Heterologous expression of influenza haemagglutinin leads to early and transient activation of the unfolded protein response in Nicotiana benthamiana

Summary The unfolded protein response (UPR) allows cells to cope with endoplasmic reticulum (ER) stress induced by accumulation of misfolded proteins in the ER. Due to its sensitivity to Agrobacterium tumefaciens, the model plant Nicotiana benthamiana is widely employed for transient expression of recombinant proteins of biopharmaceutical interest, including antibodies and virus surface proteins used for vaccine production. As such, study of the plant UPR is of practical significance, since enforced expression of complex secreted proteins often results in ER stress. After 6 days of expression, we recently reported that influenza haemagglutinin H5 induces accumulation of UPR proteins. Since up‐regulation of corresponding UPR genes was not detected at this time, accumulation of UPR proteins was hypothesized to be independent of transcriptional induction, or associated with early but transient UPR gene up‐regulation. Using time course sampling, we here show that H5 expression does result in early and transient activation of the UPR, as inferred from unconventional splicing of NbbZIP60 transcripts and induction of UPR genes with varied functions. Transient nature of H5‐induced UPR suggests that this response was sufficient to cope with ER stress provoked by expression of the secreted protein, as opposed to an antibody that triggered stronger and more sustained UPR activation. As up‐regulation of defence genes responding to H5 expression was detected after the peak of UPR activation and correlated with high increase in H5 protein accumulation, we hypothesize that these immune responses, rather than the UPR, were responsible for onset of the necrotic symptoms on H5‐expressing leaves.


Introduction
In eukaryotes, many proteins rely on the cell secretory pathway for proper maturation, targeting, and biological functions (Benham, 2012).These include secreted proteins, plasma membrane (PM) proteins, and proteins targeted to lysosomes (and the vacuole in plant cells).To enter the secretory pathway, proteins are first translated on membrane-bound ribosomes located on the external face of the ER.Nascent polypeptides then reach luminal space of the ER via translocons that act as protein channels spanning the ER membrane.Inside the ER lumen, the so-called 'client' proteins are rapidly subjected to chaperone-assisted folding and, when required, to posttranslational modifications such as disulfide bridge formation, glycosylation, or oligomer assembly.To coordinate these activities, eukaryotic cells evolved a sophisticated protein machinery known as the endoplasmic reticulum quality control (ERQC) system (Araki and Nagata, 2011).While protein disulfide isomerases (PDIs) catalyse the formation of disulfide bonds, ER-resident chaperones of the binding immunoglobulin protein (BiP) family bind to exposed hydrophobic regions of unfolded proteins, helping them to adopt a proper tertiary structure.For glycoproteins, folding assistance is further provided by ER-resident lectins of the calreticulin (CRT) and calnexin (CNX) families.
Endoplasmic reticulum quality control allows for proper folding and modification of most client proteins, which in turn proceed towards downstream steps of the secretory pathway for further maturation, suitable targeting, or secretion.Despite efficacy of the ERQC system, some client proteins still fail to fold or assemble properly.When accumulation of misfolded proteins reaches a certain threshold, a second mechanism known as the ERassociated degradation (ERAD) system is activated (Araki and Nagata, 2011).ERAD allows for the retro-transport of misfolded proteins from the ER lumen to the cytosol, where degradation through the 26S proteasome will intervene.Distinct ERAD pathways exist (Brodsky and Wojcikiewicz, 2009;Wu and Rapoport, 2018); however, the removal of soluble and ER membrane glycoproteins is mediated by luminal lectin osteosarcoma 9 (OS9; H€ uttner et al., 2012), which recognizes trimmed Nglycan chains of glycoproteins that have failed to fold properly (Liu and Howell, 2010).Together with the ER-resident chaperone glucose-regulated protein 94 (GRP94), OS9 brings misfolded proteins to the retrotranslocation complex, which comprises ER membrane proteins such as Sel1L, HRD1, and Derlins (DERs).The retrotranslocation complex also comprises ATPase motor protein CDC48, which extracts misfolded proteins from the ER lumen and releases them in the cytosol.E3 ubiquitin ligase activity of HRD1 then conjugates ubiquitin moieties to the target protein, marking it for degradation.
While ERQC and ERAD constantly function to maintain cell homeostasis, changes in environmental conditions or cell physiological status can increase the needs for protein secretion.These changes can also induce conditions that are no longer favourable for folding and maturation of ER proteins.When accumulation of misfolded proteins overwhelms basal ERQC and ERAD functions, cells start to experience ER stress.To cope with ER stress, eukaryotes have evolved refined signalling networks collectively known as the unfolded protein response (UPR; Read and Schr€ oder, 2021).In plants, the UPR comprises two branches with conserved components and activation mechanisms (Duwi Fanata et al., 2013;Howell, 2013Howell, , 2021;;Iwata and Koizumi, 2012;Liu and Howell, 2010).In the first branch, membrane-tethered transcription factors (TFs) basic leucine zipper 17 (bZIP17) and bZIP28 are released from membrane anchoring via clipping of their transmembrane domain (TMD).This process is mediated by proteases associated with the Golgi apparatus (Liu et al., 2007).In the second branch, the ER transmembrane sensor inositol requiring enzyme 1 (IRE1) uses its ribonuclease activity to unconventionally splice bZIP60 transcripts, producing a shorter TF that lacks a TMD and therefore is no longer restrained by ER membrane anchoring (Nagashima et al., 2011).Once activated, the UPR triggers a series of complementary mechanisms, including the shutdown of translational activities and the up-regulation of UPR genes.These include ERQC components such PDIs, BiPs, CNXs, and CRTs, as well as genes encoding ERAD components (Iwata et al., 2008(Iwata et al., , 2010;;Kamauchi et al., 2005).The purpose of the UPR is to restore cell homeostasis by reducing translation on one hand, and increasing ERQC and ERAD capabilities on the other hand.In case of severe ER stress, sustained activation of the UPR can also lead to the activation of programmed cell death, a mechanism that ultimately protects highly stressed tissues from cells that have become dysfunctional (Kørner et al., 2015).
Plant molecular farming collectively refers to approaches that make use of plant cells as biofactories to produce recombinant proteins or metabolites of biopharmaceutical interest (Chung et al., 2022).Recombinant proteins commonly produced in planta include therapeutic antibodies as well as surface proteins from mammalian viruses that are used for the production of vaccines.Using Nicotiana benthamiana leaf cells, the biopharmaceutical company Medicago has for instance developed a molecular farming approach to produce influenza vaccine candidates at large scale (D'Aoust et al., 2008;Landry et al., 2010).Based on the bacterial vector Rhizobium radiobacter (commonly known and hereafter referred to as Agrobacterium tumefaciens), this process relies on the transient expression of recombinant influenza haemagglutinins (HAs), along with the viral suppressor of RNA silencing P19 (Silhavy et al., 2002) that prevents silencing of recombinant HA genes delivered by the bacterium.Engineered to efficiently enter the ER, newly synthesized HA proteins travel through the plant cell secretory pathway, before being trafficked to the PM (D'Aoust et al., 2008).When sufficient HA proteins have accumulated, bending of the PM allows for budding of the so-called virus-like particles (VLPs).These nanoscale assemblies comprise trimer clusters of the engineered HA protein embedded in a lipid envelope derived from the PM of plant cells.Structurally, VLPs and influenza viruses share similar size and shape; however, the former lack other viral proteins such as the surface exposed neuraminidase, in addition to being devoid of the genetic components required for replication.Once purified and formulated into vaccine candidates, VLPs induce an immune response that protects newly immunized hosts from subsequent infection by the influenza virus (Landry et al., 2010).
Using influenza HA protein H5, we recently reported that expression of VLPs results in a unique molecular signature that affects metabolism and fitness of plant cells at 6 days postinfiltration (DPI; Hamel et al., 2023).In addition to the shutdown of chloroplast gene expression and to the activation of immune pathways, proteomics revealed that H5 expression results in the accumulation of UPR proteins, including PDIs, BiPs, and CRTs.Enforced expression of a complex secreted protein such as H5 was expected to trigger the UPR so that plant cells can manage stress associated to the transient expression system, including increased needs for both recombinant and endogenous defence protein secretion.Interestingly, thorough monitoring of the transcriptome at 6 DPI did not reveal induction of UPR genes, suggesting that increased accumulation of UPR proteins was either independent of transcriptional regulation, or that UPR gene up-regulation occurred earlier, before returning to levels that prevented their detection at 6 DPI (Hamel et al., 2023).
Using time course sampling, we here show that H5 expression does result in early and transient activation of the UPR, as inferred from the detection of unconventional splicing of NbbZIP60 transcripts and the up-regulation of UPR genes with varied functions.Up-regulation of UPR genes was detectable at 3 DPI, but expression of these genes had returned to basal level after 5 days of expression.Our data also showed that activation of the UPR peaked prior to the induction of defence genes that strongly respond to H5 protein expression.Transient nature of the H5induced UPR suggests that enhanced ERQC and ERAD functions were sufficient to cope with ER stress imposed by expression of the HA protein, as opposed to the expression of an antibody that induced stronger and more sustained activation of the UPR.Overall, this work expands our understanding of host-plant responses to foreign protein expression, in addition of providing the research community with a useful set of marker genes to study ER stress and associated UPR in N. benthamiana, a model plant and host of choice for molecular farming and study of plant immunity (Bally et al., 2018;Chung et al., 2022;Goodin et al., 2008;Ranawaka et al., 2023).Conserved molecular mechanisms linked to the activation of UPR in N. benthamiana are further discussed.

