Introduction

Plant molecular farming (PMF) is defined as the use of plant-based platforms to produce valuable proteins or metabolites that are used in the biopharmaceutical, nutraceutical or cosmetic industries, and more. Nicotiana benthamiana is the preferred host due to its susceptibility to transfection, rapid growth, and high biomass generation [1]. Despite these advantages, the accumulation of exogenously expressed proteins is often limited by the plant’s physiological transition from a growth state to a defense state upon agroinfiltration [2]. This defense response creates a proteolytic environment in the apoplast and intracellular compartments, leading to product degradation and reduced yields [3,4]. Although strategies such as co-expression of protease inhibitors have mitigated some losses [5,6], current approaches often rely on a generalized understanding of stress rather than addressing specific degradation pathways. The stress imposed during transient expression is cumulative and multi-layered. The infiltration process itself applies significant mechanical pressure to leaf tissue, potentially triggering wound responses that are distinct from the immune responses elicited by Agrobacterium tumefaciens’ pathogen-associated molecular patterns (PAMPs) [7].

The physicochemical properties of the recombinant product influence the magnitude of cellular stress. Complex proteins that transit the secretory pathway, such as enveloped virus-like particles (VLPs), require extensive protein folding and post-translational processing within the endoplasmic reticulum (ER). For such cargoes, the structural proteins transit the ER and Golgi, where they undergo folding, disulfide bond formation and glycosylation before or concomitant with higher-order assembly. In most enveloped VLP systems, intracellular trafficking occurs primarily at the level of individual structural proteins or small oligomers, with particle assembly coupled to membrane localization and budding. Recent transcriptomic studies suggest that VLP expression triggers the unfolded protein response (UPR) and immune signaling more intensely than simple soluble proteins like Green Fluorescent Protein (GFP) [8,9]. Whether cytosolic cargoes that do not enter the secretory pathway impose a comparable burden remains unclear, and proteomic evidence distinguishing these layers of stress is limited. To address these complexities, we employed 4D-DIA (Data-Independent Acquisition) mass spectrometry, which incorporates ion mobility separation to achieve superior sensitivity and proteome coverage compared to traditional methods [10].

We designed an experiment to systematically decouple the sources of host stress by comparing non-infiltrated wild-type plants (WT), buffer-infiltrated controls (mechanical stress), empty vector Agrobacterium (bacterial stress), and Agrobacterium expressing either GFP or the Foot-and-Mouth Disease Virus antigen P1_2A (product-specific stress) [11]. This study provides a high-resolution map of the N. benthamiana proteome dynamics, revealing that the host proteolytic landscape is actively reshaped by the specific nature of the recombinant cargo. We aimed to: (1) determine whether buffer infiltration alone induces a distinct proteomic signature separate from the Agrobacterium-mediated immune response; (2) define the core Agrobacterium-responsive proteome independent of cargo effects; and (3) test the hypothesis that host proteolytic responses are cargo-dependent, with a multidomain viral capsid precursor eliciting a more pronounced protease response than simple soluble proteins.

Materials and methods

Results

Deep proteome coverage via 4D-DIA captures subtle host dynamics

We utilized the 4D-DIA platform to analyze the temporal dynamics of the N. benthamiana proteome at 1, 4, and 7 dpi. The experimental design included five treatment groups to isolate specific stress factors (Fig 1A). A total of 61,379 unique peptide sequences were identified from the raw data and mapped to 9,495 protein groups across three independent biological replicates and subsequently quantified using the MS1 intensity-based label-free quantification (LFQ) method (S1 Table). Pearson correlation analysis confirmed high reproducibility across all 45 proteomes, with average correlation coefficients exceeding 0.95 within each treatment group (Fig 1B). UMAP dimensionality reduction of the global proteome revealed a clear topological separation between agroinfiltrated and non-infiltrated samples: WT and buffer controls formed a distinct cluster from the GV3101, GFP, and P1_2A groups, and this divergence intensified progressively from 1 to 7 dpi (Fig 1C). This global trajectory demonstrates that Agrobacterium-mediated transformation fundamentally reshapes the host proteome in a time-dependent manner. Per-time-point PCA (Fig 1D) quantified the same contrast within each time point separately. The agroinfiltrated and non-infiltrated groups were separated mainly along PC1, which accounted for 30.6%, 38.9% and 31.3% of the total variance at 1, 4 and 7 dpi. Agroinfiltration is therefore the largest single source of proteome variation at each time point.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Experimental design and global proteomic landscape of N. benthamiana responses to agroinfiltration.

