, with higher degradation in the latter organ. In seeds, we show that secretory versions of HA78 scFv-Fcs are targeted to the extracellular space but are deposited in newly formed ER-derived vesicles upon KDEL tagging. These results are in accordance with the obtained N -glycosylation proﬁles: complex-type and ER-typical oligomannosidic N -glycans, respectively. HA78 scFv-Fcs, expressed in seeds of an Arabidopsis glycosylation mutant lacking plant-speciﬁc N -glycans, exhibit custom-made human-type N -glycosylation. In contrast, 2G12 scFv-Fcs carry exclusively ER-typical oligomannosidic N -glycans and were deposited in newly formed ER-derived vesicles irrespective of the targeting signals. HA78 scFv-Fcs exhibited efﬁcient virus neutralization activity, while 2G12 scFv-Fcs were inactive. We demonstrate the efﬁcient generation of scFv-Fcs with a controlled N -glycosylation pattern. However, our results also reveal aberrant subcellular deposition and, as a consequence, unexpected N -glycosylation proﬁles. Our attempts to elucidate intracellular protein transport in seeds contributes to a better understanding of this basic cell biological mechanism and is a step toward the versatile use of Arabidopsis seeds as an alternative expression platform for pharmaceutically relevant proteins.

Recombinant monoclonal antibodies (mAbs) are of high therapeutic potential and have thus become a major product of the pharmaceutical industry (Aggarwal, 2009). In addition to full-length mAbs, antibodies are being engineered to alter their size, pharmacokinetics, specificity, valency, effector functions, etc. in order to better suit the intended applications (for review, see Filpula, 2007;Harmsen and De Haard, 2007). Of particular interest among these engineered fragments are single-chain variable fragments (scFvs), fusions of variable heavy and variable light domains that retain an antigen-binding function. Due to their smaller size as compared with their full-length counterparts, these molecules penetrate target tissues better and can even bind to intracellular targets. Furthermore, multimerization via disulfide bonds and/or multimerization domains allows for the production of divalent or higher order antibody-like molecules, where increased avidity 1 This work was supported by a European Union Pharma-Planta (Framework VI) Ph.D. stipend to A.L., by project L575-B13 of the Fonds zur Fö rderung der wissenschaftlichen Forschung, by project Nr239375 (Laura Bassi Centres of Expertise) of the Ö sterreichische Forschungsfö rderungsgesellschaft mbH, as well as by the European Union Cooperation in Science and Technology Action FA0804.
* Corresponding author; e-mail herta.steinkellner@boku.ac.at. The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (www.plantphysiol.org) is: Herta Steinkellner (herta.steinkellner@boku.ac.at). [C] Some figures in this article are displayed in color online but in black and white in the print edition. [W] The online version of this article contains Web-only data. [OA] Open Access articles can be viewed online without a subscription.
www.plantphysiol.org/cgi/doi/10.1104/pp. 110.171330 has positive effects on the target-binding efficacy. Very promising products in this respect are so called scFv-Fcs (i.e. fusions of a scFv to the crystallizing fragment of an antibody [Fc; hinge-C H 2-C H 3]). These molecules spontaneously fold and dimerize to give a structure similar to a full-length IgG (Zhang et al., 2007;Cao et al., 2009). Despite their small size, Fc-mediated effector functions are maintained and finally lead to the elimination of the antigen-antibody complex from the blood stream (Powers et al., 2001). An additional benefit of such molecules is their composition: one scFv-Fc consists of two polypeptides from one gene. Thus, homodimerization but not heterotetramerization is necessary for correct folding, which could allow for more efficient expression. In general, scFv-Fcs exhibit characteristics equivalent to their parent IgG (Shu et al., 1993;Powers et al., 2001;Cao et al., 2009) and have been tested for similar applications (Li et al., 2000;Yuan et al., 2006;Mori and Kim, 2008;De Lorenzo and D'Alessio, 2009;Olafsen et al., 2009;Riccio et al., 2009).