Accumulation of UPR proteins in response to H5 expression
To better define molecular responses in N. benthamiana leaves expressing influenza VLPs, we previously conducted a proteomics survey using isobaric tags for relative and absolute quantitation (iTRAQ) labelling (Hamel et al., 2023).At 6 DPI, this revealed enhanced accumulation of UPR proteins, including PDIs, BiPs, and CRTs (Table 1).Since the UPR plays a central role in protein secretion, we further investigated regulation of this pathway during foreign protein expression.Search of the N. benthamiana genome identified 21 PDIs (NbPDIs) that clustered similarly to PDIs from the model plant species Arabidopsis thaliana (AtPDIs; Figure S1a).A list providing identification numbers from all UPR genes identified is available in the Supporting information section (Table S1).Of the seven NbPDIs identified by proteomics, all but NbPDI16 and NbPDI17 displayed a C-terminal 'KDEL' motif

NbCRT1
Using iTRAQ labelling, changes in protein abundance were evaluated in mock-infiltrated leaves, leaves expressing P19 only, or leaves co-expressing P19 and the HA protein from pandemic influenza virus strain H5 Indonesia (H5/A/ Indonesia/05/2005; Hamel et al., 2023).Using mock samples as a control, pairwise comparisons were performed for P19 and H5 samples.From the resulting lists of up-regulated proteins, UPR proteins were identified.For all conditions, sampling was performed at 6 DPI.To be considered significantly up-regulated, proteins had to fulfil the following criterium for at least one of the pairwise comparisons: Log2FC  1), which works as an ER retention signal (Munro and Pelham, 1987).Interestingly, identified NbPDIs also matched with the subset of AtPDIs previously involved in the UPR (Lu and Christopher, 2008), including NbPDI16 and NbPDI17 that lack an ER retention signal (Figure S1a).Search of the N. benthamiana genome also identified seven BiPs (NbBiPs; Figure S1b), and nine ER lectins that clustered in two sub-types corresponding to CRTs (NbCRTs) and CNXs (NbCNXs; Figure S1c).For these ER-resident chaperones, clustering was again similar to corresponding homologues in Arabidopsis, emphasizing conservation of ERQC components between the two plant species.For all NbBiPs and NbCRTs identified by proteomics, a C-terminal 'HDEL' motif was identified, again suggesting retention in the ER (Table 1).
Our previous transcriptomics and proteomics investigation at 6 DPI had also revealed up-regulation of cytosolic heat shock proteins (HSPs), a response seen following expression of P19 only, but not following co-expression of P19 and H5 (Hamel et al., 2023).One noticeable exception to this was the gene model Niben101Scf04331g09018, which encodes a molecular chaperone of the HSP90 family.Unlike cytosolic HSPs mentioned above, the protein product from this gene harbours a C-terminal 'KDEL' motif that suggests ER localization.The protein also showed enhanced accumulation following both P19 expression, and co-expression of P19 and H5 (Table 1).To the best of our knowledge, this HSP90 has never been formally characterized in N. benthamiana, although it is annotated as a putative endoplasmin homologue in the improved genome assembly recently published for this plant species (Ranawaka et al., 2023).Protein sequence alignments showed that this particular HSP90 homologue shares high homology to Arabidopsis SHEPHERD (AtSHD; AT4G24190), an ER-resident chaperone involved in the folding of CLAVATA proteins that regulate meristem growth (Ishiguro et al., 2002).Product of the Niben101Scf04331g09018 gene is also closely related to the human protein GRP94 (also known as endoplasmin), a key ERAD component that interacts with OS9 to deliver misfolded proteins to the retrotranslocation complex (Christianson et al., 2008;Marzec et al., 2012;Seidler et al., 2014).Herein termed NbGRP94 (Table 1), overaccumulation of that ER chaperone further suggests that the UPR was activated during H5 protein secretion.We hypothesized that the UPR contributes to efficient expression of the HA protein, in addition of preventing negative effects of the ER stress perhaps induced during enforced expression of the secreted protein.