(A) Experimental workflow. N. benthamiana leaves were subjected to five treatments: untreated wild type (WT), buffer control, empty Agrobacterium (GV3101), and Agrobacterium expressing GFP or FMDV antigen P1_2A. Samples were harvested at 1, 4, and 7 days post-infiltration (dpi) and analyzed via 4D-DIA LC-MS/MS (n = 3 biological replicates). (B) Pearson correlation heatmap of all 45 proteomes. (C) UMAP visualization of the global proteomic trajectory, showing temporal shifts and distinct clustering between non-infiltrated controls (WT/buffer) and Agrobacterium-infiltrated groups. (D) Principal component analysis at each time point (1, 4 and 7 dpi). Dashed outlines enclose the agroinfiltrated samples (GV3101, GFP, P1_2A) and the non-infiltrated controls (WT, buffer). n = 3 biological replicates per treatment.

https://doi.org/10.1371/journal.pone.0356905.g001

Buffer infiltration triggers early defense and wound responses

To understand the “background noise” in molecular farming experiments, we analyzed the DEPs between the Buffer and WT groups. Buffer- and WT-treated plants differ only in the mechanical act of infiltration; neither receives Agrobacterium. The volcano plots revealed that the buffer treatment is biologically active, inducing significant protein changes as early as 1 dpi (Fig 2A, S2 Table). Functional enrichment analysis (GO and KEGG) indicated a significant upregulation of proteins involved in phenylpropanoid biosynthesis, MAPK signaling, and oxidoreductase activity (Fig 2B). Specifically, terms such as “cell wall biogenesis,” “apoplast,” and “hydrogen peroxide catabolic process” were enriched (Fig 2C).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Time-course proteomic profiling and functional enrichment in response to buffer infiltration.

(A) Volcano plots of differentially expressed proteins (DEPs) between buffer-infiltrated and untreated WT leaves at 1, 4, and 7 days post-infiltration (dpi). Red and blue dots indicate significantly up- and down-regulated proteins, respectively (fold change > 1.2, P < 0.05). (B) KEGG pathway enrichment analysis of the up- and down-regulated DEPs at 1, 4, and 7 dpi. (C) Gene Ontology (GO) enrichment analysis of biological processes and molecular functions for the up- and down-regulated DEPs at 1, 4, and 7 dpi.