Plant seeds are increasingly becoming an attractive alternative platform for the production of recombinant proteins. In particular, low production costs, high accumulation of recombinant proteins in the small volume of the seed, and product stability over years allows the separation of cultivation from the purification process, thereby making this system an economically feasible proposition (for review, see Boothe et al., 2010). Notably, with the model plant Arabidopsis (Arabidopsis thaliana), very high accumulation levels have been reported, with recombinant polypeptides even exceeding 35% of the total soluble protein, which corresponds to over 70 g of recombinant protein per kg of dry seeds (De Jaeger et al., 2002). Regulatory elements from highly expressed seed storage proteins enable these elevated accumulation levels to be reached and have already been used for the expression of a scFv-Fc (Van Droogenbroeck et al., 2007). However, the high accumulation of recombinant proteins in seeds, including mAbs and scFv-Fcs, not only depends Figure 1. Schematic overview of the T-DNA region of the expression constructs generated in this study. wt-Ph 2G12sc SEC , wt-Ph 2G12sc KDEL , wt-Ph HA78sc SEC , and wt-Ph HA78sc KDEL were cloned into the vector pPhasGW (F. Morandini, B. Van Droogenbroeck, and A. Depicker, unpublished data) for transformation into wild-type and TKO plants. wt-35S 2G12sc SEC and wt-35S HA78sc SEC were additionally cloned into the binary expression vector pPT2 (Strasser et al., 2005). LB, Left border; 3#ocs, 3# end of the octopine synthase gene; nptII, neomycin phosphotransferase II; Pnos, nopaline synthase promoter; Pphas, b-phaseolin promoter (1-1,470; GenBank accession no. J01263); attB1, attB2, attP1, attP2, attL1, attL2, attR1, and attR2, recombination sites for Gateway cloning (Invitrogen, 2003); CmR-ccdB, Gateway positive/negative selection cassette; 3#arc5-I, approximately 4,000 bp of 3# flanking region of the arceline 5I gene (part of Gen-Bank accession no. Z50202); RB, right border; 2S2, signal peptide of the Arabidopsis 2S2 seed storage protein (Krebbers et al., 1988); KDEL, ER retrieval motif; Tnos, nopaline synthase terminator; P35S, cauliflower mosaic virus 35S promoter; g7T, 200 bp of transcript 7 3# region (bp 398-598, Dhaese et al., 1983). upon the level of synthesis but also on their stability after deposition within a suitable compartment. In this respect, a frequently reported approach is KDELmediated retention within the endoplasmic reticulum (ER), which seems to allow for increased protein accumulation (Wandelt et al., 1992;Fiedler et al., 1997;Ko et al., 2005;Petruccelli et al., 2006;Laguía-Becher et al., 2010).
The deposition of recombinant proteins in a specific subcellular compartment has identified peculiarities of the seed secretory system, which obviously differs from the commonly accepted default pathway. The most obvious difference is the absence of a lytic vacuole, in terms of size of the dominant organelle of most plant cells, and instead the presence of protein storage vacuoles (PSVs), where the endogenous seed storage proteins are deposited (for review, see Mü ntz, 1998;Robinson et al., 2005). However, the targeting of recombinant proteins in seeds is not yet fully understood. For example, KDEL-tagged as well as secretory proteins have been detected in protein storage vacuoles, in the cell wall, in ER-derived compartments, or in a combination of the three (Herman et al., 1990;Philip et al., 1998Philip et al., , 2001Wright et al., 2001;Reggi et al., 2005;Downing et al., 2006;Petruccelli et al., 2006;Van Droogenbroeck et al., 2007;Abranches et al., 2008;Schmidt and Herman, 2008;Loos et al., 2011). As a consequence of this atypical targeting and deposition, seed-derived recombinant products may carry unusual N-glycosylation profiles (Van Droogenbroeck et al., 2007;Rademacher et al., 2008;Ramessar et al., 2008). Since the glycosylation status of many therapeutic proteins, including mAbs and scFv-Fcs, has a profound effect on their biological activity, it is of great importance to fully understand the molecular mechanisms that underlie targeting along the secretory pathway and final deposition.
In this study, we set out to explore further the potential of plant seeds as a versatile expression platform for valuable recombinant proteins. To this end, a series of scFv-Fc versions of the two antiviral mAbs 2G12 and HA78 were placed under the control of seedspecific regulatory sequences. Secretory as well as KDEL-tagged versions were used to transform Arabidopsis wild-type plants as well as a glycosylation mutant lacking plant-specific N-glycosylation (Strasser et al., 2004). For comparison, some of these constructs were additionally placed under the control of the constitutive cauliflower mosaic virus 35S promoter. Expression levels and the stability of seed-and leafproduced recombinant scFv-Fcs were determined using ELISA and immunoblotting. N-Glycosylation profiles of the different scFv-Fc variants were determined by mass spectrometry, and the actual subcellular localization of the seed-produced scFv-Fcs was demonstrated by immunogold electron microscopy (IEM). Finally, functional activities of the scFv-Fc variants were determined by antigen-binding and virus neutralization assays.