Time course sampling, stress symptoms, and recombinant protein accumulation
To assess whether the accumulation of UPR proteins was dependent on early up-regulation of the corresponding UPR genes, time course sampling was performed at 1, 3, 5, 7, and 9 DPI.For each time point, leaves from non-infiltrated (NI) plants, or plants only infiltrated with the resuspension buffer (Mock) were used as controls (absence of treatment and effects of the mechanical stress caused by infiltration without the Agrobacterium respectively).As the effects of Agrobacterium infiltration without recombinant protein expression was previously shown to largely overlap with the effects of Agrobacterium-mediated expression of P19 only (Hamel et al., 2023), sampling was also performed on agroinfiltrated leaves solely expressing the viral suppressor of RNA silencing.As such, combined effects of Agrobacterium infiltration and P19 expression served as a control for the expression of products with biopharmaceutical interest.For VLPs, leaves co-expressing P19 along with the HA protein of pandemic influenza virus strain H5 Indonesia (H5/A/Indonesia/05/2005; H5 Indo ) were employed.Hereafter, this recombinant protein will be referred to as H5 (Hamel et al., 2023).Regulation of the UPR was also assessed during co-expression of P19 along with two monoclonal antibodies (mAbs) termed mAb1 and mAb2.Within plant cells, fully assembled antibodies are produced following secretion of antibody light and heavy chains; however, the expression of these proteins does not result in the formation of VLPs.mAb1 and mAb2 were chimeric antibodies that thus comprised variable regions of murine origin and constant regions of human origin.Both antibodies belonged to the gamma (c) class and were of the same subclass (IgG1).In both cases, light chains belonged to the kappa (j) type.In other words, primary sequences from mAb1 and mAb2 were exactly the same within constant regions of the light and heavy chains, while variable regions of the light and heavy chains were different between the two antibodies.In addition to their high proximity at the primary protein sequence level, mAb1 and mAb2 were selected based on (1) the contrasted stress symptoms they induced on the host plants, and (2) their contrasting accumulation levels, either high or low, in leaf tissue following heterologous expression (see below).
The effects of foreign protein expression were first characterized by macroscopic evaluation of the stress symptoms induced on representative leaves from each condition harvested at 9 DPI (Figure 1a).Using NI leaves as a baseline, no obvious effect was visible on mock-infiltrated leaves.For agroinfiltrated leaves expressing P19 only, yellowish discoloration typical of chlorosis was visible, but no sign of plant cell death was denoted.For agroinfiltrated leaves co-expressing P19 and H5, chlorosis was more advanced and accompanied by diffused greyish necrotic flecking that spread uniformly throughout the leaf blade (see the magnified H5 leaf section; Figure 1a).For agroinfiltrated leaves co-expressing P19 and mAb1, chlorosis was also observed, but again, no sign of plant cell death was denoted.On the opposite, advanced plant cell death was detected on agroinfiltrated leaves co-expressing P19 and mAb2.Notably, necrotic lesions found on these leaves were somewhat different from those seen on H5expressing leaves, with well-delimited cell death lesions spanning the entire leaf width (see the magnified mAb2 leaf section; Figure 1a).
To confirm expression of suitable recombinant genes within harvested biomass, real-time quantitative polymerase chain reaction (RTqPCR) was performed using primers specific to P19 (Figure S2a), H5 (Figure S2b), antibody heavy chain (mAb HC; Figure S2c), and antibody light chain (mAb LC; Figure S2d).For every gene and time point examined, results confirmed the absence of recombinant gene expression in NI and Mock samples.For every recombinant gene tested, results also indicated that transcript accumulation had not started at 1 DPI.For recombinant gene P19, expression was detected in P19, H5, mAb1, and mAb2 samples, with higher expression levels when the P19 gene was expressed alone.For most conditions, P19 expression peaked around 3-5 DPI, before going down gradually at later time points (Figure S2a).Consistent with predicted strength of the promoters used to drive recombinant gene expression (plastocyanin promoter for P19 and 2X35S promoter for other recombinant genes), P19 was the least expressed when compared to H5, mAb HC, and mAb LC (compare scales from the four panels in Unfolded protein response in plant cell protein biofactories 1149 Figure S2).For recombinant gene H5, expression could only be detected in H5 samples (Figure S2b).H5 expression rapidly picked up between 1 and 3 DPI, before reaching a peak at 5 DPI.As seen for P19 (Figure S2a), H5 expression then gradually decreased at 7 and 9 DPI (Figure S2b).For recombinant genes mAb HC (Figure S2c) and mAb LC (Figure S2d), expression could only be detected in mAb1 and mAb2 samples.For both genes and for every time point examined, expression in mAb1 samples was higher than in mAb2 samples.During the whole sampling period, both genes displayed overall similar expression profiles; however, mAb LC was expressed at much higher levels compared to mAb HC (compare scales from Figure S2c,d).Taken together, RTqPCR results confirmed that harvested leaf biomass expressed the proper combinations of recombinant genes, allowing monitoring of recombinant protein accumulation.
To confirm accumulation of the H5 protein, total protein extracts were generated from H5 samples.Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by Coomassie blue staining confirmed integrity of the proteins from all extracts (upper panel; Figure 1b).As the expression phase progressed, results also revealed gradual reduction in the levels of the Ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO), with both RuBisCO subunits similarly affected (upper panel; Figure 1b).This was consistent with what was previously reported when expressing H5 (Hamel et al., 2023).From the same protein extracts, a Western blot was then performed using an antibody specific to the HA protein of influenza virus strain H5 Indonesia (a-H5 Indo ; lower panel; Figure 1b).In H5 samples, no accumulation of the H5 protein was detected at 1 DPI, while low accumulation was seen at 3 DPI.Accumulation of the H5 protein then notably increased between 3 and 5 DPI, and again between 5 and 7 DPI.At that point, H5 accumulation had pretty much reached a plateau, as the level observed at 9 DPI was roughly the same as the one seen at 7 DPI (lower panel; Figure 1b).To assess HA activity, haemagglutination (HMG) assays were conducted on H5 samples (Figure 1c).At 1 DPI, HA activity was barely detectable, consistent with the absence of measurable H5 product (lower panel; Figure 1b).HA activity then notably increased between 1 and 7 DPI, to reach a plateau maintained up to 9 DPI (Figure 1c).Overall, H5 accumulation and HA activity thus correlated tightly, confirming that the recombinant protein H5 accumulated in planta was still active against red blood cell receptors following its extraction from the leaf tissues.Using H5 samples harvested at 3, 5, 7, and 9 DPI, VLPs were then partially purified as described previously (D'Aoust et al., 2008).Transmission electron microscopy (TEM) of the purified products confirmed presence of complex structures corresponding to VLPs in both size and morphology, including a lipid membrane covered with spikes that closely resemble those of true influenza virions (Figure 1d).
To monitor the accumulation of recombinant antibodies, total protein extracts were prepared from mAb1 and mAb2 samples.Following SDS-PAGE under non-reducing conditions, Coomassie blue staining revealed higher accumulation of mAb1 compared to mAb2, an observation true for every sampling point (Figure 1e).Using stained gels and standard curves made from commercially available preparation of purified IgG1, gel densitometry measurements were next performed on assembled antibodies produced in planta.By arbitrarily setting the accumulation of mAb1 at 3 DPI to onefold, relative fold-change (FC) accumulation of both antibodies was determined at each time point (Figure 1f).For mAb1, no accumulation was detected at 1 DPI, while low but measurable levels were detected at 3 DPI.Levels of the antibody then substantially increased between 3 and 5 DPI, reaching over a fourfold increase compared to the established baseline at 3 DPI.The accumulation rate of mAb1 then decreased between 5 and 9 DPI; yet, the peak of antibody accumulation was not reached until the end of sampling at 9 DPI (Figure 1f).For mAb2, significant product accumulation could not be detected until 5 DPI (Figure 1e,f).The level of mAb2 then slightly increased at 7 and again at 9 DPI, but overall accumulation remained much lower compared to mAb1 (Figure 1e,f).Considering these poor accumulation levels and the strong stress symptoms induced by expression of mAb2 in planta (Figure 1a), we hypothesized that this product leads to an unresolved ER stress and therefore that regulation of UPR gene expression within these samples would be informative for comparison with P19, H5, and mAb1 samples.

Activation of the IRE1-bZIP60 pathway
In the absence of ER stress, IRE1 is held in a monomeric, nonsignalling state through interaction with luminal BiPs (Figure 2a).When misfolded proteins start to accumulate, BiPs get recruited for protein folding, freeing the luminal domain of IRE1 for interaction with misfolded proteins that are accumulating.Following oligomerization and phosphorylation, IRE1 gets activated and using its cytosolic RNase domain, it performs unconventional splicing of bZIP60 transcripts.The shorter TF that is produced lacks a TMD, and as a result, it translocates to the nucleus to up-regulate expression of UPR genes that carry specific cis regulatory elements in their promoter (Figure 2a; Liu and Howell, 2016;Li and Howell, 2021).
To investigate IRE1 activation and unconventional splicing of bZIP60 transcripts, we retrieved nucleotide sequences of Niben101Scf24096g00018, the gene model encoding N. benthamiana's version of the UPR-related TF bZIP60.Herein termed NbbZIP60, this gene displayed strict sequence conservation around predicted splicing sites recognized by IRE1 (Figure S3).Based on this sequence conservation, forward primers specific to the spliced (s) or unspliced (u) versions of NbbZIP60 transcripts were designed (Figure S3).Paired to a reverse primer that recognized both transcript versions, forward primers were then used to selectively profile expression of NbbZIP60u and NbbZIP60s via RTqPCR.For NbbZIP60u, results showed similar expression levels for all conditions at 1 DPI (Figure 2b).For NI and Mock samples, expression levels of NbbZIP60u remained low and unchanged up to 9 DPI.For H5 and mAb2 samples, upregulation of NbbZIP60u was seen at 3 DPI and extent of the response was similar for both conditions.In H5 samples, expression of NbbZIP60u slightly increased between 3 and 5 DPI, before going down slightly at 7 and 9 DPI.For mAb2 samples, expression level of NbbZIP60u also increased between 3 and 5 DPI, but expression levels then remained high up to the end of sampling at 9 DPI.Compared to NI and Mock controls, upregulation of NbbZIP60u was also observed in P19 and mAb1 samples, but this response arose later (5 DPI) and remained weaker compared to H5 and mAb2 samples (Figure 2b).
For NbbZIP60s, overall expression levels were lower than those observed for NbbZIP60u (compare scales from Figure 2b,c).This suggests that only a fraction of overall NbbZIP60 transcripts were unconventionally spliced by IRE1.At 1 DPI, very few spliced transcripts were detected, regardless of the conditions (Figure 2c).Significantly higher levels of NbbZIP60s transcripts were however detected for H5 and mAb2 samples at 3 DPI, with significantly higher levels for the latter.For both conditions, levels of spliced transcripts then slowly decreased between 3 and 9 DPI, with levels from mAb2 samples again remaining significantly higher compared to other conditions, including H5 samples.At 5 DPI, H5 samples still displayed significantly higher levels of spliced transcripts compared to P19 samples, but this was no longer the case at 7 and 9 DPI (Figure 2c).Unfolded protein response in plant cell protein biofactories 1151 throughout the time course.Overall, our data suggest early and transient activation of IRE1 in H5 samples, while early and more sustained activation of IRE1 was seen in mAb2 samples.Compared to NI and mock controls, IRE1 splicing activity was also detected in P19 and mAb1 samples, although arising later and remaining weaker compared to H5 or mAb2 samples.