https://doi.org/10.1371/journal.pone.0356905.g002

The core Agrobacterium-responsive proteome

Having established that buffer infiltration alone constitutes a significant abiotic stressor, we next sought to identify the proteomic changes attributable specifically to Agrobacterium. To isolate the host response specifically attributable to Agrobacterium infection, we identified the intersection of DEPs found in all bacteria-treated groups (GV3101, GFP, and P1_2A) relative to the Buffer control (Fig 3A, S2 Table). This approach filters out mechanical stress, leaving a “core Agrobacterium-responsive proteome.” This core response was characterized by extensive metabolic reprogramming consistent with the growth-defense trade-off hypothesis [17]. We observed a universal downregulation of proteins associated with photosynthesis (e.g., photosynthesis, light harvesting, chlorophyll A-B binding proteins, and Rubisco small subunits) and carbon metabolism across all time points (Fig 3B, 3C). Conversely, proteins involved in protein processing in the endoplasmic reticulum (ER), ribosome biogenesis, and vesicle trafficking were significantly upregulated (Fig 3B). KEGG pathway analysis further confirmed this metabolic shift, revealing the downregulation of photosynthesis-antenna proteins and carbon fixation pathways, while plant-pathogen interaction, MAPK signaling, and protein processing in the ER were enriched among the upregulated core proteins (Fig 3C). The enrichment of plant-pathogen interaction pathways is consistent with PAMP perception via pattern recognition receptors such as EFR, which detects the bacterial elongation factor EF-Tu [18]. At the metabolic level, Prudhomme et al. [19] demonstrated that the jasmonic acid (JA) precursor 12-OPDA and six fatty acid β-oxidation enzymes accumulated in agroinfiltrated tissue, connecting PAMP-triggered immunity to hormonal defense signaling. The extracellular proteome analysis by Grosse-Holz et al. [20] showed that hydrolytic defense proteins, including chitinases (GH18), PR2 glucanases (GH17), and xylanase inhibitors, accumulated in the apoplast by 2 dpi, while papain-like cysteine proteases (PLCPs, family C01) did not increase proportionally in the extracellular fraction. In contrast, our total cellular proteome data captured the intracellular accumulation of these PLCPs, suggesting that many PLCPs induced during agroinfiltration are retained in vacuolar or ER-associated compartments rather than being secreted. The number of core DEPs increased from 1 to 7 dpi, indicating a progressive amplification of the host defense response over time. This temporal escalation is consistent with the pattern reported by Grosse-Holz et al. [20], in which the number of differentially abundant extracellular proteins expanded from 2 to 10 dpi.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Identification and functional enrichment of the core Agrobacterium-responsive proteome.

(A) Venn diagrams displaying the overlap of DEPs among GV3101, GFP, and P1_2A treatments relative to buffer control at 1, 4, and 7 dpi. (B) Gene Ontology (GO) enrichment analysis of the up- and down-regulated core DEPs. (C) KEGG pathway enrichment analysis.

https://doi.org/10.1371/journal.pone.0356905.g003

Expanded protease landscape and the core proteolytic response

Proteases are the primary bottleneck in molecular farming. Given that proteolytic degradation is a primary barrier to recombinant protein accumulation, we next focused our analysis on the protease component of these responses. Leveraging the latest N. benthamiana genome assembly [14] and our deep proteomic data, we annotated 1,505 putative proteases, significantly expanding upon the 1,243 proteases previously reported [20]. These were classified into catalytic families based on MEROPS nomenclature [21], with serine (Ser), cysteine (Cys), and aspartic (Asp) proteases dominating the landscape (Fig 4A, S3 Table). Hierarchical clustering showed that a large subset of these proteases is upregulated upon agroinfiltration regardless of the cargo (Fig 4B). We identified specific enzymes, such as the aspartic protease NbALP1, the cysteine protease NbCTB2, and the subtilase NbSBT6, as part of the core proteolytic response (Fig 4C). These enzymes showed consistent upregulation in GV3101, GFP, and P1_2A groups compared to buffer controls. NbALP1 and NbCTB2 are known to be involved in the degradation of recombinant proteins in the apoplast [22,23].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. The expanded protease landscape of N. benthamiana and the core proteolytic response.

(A) Classification of the 1,505 proteases identified in this study into MEROPS catalytic classes (Serine, Cysteine, Aspartic, Metallo, Threonine). (B) Hierarchical clustering of protease expression profiles across all treatments. Red indicates upregulation; blue indicates downregulation. (C) Expression profiles of representative “Core” proteases including NbALP1 (Aspartic), NbCTB2 (Cysteine), NbSBT6 (Serine), and NbVPE1b (Cysteine).