Cloning and Plant Transformation
Secretory and KDEL-tagged scFv-Fc versions of two different model antibodies (2G12 against Human immunodeficiency virus [HIV] and HA78 against Hepatitis A virus; Trkola et al., 1996;Cao et al., 2008) were constructed and cloned into the pPhasGW destination  Fig. 1) and were expressed under the control of the seed-specific b-phaseolin promoter from Phaseolus vulgaris (De Jaeger et al., 2002;Van Droogenbroeck et al., 2007). The vectors were used to transform Arabidopsis Columbia wild-type (wt) plants and glycosylation mutants lacking plant-specific core a-1,3-fucosylation and b-1,2-xylosylation, referred to as "triple knockout" (TKO; Strasser et al., 2004). In addition, the secretory versions were fused to the constitutive cauliflower mosaic virus 35S promoter in the vector pPT2 (Strasser et al., 2005) and used to transform Arabidopsis wild-type plants. In total, eight different combinations of construct and genetic background were established (Table I).

Expression of scFv-Fc Molecules
At least 25 primary transformants of each series of transformants were screened by ELISA and immunoblotting ( Fig. 2; Table II). Based on expression levels, two lines each were selected and selfed to obtain homozygous plants. Maximal expression levels as determined by ELISA ranged from 0.8 to 9.4 mg recombinant protein g 21 dry seed weight (Table II). Generally, the HA78 constructs accumulated to a higher level than the 2G12 constructs and KDEL tagging did not lead to an increased accumulation. These results are in accordance with the recently expressed full-length versions of 2G12 and HA78 mAbs in Arabidopsis seeds (Loos et al., 2011). Expression levels of the 35S-driven constructs were not analyzed in detail. However, when compared with the b-phaseolin-driven constructs, they were significantly lower (as deduced from immunoblotting).
Seed extracts from transformed plants were subjected to immunoblotting and revealed strong signals consisting of double bands at approximately 60 kD (Fig.  2). The smaller, less intense band represents the nonglycosylated fraction, as previously shown by Van Droogenbroeck et al. (2007). Additionally, degradation products are visible at around 28 to 34 kD. These fragments are derived from heavy chain domains as determined by mass spectrometry (MS) analysis of tryptic peptides (data not shown). KDEL-tagged versions exhibited a slightly increased mass of the intact molecule as well as of their approximately 30-kD degradation products (Fig. 2), most likely due to the KDEL tag and a different N-glycosylation pattern. Although the KDEL tag does not lead to an increased accumulation of scFv-Fcs, it seems to confer enhanced stability, which is especially pronounced for the HA78 scFv-Fcs (Fig. 2). The degradation pattern of the 35Sdriven constructs ( wt-35S 2G12sc SEC and wt-35S HA78sc SEC ) extracted from leaves clearly differs from the degradation pattern of the scFv-Fcs extracted from seeds (Figs. 2 and 3), indicating the presence of different proteases in these two tissues. Surprisingly, 35S-driven scFv-Fcs ( wt-35S 2G12sc SEC and wt-35S HA78sc SEC ) accumulated a higher proportion of intact protein in leaves than in seeds (Figs. 2 and 3; Supplemental Fig. S5).

N-Glycosylation
All constructs carry one N-glycosylation site in the C H 2 domain. The N-glycosylation status of Protein A-purified scFv-Fcs was analyzed by liquid-chromatography (LC)-MS (Fig. 4). KDEL-tagged versions of 2G12 and HA78 ( wt-Ph HA78sc KDEL and wt-Ph 2G12sc KDEL ) exhibited as expected exclusively oligomannosidic structures, mainly Man7 and Man8. Seed-produced secretory HA78 constructs ( wt-Ph HA78sc SEC and TKO-Ph HA78sc SEC ) carry complex-type glycans, GnGnXF and GnGn, respectively, depending on the expression host, Columbia wild type or TKO. In addition, substantial amounts of oligomannosidic structures are present. This ER-typical glycosylation is completely absent when secretory HA78 is expressed in leaves: Figure 3. Immunodetection of 35S-driven scFv-Fcs extracted from seeds and leaves. Nine microliters of crude seed extract and 5 mL of crude leaf extract (corresponding to 90 mg of seeds and 375 mg of leaf material) were separated by SDS-PAGE, blotted on a nitrocellulose membrane, and detected with goat anti-human IgG (H+L) HRP conjugate (Promega; W403B) and a chemiluminescent substrate (Super-Signal West Pico Chemiluminescent Substrate; Pierce). Film exposure for 30 s reveals scFv-Fc bands as well as degradation products. Bars indicate the height of marker bands (in kD).

wt-35S
HA78sc SEC extracted from leaves exhibits exclusively complex-type N-glycans, with GnGnXF as the single dominant structure. Surprisingly, all 2G12 constructs, irrespective of their targeting information or their regulatory sequences ( wt-Ph 2G12sc SEC , wt-Ph 2G12sc KDEL , TKO-Ph 2G12sc SEC , and wt-35S 2G12sc SEC extracted from leaf), carry exclusively oligomannosidic N-glycans.
The 35S-driven constructs accumulated only at a low level in seeds (Fig. 3). Thus, it was not possible to purify sufficient amounts of 35S-driven scFv-Fc from this tissue to proceed to N-glycan analysis. However, the N-glycan analysis of HA78 degradation fragments revealed the presence of complex (GnGnXF) and oligomannosidic structures (Supplemental Figs. S5 and S6; Supplemental Results S1). Interestingly, small amounts of single GlcNAc stubs were found as well. These single GlcNAc stubs have not been found on any of our seed-or leaf-produced scFv-Fcs and mAbs (Loos et al., 2011) but are known from other systems (Islam et al., 1991;Zeng et al., 2003;Rademacher et al., 2008;Ramessar et al., 2008;Floss et al., 2009).

Man7 Isomer Distribution
Another interesting observation was the presence of a relatively high proportion of Man7 structures in seed-extracted scFv-Fcs, even in constructs designed for secretion (i.e. wt-Ph HA78sc SEC , TKO-Ph HA78sc SEC, wt-Ph 2G12sc SEC , and TKO-Ph 2G12sc SEC ). As known from other studies, this Man7 structure occurs in at least three isomers, Man7.1, Man7.2, and Man7.7 (Neeser et al., 1985;Tomiya et al., 1991). There is some evidence that Man7.1 isomers are substrates for the endoplasmic reticulum-associated protein degradation (ERAD) pathway, which directs misfolded proteins from the ER for ubiquitination and subsequent proteasomal degradation (Clerc et al., 2009). We set out to analyze the Man7 isomer composition of scFv-Fcs in more detail (Fig. 5). In fact, wt-Ph HA78sc SEC and TKO-Ph HA78sc SEC had the lowest proportion of Man7.1 (approximately 30% of the Man7 glycans and approximately 8% of the oligomannosidic fraction). The ER-retained wt-Ph HA78sc KDEL contains clearly more Man7.1 (approximately 62% of the Man7 glycans and approximately 13% of the oligomannosidic fraction), while wt-Ph 2G12sc SEC shows the highest amount (approximately 90% of the Man7 glycans and 26% of the oligomannosidic fraction).