Up-regulation of ERQC genes
Accumulation of NbbZIP60s transcripts (Figure 2c) suggests the UPR to be activated with different strength and kinetics, depending on the foreign protein combinations that were expressed.Activation of the UPR generally leads to the upregulation of genes encoding ERQC components, including PDIs and ER-resident chaperones of the BiP, CRT, and CNX families (Figure 2a; Iwata et al., 2008Iwata et al., , 2010;;Kamauchi et al., 2005).Based on proteomics results (Table 1) and homology with homologues from Arabidopsis (Figure S1), primers specific to selected ERQC genes were designed to perform RTqPCR.For NI and mock samples, expression of ERQC genes remained low and essentially unchanged throughout expression (Figure 3a for NbPDIs, Figure 3b for NbBiPs, Figure 3c for NbCRTs, and Figure 3d for NbCNXs).For the other conditions, slight differences between respective gene expression profiles were identified; however, three major trends emerged.An early and transient expression pattern was first observed for H5 samples.This pattern was characterized by the up-regulation of ERQC genes at 3 DPI, followed by expression levels that rapidly decreased to reach levels similar to those observed in P19, mAb1, or even NI and Mock control samples at 5 DPI.The second expression pattern was observed in mAb2 samples and it was also characterized by the early up-regulation of ERQC genes at 3 DPI.Unlike the first pattern, ERQC gene expression was however sustained at later time points, with levels that never dropped back to basal level and remained significantly higher compared to the other conditions.The third expression pattern, which was observed in P19 and mAb1 samples, was characterized by weaker up-regulation of ERQC genes at 3 DPI.At later time points, the expression of ERQC genes was above the expression levels observed for NI and mock controls, however expression always remained significantly lower compared to H5 and mAb2 samples.Overall, these observations suggest that at 3 DPI, the UPR was more strongly activated in H5 and mAb2 samples compared to other conditions.For some Unfolded protein response in plant cell protein biofactories 1153 ERQC genes, they also suggest a higher response level in mAb2 samples compared to H5 samples (e.g.NbBiP2a/b and NbBiP3a/b; Figure 3b).Expression patterns from all of the ERQC genes examined also tightly correlated with accumulation patterns of NbbZIP60s (Figure 2c), suggesting that these UPR genes are direct genetic targets of the TF.

Up-regulation of ERAD genes
In Arabidopsis, activation of the UPR also promotes the expression of ERAD genes (Iwata et al., 2008(Iwata et al., , 2010;;Kamauchi et al., 2005).These include Sel1L, HRD1, DER2, and DER3, which all encode components of the retrotranslocation complex (Figure 4a).To study ERAD gene expression during foreign protein accumulation, we searched the N. benthamiana genome to identify ERAD gene homologues.Expression from some of the retrieved candidates was then profiled using RTqPCR.Selected genes were NbSel1La and NbSel1Lb (Figure 4b), NbHRD1a (Figure 4c), as well as the closely related NbDER2 and NbDER3 (Figure 4d).At 1 DPI, respective expression levels from all of these genes were similar in all conditions.At 3 DPI, ERAD gene expression significantly increased to similar levels in H5 and mAb2 samples, while P19 and mAb1 samples had similar expression levels compared to NI and Mock controls.At 5 DPI, ERAD gene expression from H5 samples had decreased importantly to reach levels that were similar to P19 and mAb1 samples.After 3 DPI, ERAD gene expression from mAb2 samples also tended to decrease, but it Figure 4 The ERAD system and expression of ERAD genes.(a) Model depicting ER retrotranslocation and cytosolic degradation of a soluble glycoprotein that is irrevocably misfolded.After trimming of its N-glycan chains, the glycoprotein is recognized by luminal lectin OS9.Together with GRP94, a molecular chaperone of the HSP90 family, OS9 brings the misfolded protein to the retrotranslocation complex.The latter comprises ER membrane proteins Sel1L, HRD1, and Derlins (DERs).Through ATPase motor CDC48, the misfolded glycoprotein is extracted from the ER lumen and released in the cytosol.E3 ubiquitin ligase activity of HRD1 then conjugates ubiquitin (Ub) moieties to the misfolded protein, targeting it for degradation via the 26S proteasome.Adapted from Howell, 2021.Expression of genes encoding Sel1L (b), HRD1 (c), or DER (d) proteins, as measured by RTqPCR.For each time point in days post-infiltration (DPI), results are expressed in numbers of molecules per nanogram of RNA.Condition names are as follows: NI, non-infiltrated leaves; Mock, leaves infiltrated with buffer only; P19, agroinfiltrated leaves expressing P19 only; H5, agroinfiltrated leaves co-expressing P19 and H5; mAb1 (mAb2), agroinfiltrated leaves co-expressing P19 and monoclonal antibody 1 (monoclonal antibody 2).
ª 2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163 always remained significantly higher compared to the other conditions.Overall expression patterns from the ERAD genes examined were somewhat similar to those seen previously for NbbZIP60s (Figure 2c) and ERQC genes (Figure 3).
Other UPR genes: Subunits of the Sec61 translocon and Bax inhibitor-1 Translocons are protein complexes that transport nascent proteins from their translation site in the cytosol to the ER lumen.These structures comprise several subunits, including the heterotrimeric complex Sec61, which forms the translocon pore and is made from assembly of secretory (Sec) proteins Sec61a, Sec61b, and Sec61c (Itskanov and Park, 2023).As a mean to favour protein secretion under stress conditions, genes encoding Sec proteins were found to be induced during the UPR (Iwata et al., 2008(Iwata et al., , 2010;;Kamauchi et al., 2005).Search of the N. benthamiana genome identified three Sec61a genes, six Sec61b genes, and five almost identical Sec61c genes (Figure S4a).To examine expression of some of these candidates, RTqPCR primers were designed to target NbSec61a-1 (Figure S4b), as well as closely related NbSec61b-1 and NbSec61b-2 (Figure S4c).Similar to ERQC and ERAD genes, results showed enhanced transcript accumulation that started at 3 DPI, a response more importantly induced in H5 and mAb2 samples compared to P19 and mAb1 samples.For mAb2 samples, NbSec61 gene expression then remained high until the end of sampling at 9 DPI.For H5 samples, expression of these genes on the other hand rapidly decreased between 3 and 5 DPI, reaching levels similar to those observed in P19 and mAb1 samples.No gene induction was seen in NI and mock control samples.
Genes of the Bax inhibitor-1 (BI-1) family encode ER transmembrane proteins first identified as suppressors of the cell death induced by pro-apoptotic protein Bax from yeast and mammals (H€ uckelhoven, 2004).In severe ER stress cases, BI-1 protein levels are thought to act as some sort of a rheostat controlling activation of programmed cell death, including in plants (Ishikawa et al., 2011).This function perhaps explains why BI-1 genes are also induced during the UPR, including in plants (Kamauchi et al., 2005;Iwata et al., 2008Iwata et al., , 2010)).Search of the N. benthamiana genome identified at least four BI-1 homologues (NbBI-1 s), which were most closely related to AtBI-1a of Arabidopsis (AT5G47120; Figure S4d).Primers specific to NbBI-1a and NbBI-1b showed these genes to be induced at 3 DPI in all agroinfiltrated conditions compared to NI and mock controls.However, up-regulation levels were significantly higher in H5 and mAb2 samples compared to P19 and mAb1 samples (Figure S4e).At later time points, higher expression seen in H5 samples reverted to levels similar to those seen in P19 and mAb1 samples, while expression remained significantly higher in mAb2 samples.Overall, these results confirmed that the expression of UPR genes of varied functions was induced in H5 and mAb2 samples, and again that these expression profiles mirrored the expression profile of NbbZIP60s (Figure 2c).