https://doi.org/10.1371/journal.pone.0356905.g004

P1_2A triggers a product-specific proteolytic response

While the above analyses identified a common proteolytic response to Agrobacterium, a key question for molecular farming is whether the nature of the recombinant cargo further shapes the host response. A direct comparison between the GFP and P1_2A expression groups revealed marked differences in the host response. P1_2A, an animal viral capsid precursor reported to self-assemble into VLPs [11], induced a higher number of DEPs compared to the soluble GFP reporter, particularly at 4 dpi (Fig 5B, S2 Table). Functional enrichment analysis of the P1_2A-specific response highlighted pathways related to cytoskeleton organization (“microtubule-based process”), nuclear pore complex, and chromatin remodeling (Fig 5A). This likely reflects the cellular demands imposed by a large, multidomain cargo [8]. Most critically, P1_2A expression triggered the specific and intense upregulation of a distinct subset of proteases (Fig 5D). Time-course analysis of recombinant protein accumulation revealed markedly different dynamics between the two cargoes: GFP protein intensity peaked sharply at 4 dpi before declining by 7 dpi, whereas P1_2A protein levels remained consistently low throughout the time course (Fig 5B). A similar trend was observed for relative expression levels, with both recombinant proteins approaching baseline by 7 dpi (Fig 5C). While GFP achieved substantially higher peak accumulation, the P1_2A group exhibited a more sustained and intense stress-related proteome. The heatmap of cargo-specific differentially expressed proteases revealed that 10 proteases spanning three catalytic classes [aspartic (Asp), cysteine (Cys), and serine (Ser)] were significantly differentially regulated between GFP and P1_2A at 4 dpi (Fig 5D). Specific proteases, including NbSBT9 (serine), NbRD21B (cysteine), and NbVPE1b (Vacuolar Processing Enzyme, cysteine), were hyper-induced in the P1_2A group, along with previously uncharacterized proteases Nbe16g06900.1 (Asp), Nbe19g17120.1 (Cys), and Nbe14g08810.1 (Cys). This defines a cargo-specific protease signature at 4 dpi that includes vacuolar processing enzymes (VPEs) [24,25]. As shown below, however, this signature was not accompanied by cargo-specific differences at the transcript level, suggesting that it may reflect differences in the level or kinetics of recombinant-protein accumulation rather than a distinct P1_2A-specific stress program.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Cargo-specific proteomic responses and RT-qPCR validation.

(A) Gene Ontology (GO) functional enrichment analysis of differentially expressed proteins between P1_2A and GFP groups at 4 dpi. Node color indicates -log10(P-value) and node size indicates the protein ratio. (B) Time-course comparison of relative protein intensity for the recombinant proteins GFP and P1_2A at 1, 4, and 7 dpi (mean ± SD, n = 3). (C) Relative expression levels of GFP and P1_2A over the infiltration time course, normalized to 1 dpi values. (D) Heatmap of cargo-specific differentially expressed proteases between GFP and P1_2A groups at 4 dpi. Proteases are classified by catalytic type: Asp (aspartic), Cys (cysteine), and Ser (serine). Red indicates upregulation; blue indicates downregulation. (E) qRT-PCR validation of ten selected protease genes. Bar plots show relative expression levels in GV3101, GFP, and P1_2A groups at 4 dpi (mean ± SD, n = 3). Asterisks indicate statistical significance (*P < 0.05, Student’s t-test).

https://doi.org/10.1371/journal.pone.0356905.g005

Discussion

Our study provides a high-resolution, temporally resolved proteomic atlas of the multi-layered host response during agroinfiltration of N. benthamiana, revealing that the widely assumed “neutral” controls (WT and buffer-infiltrated plants) in molecular farming are biologically active. The buffer infiltration alone induced significant changes in the proteome, including the upregulation of cell wall remodeling enzymes, MAPK signaling components, and oxidoreductases as early as 1 dpi (Fig 2). This finding is consistent with the “infiltration stress” concept described by Hamel [2] and the metabolic perturbations reported by Drapal et al. [7] in different N. benthamiana accessions. The activation of phenylpropanoid biosynthesis pathways in our buffer-infiltrated samples mirrors the wound-induced lignification response documented in other systems. Prudhomme et al. [19] similarly noted that bacterial growth alone reshapes the N. benthamiana metabolome, but our data extend this by showing that even the mechanical component of infiltration, without any bacterial presence, is sufficient to induce defense-associated proteins. This has practical implications: experiments using buffer-infiltrated controls as a “baseline” may systematically underestimate the total stress burden imposed on the plant, as these controls already represent a perturbed state.