Subcellular Localization
In order to reveal the stations of intracellular transport and the final destination of the recombinant scFv-Fcs, IEM was carried out on mature and developing seeds. The final deposition status of the target proteins can be determined in mature seeds; however, more organelles are visible in developing seeds, therefore enabling a more detailed investigation of intracellular trafficking. Plants that were transformed with scFv-Fcs driven by the seed-specific phaseolin promoters (i.e. wt-Ph 2G12sc SEC , wt-Ph 2G12sc KDEL , wt-Ph HA78sc SEC , and wt-Ph HA78sc KDEL ) were analyzed. The results for mature seeds are shown in Supplemental Figures S1 to S4 and in Supplemental Results S1.
Intense labeling of the extracellular space was obtained in seeds expressing wt-Ph HA78sc SEC , showing the efficient secretion of the scFv-Fc to that compartment (Fig. 6A). In addition, dense vesicles were intensely labeled (Fig. 6B), but minor amounts of gold particles were also detected in the Golgi stack itself (Fig. 6C). This labeling pattern is reminiscent of the expression of the secretory full-length antibody versions of 2G12 and HA78 in Arabidopsis seeds, which also localize to the same structures (Loos et al., 2011).

wt-Ph
HA78sc KDEL accumulated in globular, membranedelimited structures of around 200 to 400 nm diameter (Fig. 7). These structures were partially studded with ribosomes, indicating their ER origin, and are thus called endoplasmic reticulum-derived vesicles (ERVs). The PSVs were consistently only slightly labeled (Fig.  7C). However, none of the other compartments, like the Golgi apparatus (Fig. 7A), putative multivesicular bodies (Fig. 7B), or the extracellular space (data not shown), was labeled. Mature seeds expressing wt-Ph HA78sc SEC also showed gold particles in the extracellular space; however, in contrast to developing seeds, ERVs were additionally present in the cytoplasm and labeled (Supplemental Fig. S1). Mature seeds expressing wt-Ph HA78sc KDEL exhibited labeling exclusively in ERVs and dilated nuclear envelope (Supplemental Fig. S2).
Surprisingly, IEM studies in developing seeds expressing wt-Ph 2G12sc SEC (Fig. 8) exhibited a similar labeling pattern as obtained for wt-Ph HA78sc KDEL (Fig.  7). Gold label was found exclusively in globular, partly ribosome-studded ER-derived vesicles of around 200 to 400 nm. In addition, the nuclear envelope was partially dilated and labeled (Fig. 8A), which was not observed for the HA78 scFv-Fc constructs. Golgi stacks, the extracellular space, as well as the PSV were devoid of label (Fig. 8, B, D, and E). Such an unexpected labeling pattern was also detected in mature seeds (Supplemental Fig. S3). These findings were surprising, as the protein was designed for secretion; however, they were consistent with the observed ER-typical oligomannosidic N-glycan profile. For wt-Ph 2G12sc KDEL , IEM was conducted on mature seeds only (Supplemental Fig. S4) and gave a very similar result as for wt-Ph 2G12sc SEC, with intensive labeling of ERVs. Notably, ERVs were not detected in nontransformed Arabidopsis seeds (data not shown).

Antigen-Binding Activity and in Vitro Virus Neutralization
Functional activities of seed-produced scFv-Fcs were also determined. First, the antigen-binding capacities of purified wt-Ph 2G12sc KDEL , wt-Ph HA78sc SEC , wt-Ph HA78sc KDEL , as well as TKO-Ph HA78sc SEC were determined by ELISA. Whereas all HA78 scFv-Fc variants exhibited binding comparable to a Chinese hamster ovary (CHO)-produced full-length mAb (Fig. 9), none of the 2G12 constructs was able to bind the respective antigen (data not shown). In vitro virus neutralization assays were conducted with the same scFv-Fcs. Again, the three HA78 scFv-Fc variants exhibited a hepatitis A virus neutralization efficiency similar to CHOand plant-produced full-length mAbs (Table III), while wt-Ph 2G12sc KDEL did not show any neutralizing activity (Table IV).