Kinetics of defence gene expression versus activation of the UPR
At 6 DPI, H5 protein expression was shown to induce a number of plant immune responses, including the up-regulation of oxylipin regulatory and response genes (Hamel et al., 2023).As these observations were made at a single time point, we here used RTqPCR to profile expression from some of these defence genes and compared regulation of immune responses with activation timeline of the UPR.For oxylipin regulatory genes, we monitored expression of the patatin gene NbPAT1, of the 9-lipoxygenase gene NbLOX1, of divinyl ether synthase (DVE), or epoxy alcohol synthase (EAS) genes NbCYP74a and NbCYP74b, and of the allene oxide synthase gene NbAOS1 (Hamel et al., 2023).For NbPAT1 (Figure 5a), NbLOX1 (Figure 5b), and NbCYP74 genes (Figure 5c), enhanced gene expression was mostly seen in H5 and mAb2 samples, with higher expression levels detected in the former case.Up-regulation of these genes also arose faster in H5 samples, with significantly enhanced expression detected at 3 DPI.To a much lesser extent, up-regulation of these genes was also detected in P19 and mAb1 samples compared to NI and mock controls.This was consistent with the reported low induction of these genes by Agrobacterium (Hamel et al., 2023).For NbAOS1, weak gene up-regulation was detected in mAb2 samples at 7 and 9 DPI (Figure 5d).Again, gene induction however started earlier (5 DPI) and reached much higher levels in H5 samples.
For oxylipin response genes, expression patterns of the plant defensin gene NbPDF1 (Figure 5e) and of the Kunitz trypsin inhibitor gene NbKTI3 (Figure 5f) were examined.As the candidates selected above, these defence genes were previously shown to be highly responsive to H5 protein expression (Hamel et al., 2023).Consistent with the induction of oxylipin regulatory genes, RTqPCR revealed strongest up-regulation in H5 samples.In both cases, enhanced expression was not detected at 3 DPI, but was obvious at 5 DPI (Figure 5e,f).Expression levels from both genes then continued to increase until the end of sampling at 9 DPI.At 7 and 9 DPI, the two genes were also up-regulated in P19 and mAb2 samples, but expression levels were significantly lower compared to H5 samples.Together, these results indicate that up-regulation of oxylipin regulatory genes preceded upregulation of oxylipin response genes.For H5 samples, these results also indicate that activation of the oxylipin pathway occurred after activation of the UPR, which in this condition peaked at 3 DPI.
At 6 DPI, transcriptomics analyses had also shown strong, and in many cases, H5-specific induction of genes promoting the activation of oxidative stress, including in the apoplast where VLP accumulation takes place (Hamel et al., 2023).In view of this, we profiled the expression from some of these genes, namely the NADPH oxidase gene NbRBOHd, closely related polyphenol oxidase genes NbPPO1 and NbPPO3, secreted carbohydrate oxidase gene NbBBE2, as well as secreted ascorbate oxidase genes NbAO1 and NbAO2 (Hamel et al., 2023).At 1 DPI, results indicated a similar up-regulation of NbRBOHd in mock and agroinfiltrated samples compared to NI samples (Figure 6a).A comparable expression pattern was observed at 3 DPI but not at 5 DPI, as NbRBOHd was expressed at significantly higher level in H5 samples compared to the other agroinfiltrated conditions.In H5 samples, expression of NbRBOHd then decreased at 7 and 9 DPI, yet it remained significantly higher compared to P19, mAb1, or mAb2 samples.For NbPPO genes (Figure 6b) and NbBBE2 (Figure 6c), RTqPCR revealed similar expression profiles, with stronger up-regulation again seen in H5 samples.In the latter samples, up-regulation was first detected at 5 DPI and then it steadily increased until the end of sampling at 9 DPI.For the last two time points, enhanced gene expression was also detected in mAb2 samples and, to a lower extent, in P19 samples.In both cases, defence gene expression however remained significantly lower compared to H5 samples.No or very limited expression of these genes was detected in NI, mock, or mAb1 samples.For Unfolded protein response in plant cell protein biofactories 1155 NbAO genes (Figure 6d), significant gene induction was detected in all agroinfiltrated leaf samples compared to NI and mock controls.In all cases, the highest up-regulation level was reached at 5 DPI, even though gene induction started earlier and reached significantly higher levels in H5 samples.Taken together, these results suggest that oxidative stress signalling was stronger in H5 and mAb2 samples, correlating with cell death symptoms observed on leaves expressing these products (Figure 1a).For H5expressing samples, up-regulation of oxidative stress-related genes again came after the peak of UPR activation at 3 DPI.

Discussion
Plant molecular farming allows for the large-scale production of many clinically useful proteins, including therapeutic antibodies and virus surface proteins such as influenza HAs.Plant biotechnology offers alternative approaches to prevent the spreading of infectious diseases and to limit the societal and economic burdens associated with these diseases (Ortiz de Lejarazu-Leonardo et al., 2021).Despite promising developments, mass infiltration of Agrobacterium and high yield production of Figure 5 Expression of oxylipin-related genes.Expression of oxylipin regulatory genes NbPAT1 (a), NbLOX1 (b), NbCYP74a and NbCYP74b (c), as well as NbAOS1 (d), as measured by RTqPCR.The expression of oxylipin response genes NbPDF1 (e) and NbKTI3 (f) was also assessed.These genes were selected because of their responsiveness to H5 protein expression (Hamel et al., 2023).For each time point in days post-infiltration (DPI), results are expressed in numbers of molecules per nanogram of RNA.Condition names are as follows: NI, non-infiltrated leaves; Mock, leaves infiltrated with buffer only; P19, agroinfiltrated leaves expressing P19 only; H5, agroinfiltrated leaves co-expressing P19 and H5; mAb1 (mAb2), agroinfiltrated leaves co-expressing P19 and monoclonal antibody 1 (monoclonal antibody 2).foreign proteins unavoidably put pressure on host plants, including on the cellular machinery mediating protein folding and maturation within the ER.Understanding the impact of those stresses is an important step to ensure sustainability of molecular farming approaches, especially since proteins with biopharmaceutical interest generally require refined protein domain folding and post-translational modifications such as sophisticated glycosylation patterns or assembly in precise quaternary structures.Overloading the host cell ER with such complex proteins can lead to ER stress and, in turn, in activation of the UPR.When unresolved, ER stress can dramatically compromise plant cell fitness and viability, resulting in poor biomass quality and insufficient protein yields at harvest.

Protein-specific activation of the UPR
At 6 DPI, iTRAQ proteomics revealed that Agrobacterium-mediated expression of influenza protein H5 induces accumulation of UPR proteins with varied functions (Table 1; Hamel et al., 2023).Interestingly, transcriptomics at 6 DPI did not reveal up-regulation of the corresponding UPR genes, suggesting this response to be independent of UPR gene induction, or to involve early and transient up-regulation of these genes prior to biomass harvesting.Here, time course sampling showed that H5 protein expression does in fact result in the induction of UPR genes, a response initiated early after agroinfiltration, that is between 1 and 3 DPI.
UPR gene up-regulation also correlated tightly with unconventional splicing of NbbZIP60 transcripts, the closest homologue of AtbZIP60 that largely controls activation of the UPR in Arabidopsis (Iwata et al., 2008;Iwata and Koizumi, 2005).Since unconventional splicing of bZIP60 transcripts reflects the activation status of IRE1, this process is a reliable marker of UPR activation.Consistent with the lack of UPR gene induction at 6 DPI (Hamel et al., 2023), H5-mediated activation of the UPR was transient and could no longer be detected after 5 days of the heterologous protein expression.
In our attempt to better define importance of the UPR during foreign protein expression in N. benthamiana, time course sampling of mAb1-and mAb2-expressing samples was also very informative.Despite high similarity between the two IgG1 products, expression profiles of plant genes revealed completely different activation patterns of the UPR.Whereas mAb2 expression induced early, strong, and sustained activation of the UPR, expression of mAb1 resulted in late and much weaker activation of this pathway, with measured levels comparable to those induced by the expression of P19 only.Consistent with this, mAb2 resulted in heavy necrotic symptoms and poor yields, while the mAb1 had little impact on biomass quality despite high product accumulation (Figure 1).Taken together, these results suggest that expression of mAb2 led to an unresolved ER stress, perhaps induced by improper folding of its variable regions, or Unfolded protein response in plant cell protein biofactories 1157 unsuccessful quaternary protein structure assembly of its light and heavy chains.In any case, these contrasted outcomes highlight importance of the UPR molecular components in guaranteeing success and sustainability of plant molecular farming approaches.
Considering contrasted UPR activation patterns seen for the related mAb1 and mAb2 products, we conclude that H5 protein expression results in moderate activation of the UPR, at least for the viral strain H5 Indo that was employed in this study.While induction of UPR genes was obviously higher in H5 samples compared to P19 or mAb1 samples, it was similar or lower than in mAb2 samples, depending on which UPR gene is considered.As for timing, initiation of the UPR occurred within the same time frame in both H5 and mAb2 samples.With regard to UPR activation, the major difference between these two products therefore lied in duration of the response, which was transient in H5 samples and more sustained in mAb2 samples.When expressing recombinant gene H5, activation of the UPR also correlated with the initiation of H5 protein accumulation in the biomass between 1 and 3 DPI (Figure 1b).This strongly suggests that the cellular machinery involved in ER protein folding and maturation started to be mobilized as soon as the production of H5 began.Considering that this protein accumulates to levels that sustain the commercial production of influenza vaccine candidates (D'Aoust et al., 2008;Landry et al., 2010), and that this process results in somewhat mild stress symptoms on the plants (Figure 1a; Hamel et al., 2023), early and transient activation of the UPR was apparently sufficient to cope with increased cellular needs associated to foreign H5 protein secretion and to prevent deleterious effects that would have been caused by an unresolved ER stress.It is also important to keep in mind that despite reduced expression of UPR genes between 3 and 5 DPI in H5 samples, proteomics confirms that the level of several UPR proteins was still enhanced at 6 DPI (Table 1; Hamel et al., 2023).Cellular benefits from a transient activation of the UPR thus appear to last for a longer time frame, at least during expression of the foreign protein H5.