DISCUSSION
In this study, we focused on the expression, stability, intracellular trafficking, deposition pattern, and N-glycosylation status of recombinantly expressed scFv-Fc versions of two antiviral antibodies. We have exploited the capability of plant seeds to produce high amounts of these complex, potentially therapeutic proteins. The accumulation levels of the scFv-Fcs expressed in Arabidopsis seeds lie between 0.8 and 9.4 mg mg 21 dry seeds (Table II). This is in the same range as recently published for the respective fulllength IgGs (Loos et al., 2011) but clearly lower than previously reported for scFvs and scFv-Fcs (74 mg mg 21 dry seeds [De Jaeger et al., 2002] and 26 mg mg 21 dry seeds [Van Droogenbroeck et al., 2007], respectively). The reason for the lower accumulation level  is not clear but may lie in sequence elements with unfavorable codon usage. Thus, codon optimization might further increase expression levels, as has been frequently shown (e.g. Geyer et al., 2010).
Besides the intact protein, secretory versions of both 2G12 and HA78 scFv-Fcs exhibit a substantial amount of degradation products. This degraded fraction is more pronounced for the scFv-Fc constructs than for the recently reported respective IgGs (Loos et al., 2011), indicating a reduced stability of the engineered scFv-Fc compared with the full-length counterparts. Surprisingly, seed-produced scFv-Fcs seem to be more stable when expressed to a higher level: under the control of the stronger, seed-specific b-phaseolin promoter, the ratio of intact protein to degraded peptides is higher than for the 35S-driven constructs (compare the ratio of intact scFv-Fcs versus degraded polypeptides in Fig. 2, lane 4, and Fig. 3, lane 2). One possible explanation is that proteases responsible for degradation are overloaded by the vast amount of scFv-Fc produced by the b-phaseolin promoter but still cope with the smaller amount produced by the 35S promoter. Alternatively, a different temporal and spatial control of expression might be responsible for the observed differences in degradation (Takaiwa et al., 1995).
When comparing seed-and leaf-produced scFv-Fcs, a clear difference in the degradation pattern is visible (Figs. 2 and 3). This indicates that a different set of proteases is active in the different tissues or might represent differences in the secretory pathways of these two tissues. The higher ratio of intact scFv-Fcs versus degraded peptides in leaves compared with seeds suggests that 35S-driven scFv-Fcs are more stable in leaves (Fig. 3). This contrasts with the prevailing notion that seeds possess a lower lytic potential and are thus especially well suited for recombinant protein production (Fiedler and Conrad, 1995;Stoger et al., 2002). An alternative suggestion to increase the protein stability is retention in the ER via a Figure 7. Subcellular localization of wt-Ph HA78sc KDEL in developing Arabidopsis seeds by IEM. A, Gold label was nearly exclusively found in globular structures that were partially ribosome studded (arrowheads), indicating an ER origin (ERVs). The nuclear envelope was neither swollen nor labeled. B, A putative multivesicular body was not labeled. C, ER-derived globular structures were strongly labeled, the PSV was slightly labeled, and oil bodies were free of labeling. N, Nucleus; NE, nuclear envelope; MVB, multivesicular body; OB, oil body. Bars = 200 nm. Figure 8. Subcellular localization of wt-Ph 2G12sc SEC in developing Arabidopsis seeds by IEM. A, Gold label was mainly found in ERVs. The nuclear envelope was partially swollen and labeled in a density similar to the ERVs. B, The ERVs were partially ribosome studded (arrowheads), indicating ER origin. Gold label was not found in the Golgi apparatus. C, ERVs were abundant structures. Some ERVs are indicated by arrows. D, The PSV was not labeled, but a Golgi stack was slightly labeled. E, The extracellular space was devoid of labeling. cl, Chloroplast; CW, cell wall; EC, extracellular space; G, Golgi apparatus; N, nucleus; NE, nuclear envelope; OB, oil body; SNE, swollen nuclear envelope. Bars = 200 nm (A, B, D, and E) and 1,000 nm (C). C-terminal KDEL tag, which has also been reported to additionally increase the accumulation level (Wandelt et al., 1992;Schouten et al., 1996;Fiedler et al., 1997;Ko et al., 2005;Petruccelli et al., 2006;Laguía-Becher et al., 2010). In our study, the addition of a KDEL tag increased the stability of the protein (especially for the HA78 constructs); however, the accumulation level actually decreased. Our observations suggest that KDEL tagging is not a general technique to increase the accumulation level of recombinant proteins in Arabidopsis seeds. Moreover, the ER-typical oligomannosidic N-glycans on potential pharmaceutical products are only in rare cases appropriate structures (Raju, 2008).
The investigation of N-glycosylation status in combination with the subcellular localization of the recombinant proteins revealed some unexpected outcomes. Seed-expressed 2G12 scFv-Fcs accumulated virtually exclusively in newly formed globular structures, irrespective of their targeting signals. As ribosomes are attached to the delimiting membranes of these newly formed structures, it is anticipated that they are ERVs. This assumption is corroborated by the ER-typical oligomannosidic N-glycosylation pattern of the 2G12 scFv-Fcs. The formation of such ER-derived structures in dicot seeds has been suggested by others (Moravec et al., 2007;Van Droogenbroeck et al., 2007;Schmidt and Herman, 2008;Floss et al., 2009); however, the presented data did not allow drawing an unequivocal conclusion on the origin of these structures.
Particularly, the deposition of the secretory 2G12 scFv-Fcs in ERVs came as a surprise, since it was designed for secretion and thus expected to localize to the extracellular space, as reported for the full-length version of the same antibodies expressed in seeds (Loos et al., 2011). Notably, also the secretory version of 2G12 scFv-Fc expressed in leaves carries ER-typical oligomannosidic N-glycans instead of the expected complex-type N-glycosylation, indicating retention and not secretion. The reasons for this unusual subcellular deposition are not clear yet but could be due to improper folding. The complete absence of antigen-binding and virus-neutralization activity indicates an improper conformation of the polypeptide. A detailed structural analysis of 2G12 IgG had revealed an unusual conformation of this antibody (Calarese et al., 2003), with the V H domain of one heavy chain pairing with the V L domain of the opposing light chain. We hypothesize that the correct threedimensional assembly of this unusual structure is impeded in the scFv-Fc, leading to the observed findings. The 2G12 scFv-Fcs seem to be recognized by the ER-based quality-control systems, and their passage farther down the secretory pathway is prevented. The increased presence of Man7.1 on 2G12 scFv-Fcs further corroborates our hypothesis, as this isoform has been reported to act as a signal for ERAD, an intracellular control mechanism for the elimination of incorrectly folded proteins (Clerc et al., 2009;Liebminger et al., 2010). In contrast, Man7 of secretory HA78 scFv-Fcs consists mainly of the Man7.2 isomer, which does not seem to be involved in the ERAD pathway (Clerc et al., 2009). Surprisingly, also the KDEL-tagged HA78 scFv-Fc has an elevated fraction of Man7.1 isomers. Thus, it seems unlikely that Man7.1 structures alone are a mandatory signal for ERAD in Arabidopsis.
KDEL-tagged HA78 scFv-Fc was detected in newly formed ER-derived vesicles. This deposition pattern is in accordance with the N-glycosylation pattern. However, it is not in full agreement with a similar KDELtagged HA78 scFv-Fc construct, which was previously expressed in Arabidopsis seeds and had accumulated to a large extent in the periplasmic space between the plasma membrane and the cell wall (Van Droogenbroeck et al., 2007). Such a periplasmic space was not detected in this study. Yet, another subcellular localization was found for a KDEL-tagged full-length version of 2G12 recently expressed in Arabidopsis seeds. This protein was targeted nearly exclusively to protein storage vacuoles, and neither formation of ERVs nor a periplasmic space was detected (Loos et al., 2011). The factors responsible for the different deposition patterns of these related KDEL-tagged polypeptides are Figure 9. Antigen-binding activity of the seed-produced HA78 scFv-Fc variants. All seed-produced HA78 scFv-Fc variants exhibit the same antigen-binding activity as a CHO-produced full-length HA78 IgG. OD, Optical density.

Table III. In vitro virus neutralization activity of seed-produced HA78 scFv-Fcs
In vitro virus neutralization efficiencies of the plant-produced HA78 scFv-Fcs as well as of a CHO-produced HA78 mAb are all in the same range and thereby similar to previously reported seed-produced HA78 mAbs and a scFv-Fc (Van Droogenbroeck et al., 2007;Loos et al., 2011). As a negative control, a recombinant human antibody against Hantavirus was used. so far unknown. Possible reasons could be differences in the protein sequences and/or differences in expression level. However, despite their different subcellular depositions (between plasma membrane and cell wall, in ERVs, and PSVs), all these KDEL-tagged proteins carry comparable oligomannosidic N-glycosylation profiles, indicating identical trafficking routes along the early secretory pathway. In contrast to the 2G12 constructs, the secretory versions of HA78 scFv-Fcs ( wt-Ph HA78sc SEC , TKO-Ph HA78sc SEC , and wt-P35S HA78sc SEC purified from leaves) carry the expected complex N-glycans, indicating efficient secretion via the Golgi apparatus, where these modifications occur. Intensive labeling of the extracellular space and the Golgi found in our IEM studies confirms this assumption. The complete absence of plant-specific b-1,2-Xyl and a-1,2-Fuc on the HA78 scFv-Fc expressed in the N-glycosylation mutant TKO ( TKO-Ph HA78sc SEC ) confirms the suitability of this expression host for the production of glycoproteins with a humanized N-glycosylation profile. This is in accordance with our previous studies, where glyco-engineered Arabidopsis served as an expression system for the generation of mAbs with a humanized N-glycosylation profile (Strasser et al., 2004;Schähs et al., 2007;Loos et al., 2011). Glycan structures as obtained in these studies (i.e. GnGn structures) have been shown to enhance Fc-mediated effector functions of mAbs and thus may contribute to the development of pharmaceuticals with enhanced in vivo efficacy (Forthal et al., 2010).
An unexpected observation was the presence of substantial portions of oligomannosidic N-glycans on b-phaseolin-driven secretory HA78 scFv-Fcs. Such ERtypical glycans were hardly found on leaf-produced HA78 scFv-Fc ( wt-35S HA78sc SEC purified from leaf). This observation seems to be due to the special characteristics of the seed secretory system rather than a consequence of the substantially increased accumulation level of the b-phaseolin-driven constructs, as degradation products of 35S-driven secretory HA78 scFv-Fcs ( wt-35S HA78sc SEC ) purified from seeds also carry a substantial portion of oligomannosidic structures besides complex N-glycans (Supplemental Figs. S5 and S6; Supplemental Table S1). However, this increased fraction of ER-typical, oligomannosidic N-glycans is in line with the deposition pattern in mature seeds, where wt-Ph HA78sc SEC is secreted to the extracellular space but is also partially deposited intracellularly in ER-derived structures (Supplemental Fig. S1). This partial retention, which is not found in developing seeds, might occur in an advanced developmental stage. Overloading of the translocation machinery, the onset of senescence, and, related to that, decreased translocation activity or a development-related switch in pathways can all be envisaged to explain this unusual phenomenon. The same difference between immature and mature seeds (i.e. a stronger tendency toward ER-derived structures in later stages of maturation) was also observed in maize (Zea mays) seeds expressing a recombinant protein (Arcalis et al., 2010).
Recently, full-length versions of 2G12 expressed in maize seeds and tobacco (Nicotiana tabacum) seeds have been shown to carry a substantial fraction of single GlcNAc residues Ramessar et al., 2008;Floss et al., 2009). With the exception of some degradation products of 35S-driven HA78 scFv-Fc, this specific glycoform has not been found in this study or in related studies on the expression of antibodies or fragments in Arabidopsis seeds (Van Droogenbroeck et al., 2007;Loos et al., 2011), nor has it been reported upon expression of other mAbs in other plant seeds.
To our knowledge, our results represent the first study that examines in a systematic way the stability, subcellular targeting, and deposition of recombinantly expressed proteins in combination with a detailed monitoring of the N-glycosylation status. This important posttranslational modification occurs along the secretory pathway and allows conclusions on the subcellular trafficking of the target protein. In summary, we demonstrate the efficient expression of recombinant proteins in Arabidopsis seeds, the targeted subcellular deposition, and the generation of customized N-glycosylation patterns. These results emphasize the high potential of plant seeds as reliable and versatile expression hosts for the production of complex mammalian glycoproteins. However, we also reveal some unusual phenomena within the secretory pathway of transgenic seeds, like the formation of ERderived vesicles upon the expression of recombinant proteins. The exact mechanism that initiates the formation of such vesicles is not yet fully understood. Further studies are required to elucidate intracellular trafficking mechanisms in seeds, which are of particular importance for the production of proteins that need special posttranslational modifications, like glycosylation, to obtain full biological activity. So far, this elucidation remains a case-to-case study.