Conservation of UPR signalling branches in N. benthamiana
For H5 and mAb2 samples, in which activation of the UPR was the strongest, one of the most striking observations is the fact that expression patterns from most UPR genes examined closely mirrored the expression pattern of NbbZIP60s (Figure 2c).Since unconventional splicing of bZIP60 transcripts reflects IRE1 activity, this strongly suggests that this branch of the UPR is involved in foreign protein expression and that UPR genes examined are direct genetic targets of NbbZIP60.In plants, the UPR also comprises another key signalling branch, which is initiated by dissociation of BiPs from the ER membrane-tethered TFs bZIP17 and bZIP28 (Figure 2a; Liu and Howell, 2010;Iwata and Koizumi, 2012;Duwi Fanata et al., 2013).These bZIPs can then cruise along the endomembrane system and end up in the Golgi apparatus, where clipping mediated by Golgi-associated proteases allows for their release from anchoring membrane.Free TFs then translocate to the nucleus, where they up-regulate the expression of UPR genes in conjunction with bZIP60.Using protein sequences of Arabidopsis bZIP17 and bZIP28 as queries, blast of the N. benthamiana genome identified three geneencoding proteins that share about 50% identify with their respective homologues in Arabidopsis (Figure S5a).Although low overall, sequence identity of the protein products from Niben 101Scf32851g00038 (herein termed NbbZIP17), Niben101Sc f03647g01004 (NbbZIP28a), and Niben101Scf00077g08013 (NbbZIP28b) was much higher within functional regions, inc luding the bZIP domain, the TMD, and clipping sites targeted by Golgi-associated proteases (Figure S5b).Since the promoter region of every UPR gene examined here contains at least one cis regulatory element involved in plant UPR signalling (Figure S6), these TFs likely function in conjugation with NbbZIP60 to promote the expression of UPR genes during foreign protein expression.Development of an assay to monitor NbbZIP17 and NbbZIP28 clipping would however be required to formally confirm this hypothesis.

Activation of the UPR and plant defence gene expression
Considering that expression of influenza H5 induces transient activation of the UPR, we also hypothesize that necrotic symptoms observed on leaves expressing this product are more likely caused by later activation of plant immunity.At 6 DPI, H5 expression was shown to induce a unique molecular signature, including strong and specific up-regulation of genes involved in oxylipin signalling and responses, as well as in the activation of oxidative stress responses (Hamel et al., 2023).As many of these genes typically respond to wounding and herbivory, their upregulation is likely associated with budding of the VLPs, which hijack lipids from the PM of plant cells.At the molecular level, these numerous membrane budding events may be perceived as 'micro-wounds' that induce specific plant immunity pathways.While this portrait was drawn using biomass harvested at a single time point after 6 days of expression, the time course approach employed here confirmed that up-regulation of these defence genes comes after the peak of UPR activation at 3 DPI.Immunoblotting also indicated that defence gene up-regulation correlated with a high increase in H5 protein accumulation (Figure 1b).Taken as a whole, these findings suggest that early activation of the UPR favours high production of the H5 protein, in turn leading to the activation of plant immunity and eventually in the onset of leaf necrosis.This view is also supported by the fact that spraying of the leaves with an antioxidant solution of ascorbic acid helps to reduce H5-induced defences and development of leaf necrotic symptoms, with no negative impact on foreign protein accumulation in planta (Hamel et al., 2023).
In the case of mAb2 samples, in which the strongest necrotic symptoms were induced, expression from most defence genes examined was also induced compared to controls.When compared to H5 samples, induction of these defence genes was however of lower intensity and delayed by several days.As a result, late up-regulation of these defence genes was perhaps caused by the fact that at this stage of expression, plant cells were already dysfunctional and on the verge of collapsing.On the opposite, expression of UPR genes was induced early at 3 DPI, despite undetectable levels of the mAb2 product until 5 DPI (Figure 1e,f).High UPR gene expression was then sustained up to end of leaf sampling at 9 DPI, suggesting that induced countermeasures never solved the issue associated to the expression of this antibody.In turn, this led to poor accumulation of the recombinant product and to strong activation of plant cell death.
While defence genes that strongly respond to H5 expression were only weakly induced in mAb2 samples, this does not preclude that other defence responses were strongly induced in this condition.In fact, several examples of signalling crosstalk ª 2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163 exist between the UPR and plant immunity.In Arabidopsis for instance, the IRE1-bZIP60 pathway was shown to participate in pathogen responses, with the two IRE1 isoforms having shared and unique functions that can be dependent or independent of bZIP60 (Moreno et al., 2012).GTP-binding protein b1 (AGB1), which is involved in biotic and abiotic stress signalling, was also shown to act synergistically with IRE1 to control activation of the UPR, in addition to immune responses against Pseudomonas syringae (Afrin et al., 2022).Considering these examples of crosstalk between the UPR and plant immunity, it seems reasonable to think that sustained activation of the UPR in mAb2 samples also led to strong activation of plant immunity in this condition, in turn leading to strong plant cell death activation.While leaves expressing VLP and mAb2 products both displayed necrotic symptoms, the signalling pathways that resulted in plant cell death were most likely different between these two conditions.This perhaps also explains why cell death lesions seen on H5-expressing leaves were macroscopically different from those seen on leaves expressing mAb2 (Figure 1a).

Optimization of molecular farming approaches using components of the UPR
Achieving high recombinant protein yields in planta is one of the key aspects to ensure success and sustainability of plant molecular farming approaches.While the expression of many foreign proteins is currently performed using wild-type plants, editing of the host plant genome can be envisioned as a way to optimize productivity or product quality.Alternatively, the co-expression of helper proteins can lead to an increase in foreign protein yields, as shown in N. benthamiana with the use of protease inhibitors that favour the accumulation of recombinant antibodies (Goulet et al., 2012;Grosse-Holz et al., 2018;Jutras et al., 2016;Robert et al., 2013).As for UPR proteins, heterologous expression of a human CRT was shown to improve production of viral glycoproteins that initially showed poor accumulation levels in planta (Margolin et al., 2020).The results presented here highlight a series of N. benthamiana genes that could be used as efficient markers of the UPR activation during stress response or foreign protein expression in planta.For complex or even refractory products, the UPR genes here identified may also constitute an interesting list of potential endogenous protein helpers to be co-expressed as a way to enhance ERQC functions of plant cell biofactories, or to prevent undesirable effects provoked by an unresolved ER stress during foreign protein expression.

Seed germination and plant growth
Seeds of N. benthamiana were spread on pre-wetted peat mix plugs (Ellepot) and placed in a germination chamber for 2 days, where conditions were as follows: 28 °C/28 °C day/night temperature, 16 h photoperiod, 90% relative humidity, and light intensity of 7 lmol m À2 s À1 .Germinated plantlets were next transferred in a growth chamber for 15 days, where conditions were as follows: mean temperature of 28 °C over 24 h, 16 h photoperiod, mean relative humidity of 66% over 24 h, 800 ppm carbon dioxide (CO 2 ) injected only during the photo-phase, and light intensity of 150 lmol m À2 s À1 .During this time, watering and fertilization were provided as needed.After 2 weeks, peat mix plugs were transferred to 4 inches pots containing prewetted peat-based soil mix (Agro-Mix).Freshly transferred plantlets were then moved to a greenhouse, where conditions were as follows: mean temperature of 25 °C over 24 h, 16 h photoperiod, mean relative humidity of 66% over 24 h, 800-1000 ppm CO 2 injected only during the photo-phase, and light intensity according to natural conditions, but supplemented with artificial high pressure sodium lights at 160 lmol m À2 s À1 .In the greenhouse, watering and fertilization were provided as needed.Growth was allowed to proceed for an average of 20 additional days, until the plants were ready for agroinfiltration.

Binary vector constructs
For VLP expression, sequences from the mature HA protein of pandemic influenza virus strain H5 Indonesia (H5/A/Indonesia/05/2005; H5 Indo ) were fused to the signal peptide of a Medicago sativa (alfalfa) PDI using PCR-based methods.Once assembled, chimeric H5 gene was reamplified by PCR and then introduced in the T-DNA region of a customized pCAMBIA0380 binary vector previously linearized with restriction enzymes SacII and StuI using the In-Fusion cloning system (Clontech).The same methods and PDI signal peptide were used to express genes encoding antibody light and heavy chains, which allowed expression of recombinant antibodies mAb1 and mAb2.Expression of recombinant genes H5, mAb LC, and mAb HC was driven by a 2X35S promoter from the cauliflower mosaic virus (CaMV).Expression cassettes also comprised 5 0 -and 3 0 -untranslated regions (UTRs) from the cowpea mosaic virus (CPMV), and the Agrobacterium nopaline synthase (NOS ) gene terminator.To prevent silencing induced by recombinant gene expression in planta, T-DNA region of the binary vectors used to express H5 and antibodies also included the suppressor of RNA silencing gene P19, under the control of a plastocyanin promoter and terminator.For P19 samples, a binary vector allowing expression of P19 only was employed.

Agrobacterium cultures and plant infiltration
Binary vectors were transformed by heat shock in Agrobacterium strain AGL1.Transformed bacteria were plated on Luria-Bertani (LB) medium, with appropriate antibiotics selection (kanamycin 50 lg/mL).Colonies were allowed to develop at 28 °C for 2 days.Using isolated colonies, frozen glycerol stocks were prepared and placed at À80 °C for long term storage.When ready, frozen bacterial stocks were thawed at room temperature before transfer in pre-culture shake flasks containing LB medium with antibiotics selection (kanamycin 50 lg/mL).Bacterial precultures were grown for 18 h at 28 °C with shaking at 200 rpm.While keeping kanamycin selection, pre-cultures were transferred to larger shake flasks and bacteria were allowed to develop for an extra 18 h at 28 °C with shaking at 200 rpm.Using a spectrophotometer (Implen), bacterial inoculums were prepared by diluting appropriate volumes of the bacterial cultures in resuspension buffer (10 mM MgCl 2 , 5 mM MES, and pH 5.6).For P19, H5, mAb1, and mAb2 samples, a final OD 600 of 0.6 was employed for all experiments.Vacuum infiltration was performed by placing whole plant shoots upside down in an airtight stainless-steel tank containing the appropriate bacterial suspension.To draw air out of the leaves, vacuum pressure was applied for 1 min before pressure release to force the bacterial inoculum into the leaves.

Transient protein expression and biomass harvesting
Recombinant protein accumulation was allowed to proceed for several days, as indicated.For all experiments, expression took Unfolded protein response in plant cell protein biofactories 1159 place in condition-controlled plant growth chambers, where settings were as follows: 20 °C/20 °C day/night temperature, 16 h photoperiod, 80% relative humidity, and light intensity of 150 lmol m À2 s À1 .Watering was performed every other day, with no fertilizer supplied during the expression phase.For biomass harvesting, leaves of similar developmental stage were selected using the leaf plastochron index (Meicenheimer, 2014).The fourth and fifth fully expanded leaves starting from the top of each plant were harvested without petiole.Freshly cut leaves were placed in pre-frozen 50 mL Falcon tubes, before flash freezing in liquid nitrogen.Frozen biomass was stored at À80 °C until ready for analysis.Using pre-chilled mortars and pestles, foliar tissue was ground and homogenized into powder using liquid nitrogen.Each sample was made from four leaves collected on two randomly selected plants.The average results presented were obtained from at least three biological replicates.

Protein extraction and quantification
For protein extraction, 1 g of frozen biomass powder was taken out of the À80 °C freezer and placed on ice.A 2 mL volume of extraction buffer (50 mM Tris, 500 mM NaCl, and pH 8.0) was added, followed by 20 lL of 100 mM phenylmethanesulfonyl fluoride (PMSF) and 2 lL of 0.4 g/mL metabisulfite.Quickly after addition of all solutions, the samples were crushed for 45 s using a Polytron homogenizer (ULTRA-TURRAXâ T25 basic) at maximum speed.One millilitre of each sample was then transferred to a pre-chilled Eppendorf tube and centrifuged at 10 000 9 g for 10 min at 4 °C.Supernatants were carefully recovered, transferred to new Eppendorf tubes, and kept on ice until determination of protein concentration.To quantify protein content from crude extracts, the Bradford method was employed, with bovine serum albumin as a protein standard.

Western blotting, HMG assays, and gel densitometry
For Western blotting, total protein extracts were diluted in extraction buffer and mixed with 5X Laemmli sample loading buffer to reach a final concentration of 0.5 lg/lL.Protein samples were denatured at 95 °C for 5 min, followed by a quick spin using a microcentrifuge.Twenty microlitres of each denatured protein extract (10 lg) was then loaded on Criterion TM XT Precast polyacrylamide gels 4%-12% Bis-Tris and separated at 110 volts for 105 min.Using transfer buffer (25 mM Tris, 192 mM Glycine, and 10% methanol), proteins were next electrotransferred onto a polyvinylidene difluoride (PVDF) membrane at 100 volts.After 30 min, the membranes were placed in blocking solution: 1X Tris-Buffered Saline with Tween-20 (TBS-T; 50 mM Tris, pH 7.5, 150 mM NaCl, and 0.1% (v/v) Tween-20), with 5% nonfat dried milk.Membranes were blocked overnight at 4 °C with gentle shaking.The next morning, blocking solution was removed and primary antibodies were incubated at room temperature for 60 min with gentle shaking in 1X TBS-T, 2% nonfat dried milk solution.After four washes in 1X TBS-T, secondary antibodies were added and incubated at room temperature for 60 min with gentle shaking in 1X TBS-T, 2% nonfat dried milk solution.After four extra washes in 1X TBS-T, Luminata TM Western HRP Chemiluminescence Substrate (Thermo Fisher Scientific) was added to the membranes and protein complexes were visualized under the chemiluminescence mode of an Imager 600 apparatus (Amersham).Antibody dilutions were as follows: anti-HA A/Indonesia/05/2005 (H5N1; CBER): 1/5000 (primary antibody).Rabbit anti-sheep (JIR): 1/10 000 (secondary antibody).
For HMG assays, turkey red blood cells were diluted to a concentration of 0.25% (v/v) in phosphate-buffered saline solution (PBS; 0.1 M PO 4 , 0.15 M NaCl, and pH 7.2).While keeping red blood cells on ice, protein samples were diluted in extraction buffer using 1/384 and 1/576 ratios.For each dilution, 200 lL of total protein extract was transferred to the first row of a 96-well plate.Eight serial dilutions were next performed using 100 lL of protein extract mixed to 100 lL of PBS buffer previously poured in each plate well.Following serial dilutions, 100 lL of the red blood cell solution was added to protein extracts.After thorough mixing, samples were incubated overnight at room temperature.HA activity was scored visually on the next day.
For gel densitometry, total protein extracts were diluted in extraction buffer and mixed with 5X Laemmli sample loading buffer without reducing reagent.Protein samples were then denatured at 95 °C for 5 min, followed by a quick spin using a microcentrifuge.For each sample, 2 and 4 lg of denatured protein extracts were loaded on Criterion TM XT Precast polyacrylamide gels 4%-12% Bis-Tris.To quantify fully assembled antibodies, a standard curve consisting of 1.5 lg-0.125 lg of purified IgG1 (Sigma) was added on each gel.Protein separation was done at 110 volts for 105 min and staining of the gels was done using Coomassie blue.

RNA extractions and RNA quantification
Using the RNeasy commercial kit (Qiagen), 100 mg of frozen biomass powder was used for RNA extractions.Residual DNA was removed using the RNase-free DNase Set (Qiagen).Concentration of RNA extracts was determined using a spectrophotometer (Implen) and integrity evaluated using a 2100 BioAnalyzer (Agilent).For long-term storage, RNA extracts were stabilized by adding the RNAseOUT recombinant ribonuclease inhibitor (Thermo Fisher Scientific), before freezing at À80 °C until further analysis.