Cloning and Plant Transformation
2G12 V H and V L were amplified with forward primers containing SfiI and ApaLI sites, respectively, and reverse primers containing XhoI and NotI sites, respectively. The fragments were cloned into the pHEN2 phagemid vector In contrast to CHO-and seed-produced 2G12 IgG (Loos et al., 2011), the seed-produced scFv-Fc does not show in vitro neutralization activity.  (Van Droogenbroeck et al., 2007); from there, the scFv was excised via SfiI and NotI sites and cloned in frame into pPICZaB (Invitrogen) containing a short linker fragment and the hinge-C H 2-C H 3 domains of the full-length 2G12 antibody, thereby constructing the 2G12 scFv-Fc. HA78 scFv-Fc was amplified from a scFv-Fc clone described previously (Van Droogenbroeck et al., 2007) with changes in the 5# and 3# ends to be the same as in the full-length HA78 construct (Loos et al., 2011) and to comply with the Kabat catalog.
The signal peptide of the 2S2 albumin of Arabidopsis (Arabidopsis thaliana) as well as the attB sites necessary for Gateway cloning were added by PCR to the scFv-Fcs. Both constructs were cloned with and without C-terminal addition of KDEL at the 3# end before the stop codon. The four constructs were inserted via Gateway cloning (via an intermediate step in a pDONR clone) into the binary plant destination vector pPhasGW (F. Morandini, B. Van Droogenbroeck, A. Depicker, unpublished data) that allows high-level seedspecific expression.
Additionally, the secretory version of both scFv-Fcs were PCR amplified from the pDONR clones, and XbaI and BglII sites were added 5# of the 2S2 signal peptide and 3# of the stop codon, respectively. The fragments were cloned into the XbaI and BamHI sites of the binary plant vector pPT2 (Strasser et al., 2005), where they are under the control of the cauliflower mosaic virus 35S promoter, allowing systemic expression in plants.
Arabidopsis wild-type plants were transformed by floral dip with all six constructs, and Arabidopsis TKO plants (Strasser et al., 2004) were transformed by the same method with the pPhasGW vector containing the secretory versions. Altogether, eight combinations of vector and genetic background were established (Table I).
Transgenic plants were selected on kanamycin containing Murashige and Skoog (Duchefa Biochemie) plant medium plates. The expression level was controlled by sandwich ELISA (coated, goat anti-human IgG [g-chain], Sigma catalog no. I-7883; detected, goat anti-human IgG [g-chain] conjugated to alkaline phosphatase, Sigma, catalog no. A-3187) and confirmed by immunoblotting against the heavy chain. Two lines each were taken to the homozygous state.

Protein Extraction from Leaves
Leaf material was weighed into 2-mL safe-lock, round-bottom Eppendorf tubes, two steel balls (5 mm) were added, and the tubes were frozen in liquid nitrogen and crushed in a ball mill (1 min, 70 2s ). For every 1 mg of leaf material, 10 mL of extraction buffer (40 mM ascorbic acid, 500 mM NaCl, 1 mM EDTA, 45 mM Tris, pH 6.8, with HCl) was added, the samples were incubated for 15 min on ice and centrifuged for 5 min at maximum speed (tabletop centrifuge), and the supernatant was taken off.

Protein Extraction from Seeds and Antibody Purification
Five to 10 mg of seeds and two steel balls were added to appropriate reaction tubes, frozen in liquid nitrogen, and crushed in a ball mill. Fatty components were extracted with 950 mL of hexane. After 5 min of centrifugation at 10,000g, the supernatant was discarded and the samples were vacuum dried. Freezing, crushing, hexane extraction, and drying were repeated, and the pellets were completely resuspended in 900 mL of icecold extraction buffer (40 mM ascorbic acid, 500 mM NaCl, 1 mM EDTA, 45 mM Tris, pH 6.8, with HCl) and incubated on ice for 15 min. The insoluble material was removed by centrifugation in a tabletop centrifuge at 20,000g (4°C) for 5 min. The supernatant was kept on ice while the extraction procedure was repeated. The supernatants were pooled, centrifuged at 35,000g (4°C) for 30 min, and filtered through a 2.5-mm filter. The centrifugation was repeated, and the cleared extract was filtered through a 0.45-mm filter. The filtrate was then applied to a column packed with 1 mL of Protein A Fast Flow Sepharose (GE Healthcare). scFv-Fcs were eluted with 100 mM Gly, pH 3.0, and immediately neutralized with 500 mM Tris, pH 8.0. wt-35S HA78sc SEC purified from seeds was concentrated via an additional Protein A purification step described below.

Purification of scFv-Fcs Expressed under the Control of the 35S Promoter
Leaf extracts of 200 mL or approximately 1 mL of wt-35S HA78sc SEC purified from seeds as described above were mixed with 25 mL of washed Protein A-Sepharose (GE Healthcare catalog no. 17-1279-01) and shaken at 4°C for 120 min. The beads were centrifuged for 5 min at 5,000 rpm (tabletop centrifuge), the supernatant was discarded, and the beads were resuspended in 750 mL of PBST (phosphate-buffered saline [PBS] and 0.1% [v/v] Tween 20) and transferred to a column (Micro Bio-Spin Chromatography Column; Bio-Rad catalog no. 732-6204). The beads were washed at least twice with 700 mL of PBS. Thirty-five microliters of 13 Laemmli loading buffer was added, and the column was sealed, heated for 3 min at 96°C, and centrifuged for 1 min at 1,000g. The flow-through was reapplied and heating and centrifugation were repeated.

N-Glycan Analysis
N-Glycan analysis was carried out by LC-electrospray ionization (ESI)-MS of tryptic glycopeptides . In short, the scFv-Fc bands of Coomassie Brilliant Blue-stained SDS-PAGE gels were excised, protein S alkylated, digested with trypsin, and subsequently analyzed by LC-ESI-MS. Due to incomplete tryptic digest, each glycan structure is represented by two glycopeptides that differ by 482 D (glycopeptide 1, EEQYNSTYR; glycopeptide 2, TKPREEQYNSTYR).
For Man7 isomer analysis, N-glycans were released by peptide:N-glycosidase A or F and purified by passage over a reverse-phase cartridge followed by borohydride reduction and purification over a graphitic carbon cartridge as described previously . The reduced oligosaccharides were separated on a capillary porous graphitic carbon column and detected by LC-ESI-MS (Pabst et al., 2007). Oligomannosidic reference glycans were prepared from Phaseolus vulgaris in a similar manner. Their structure is known from previous NMR analyses (Neeser et al., 1985). The assignments of these reference glycans were corroborated by parallel analysis of 2-aminopyridinelabeled glycans (Tomiya et al., 1991;Liebminger et al., 2009).

Antigen-Binding Activity of Purified scFv-Fcs
Antigen-binding activity was determined by ELISA as described previously (Van Droogenbroeck et al., 2007;Strasser et al., 2008). For HA78 scFv-Fcs, microtiter plates were coated overnight at 4°C with hepatitis A antigen diluted in coating buffer (0.05 M sodium carbonate buffer, pH 9.6). After washing four times, the plates were blocked with 2% skim milk in PBS. One hundred microliters of the purified antibody preparations was added per well in a 2-fold dilution series beginning with 0.5 mg mL 21 and incubated at 37°C for 1.5 h. Wells were washed and incubated with horseradish peroxidase (HRP)-conjugated anti-human IgG Fc (Sigma) antibodies followed by tetramethylbenzidine for development. PBS was used as a negative control. For 2G12 scFv-Fcs, ELISA plates were coated with 100 ng of purified gp120. After washing, 50 mL of a 1:2 serial dilution of the 2G12 scFv-Fc was added per well. As a standard, recombinant 2G12 purified from CHO culture supernatants was used. Plates were incubated for 1 h at room temperature (gentle shaking), and after washing, the wells were incubated with a goat anti-human IgG (k chain) polyclonal antibody conjugated to alkaline phosphatase (1:1,000; Sigma). After intensive washing, alkaline phosphatase was detected by p-nitro-phenylphosphate.

Virus Neutralization Assays
wt-Ph 2G12sc KDEL was tested with an syncytium inhibition assay as described previously (Purtscher et al., 1994) for HIV neutralization efficiency. As the other 2G12 constructs displayed the same N-glycosylation profile, they were not tested. In brief, after preincubating a dilution series of scFv-Fcs with HIV, CD4-positive human AA-2 cells were added. Five days later, the presence of at least one syncytium was evaluated as positive and the antibody concentration was calculated, at which 50% inhibition of the initial syncytium-forming activity was observed (IC 50 ; Reed, 1938). The assays were performed in duplicate.
Hepatitis A virus neutralization was tested as described previously (MacGregor et al., 1983;Van Droogenbroeck et al., 2007). Then, HepA virus strain JN was preincubated with a dilution series of purified scFv-Fcs (Gao, 1989), after which appropriate virus target cells (FRhK-4) were infected. Hepatitis A infection was determined after 21 d of incubation by ELISA with human anti-HepA polyclonal antibody as coating antibody and HRP-conjugated anti-HepA mAb from mouse for detection. Neutralization efficiencies were calculated according to the method of Reed (1938).

IEM
IEM on developing seeds was performed as described previously (Loos et al., 2011). Briefly, developing embryos were high-pressure frozen, freeze substituted with acetone, and infiltrated with HM20. Polymerization was conducted with UV light at 235°C. Sections were labeled with g-chain-specific antibodies and secondary gold-coupled antibodies. After postcontrasting, the samples were analyzed with a JEOL JEM-1400 transmission electron microscope at 80 kV. IEM on mature seeds was performed as described in Supplemental Materials and Methods S1.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Subcellular localization of wt-Ph HA78sc SEC in mature Arabidopsis seeds by immunogold electron microscopy.
Supplemental Figure S2. Subcellular localization of wt-Ph HA78sc KDEL in mature Arabidopsis seeds by immunogold electron microscopy.
Supplemental Figure S3. Subcellular localization of wt-Ph 2G12sc SEC in mature Arabidopsis seeds by immunogold electron microscopy.
Supplemental Figure S4. Subcellular localization of wt-Ph 2G12sc KDEL in mature Arabidopsis seeds by immunogold electron microscopy.
Supplemental Figure S6. N-glycosylation pattern of the degradation products of seed-produced wt_35S HA78sc SEC .
Supplemental Table S1. Relative distribution of N-glycans of the degraded fragments of seed-extracted wt-35S HA78sc SEC , wt-Ph HA78sc SEC , and TKO-Ph HA78sc SEC .
Supplemental Materials and Methods S1. IEM of mature seeds.
Supplemental Results S1. Subcellular localization in mature seeds and N-glycosylation of degradation products.