RTqPCR analyses
For each sample, 1 lg of RNA was reverse transcribed into cDNA using the QuantiTect Reverse Transcription Kit (Qiagen).Transcript quantification was performed in 96-well plates, using the ABI PRISM 7500 Fast real-time PCR system and custom data analysis software (Thermo Fisher Scientific).Each reaction contained the equivalent of 5 ng cDNA as a template, 0.5 lM of forward and reverse primers, and 1X QuantiTect SYBR Green Master Mix (Qiagen) for a total reaction volume of 10 lL.RTqPCR runs were done under the SYBR Green amplification mode and cycling conditions were as follows: 15 min incubation at 95 °C, followed by 40 amplification cycles at 95 °C for 5 s, 60 °C for 30 s, and 65 °C for 90 s.Reactions in the absence of cDNA template were conducted as negative controls and melting curve analyses were performed to confirm the lack of primer dimer formation and amplification specificity.Resulting fluorescence and cycle threshold (Ct) values were next exported to the Microsoft Excel software.To correct for biological variability and technical variations during RNA extraction, RNA quantification, or reverse transcription, expression from six housekeeping genes (NbACT1, NbVATP, NbSAND, NbUBQ1, NbEF1-a, and NbGAPDH1) was used to normalize expression data (Hamel et al., 2023;Vandesompele et al., 2002).Normalized numbers of transcript molecules per nanogram of RNA were deduced using the 2 ÀDDCt method (Bustin et al., 2009;Livak and Schmittgen, 2001) and standard curves derived from known quantities of ª 2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163 phage lambda DNA.Standard deviation related to the withintreatment biological variation was calculated in accordance with the error propagation rules.Sequences of all primers used in this study are available in the Table S2.

Phylogenetic analyses
For each protein family to be analysed, the genome of N. benthamiana (https://solgenomics.net/) was searched using fulllength amino acid sequences of Arabidopsis homologues as queries.Full-length protein sequences retrieved from blast analyses were next aligned with ClustalW.Alignment parameters were as follows: for pairwise alignments, 10.0 for gap opening and 0.1 for gap extension.Parameters for multiple alignments were 10.0 for gap opening and 0.20 for gap extension.The resulting alignments were submitted to the MEGA5 software and neighbour-joining trees derived from 5000 replicates were generated.Bootstrap values are indicated on the node of each branch.More details on phylogenetic analyses are provided in the legend of relevant figures.

Statistical analyses
Statistical analyses on RTqPCR data were performed on Graph Pad Prism 9.5.0, using two-way ANOVA with time points as row factors and conditions as column factors.Conditions were then analysed by a post hoc Tukey's multiple comparison test with an alpha threshold of 0.05.For clarity and to limit the number of statistical groups generated, control conditions (NI and mock) were omitted from statistical analyses, which were thus only performed on agroinfiltrated conditions (P19, H5, mAb1, and mAb2).Groups are labelled with a compact letter display.Groups that do not share the same letter(s) are statistically different.Complete results from statistical analyses can be found in the Table S3.

Search for cis regulatory elements
To locate cis regulatory elements within the promoter of UPR genes, the putative promoter regions of each gene tested were retrieved from the genome sequences of N. benthamiana (https://solgenomics.net/).For the analysis, a 1000 bp region located upstream of the annotated start codon was employed.Motif search was performed in both sequence orientations using the Find Individual Motif Occurrences (FIMO) software (Grant et al., 2011).As they were previously involved in the UPR of plants (Iwata et al., 2008;Iwata and Koizumi, 2005;Liu and Howell, 2016), cis regulatory elements that were examined were as follows: PUPRE (ATTGGTCCACGTCATC), ERSE (CCAAT N 10 CACG), ERSE2 (ATTGGN 2 CACG), UPRE2 (GATGACGCGT AC), UPRE3 (TCATCG), and XBP-BS (GATGACGTGK).
Table S3 Statistical analyses of the RTqPCR results.

ª
2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163

Figure 1
Figure 1 Stress symptoms and recombinant protein accumulation.(a) Stress symptoms observed on representative leaves from each condition harvested 9 days post-infiltration (DPI).Magnified leaf sections highlight plant cell death on H5 and mAb2 leaves.(b) Total protein extracts from H5 samples following SDS-PAGE and Coomassie blue staining (upper panel).Arrows highlight RuBisCO small and large subunits (RbcS and RbcL respectively).A Western blot also depicts H5 accumulation at each time point (lower panel).A NI sample harvested at 9 DPI was used as a control lacking H5 expression.(c) For H5 samples, haemagglutination (HMG) assays depict HA protein activity at each time point.(d) For H5 samples, transmission electron microscopy (TEM) confirms production of VLPs at 3, 5, 7, and 9 DPI.Scale bars equal 100 nm.(e) Total protein extracts from mAb1 (top panels) and mAb2 (lower panels) samples after SDS-PAGE and Coomassie blue staining.Using standard curves made with purified IgG1, gels were used to quantify accumulation of complete monoclonal antibodies (mAb), as measured by densitometry.For each of three repetitions (R1-R3), 2 and 4 lg of total soluble proteins was loaded.Arrows indicate complete IgGs and a purified IgG1 control (CTL) is shown on the left.(f) From densitometry results, fold-change (FC) accumulation of each mAb was deduced, with accumulation of mAb1 at 3 DPI arbitrarily set at onefold (dashed line).Asterisks denote mAb levels below the limit of quantification, as defined by the standard curves.Condition names are as follows: NI, non-infiltrated leaves; Mock, leaves infiltrated with buffer only; P19, agroinfiltrated leaves expressing P19 only; H5, agroinfiltrated leaves co-expressing P19 and H5; mAb1 (mAb2), agroinfiltrated leaves co-expressing P19 and monoclonal antibody 1 (monoclonal antibody 2).
For NI and mock samples, levels of spliced transcripts remained low and unchanged ª 2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163

Figure 2 Figure 3
Figure2The IRE1-bZIP60 pathway and up-regulation of NbbZIP60.(a) Model depicting unconventional splicing of bZIP60 transcripts by IRE1, an ER transmembrane sensor with protein kinase (PK) and ribonuclease (RN) activity.In the absence of stress, unspliced bZIP60 transcripts (bZIP60u) encode a transcription factor (TF) with a transmembrane domain (TMD).Membrane tethering of bZIP60 prevents translocation to the nucleus.ER stress results in misfolded protein accumulation and consequent ERQC and ERAD activation.This triggers IRE1 activation (IRE1*) through oligomerization and phosphorylation (P).Unconventional splicing of bZIP60 transcripts (bZIP60s) results in translation of a shorter TF that lacks a TMD and that can thus induce UPR gene expression in the nucleus.Within the promoter of UPR genes, bZIP60s recognizes cis regulatory elements such as the plant unfolded protein response element (UPRE) or the ER stress-response element (ERSE).Adapted from Duwi Fanata et al., 2013.To assess activation of the UPR, expression of NbbZIP60u (b) and NbbZIP60s (c) was measured by RTqPCR.For each time point in days post-infiltration (DPI), results are expressed in numbers of molecules per nanogram of RNA.Condition names are as follows: NI, non-infiltrated leaves; Mock, leaves infiltrated with buffer only; P19, agroinfiltrated leaves expressing P19 only; H5, agroinfiltrated leaves co-expressing P19 and H5; mAb1 (mAb2), agroinfiltrated leaves co-expressing P19 and monoclonal antibody 1 (monoclonal antibody 2).

ª
2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163

ª
2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163

ª
2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163

Figure S1
Phylogeny of N. benthamiana PDIs and ER-resident chaperones.Figure S2 Expression of recombinant genes.

Figure S3
Primer design to assess unconventional splicing of NbbZIP60.

Figure S4
Figure S4Phylogeny of other UPR proteins and expression of some of their corresponding genes.FigureS5Phylogeny of UPR activating bZIPs and conservation of bZIP17 and bZIP28 in N. benthamiana.FigureS6Cis regulatory elements in the promoter of UPR genes.

Table 1
UPR proteins up-regulated by P19 expression or co-expression of P19 and influenza H5.
ª 2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163 Unfolded protein response in plant cell protein biofactories 1147 Log2FC values that do not meet the minimum threshold are highlighted in red.Niben number of the best corresponding gene is provided, as is the presence or absence of an ER retention signal.For each protein, sequences of the last 25 amino acids are shown, including ER retention signals highlighted in blue.Stop codons are depicted by a black dot (•) and total predicted number of amino acids is shown in parenthesis.On the right, names attributed to each UPR gene is shown.
≥ 1.For each protein retained, UniProt number, annotation, Log2FC value compared to the mock treatment, and deduced Z-score are indicated.ª 2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 1146-1163 (Table