By subtracting the buffer infiltration background, we defined a core Agrobacterium-responsive proteome that was shared across all bacteria-treated groups regardless of the expressed cargo. This core response was dominated by a growth-defense trade-off, with photosynthetic proteins consistently downregulated and defense-related proteins upregulated (Fig 3). This pattern aligns closely with the transcriptomic findings of Grosse-Holz et al. [20], who reported a similar suppression of photosynthesis-related transcripts during agroinfiltration. Our proteomic data confirm that this trade-off is also manifest at the protein level and persists across the 1–7 dpi time window. The metabolic cost of mounting this defense was recently shown to be partially reversible through light and cytokinin co-treatments [17], which underlines the need to understand these basal responses for yield optimization. The upregulation of ER-associated protein processing and vesicle trafficking components in the core response suggests that PAMP recognition by receptors such as EFR [18] triggers a sustained secretory pathway remodeling, which may compete with recombinant protein folding and secretion.

Perhaps the most consequential finding for the molecular farming is the demonstration that host proteolytic responses are not uniform but cargo-dependent. The P1_2A viral capsid precursor induced a quantitatively and qualitatively distinct protease profile compared to soluble GFP (Fig 5D). The specific upregulation of subtilases (e.g., NbSBT9), papain-like cysteine proteases (e.g., NbRD21B), and vacuolar processing enzymes (e.g., NbVPE1b) in the P1_2A group points to the activation of both apoplastic and vacuolar degradation pathways. The two cargoes also differ in size and architecture. GFP is a 27 kDa single-domain protein that folds on its own; P1_2A is an 85 kDa polyprotein of four capsid domains that must be cleaved at the 2A site before assembly. The larger cargo folds more slowly and is processed during maturation, so it stays exposed to proteases for longer. That alone could account for part of the cargo-dependent difference, without invoking the particle. We initially considered whether this cargo-specific difference might reflect differential ER stress, by analogy to reports that secretory glycoproteins such as influenza haemagglutinin and coronavirus-like particles trigger an intense UPR in N. benthamiana [8,9]. However, those cargoes are ER-resident secretory glycoproteins, whereas GFP and the FMDV P1_2A capsid precursor are cytosolic proteins that do not transit the secretory pathway. Consistent with this, the UPR-marker RT-qPCR (NbBiP, NbPDI, NbbZIP60, NbCRT) showed that the UPR was broadly induced by agroinfiltration but scaled with recombinant-protein accumulation rather than with cargo identity, being lowest in the poorly accumulating P1_2A group, indicating that the cargo-specific protease differences are unlikely to be driven by a P1_2A-specific UPR (S1 Fig). Our data cannot say whether the P1_2A-specific protease response comes from the unassembled precursor, from assembled particles, or from degradation intermediates of either. Because P1_2A accumulated poorly, unassembled and partially folded precursor is the more likely source. Distinguishing these would require fractionating infiltrated tissue by density gradient or imaging it by electron microscopy, or comparing P1_2A with an assembly-deficient variant that accumulates precursor but forms no particles. Grosse-Holz et al. [20] identified approximately 1,243 proteases from their transcriptome and secretome analyses; our 4D-DIA approach expanded this repertoire to 1,505 enzymes, likely reflecting the improved sensitivity afforded by ion mobility separation and the use of an updated genome assembly. Our RT-qPCR data revealed that the cargo-specific differences in protease abundance are not fully reflected at the transcript level, as P1_2A vs. GFP comparisons did not reach statistical significance for any of the ten tested protease genes (Fig 5E). This discordance suggests that post-transcriptional regulation, such as differential mRNA stability or protein turnover, plays a significant role in shaping the cargo-specific proteolytic landscape. These findings argue against the “one-size-fits-all” approach to protease inhibitor co-expression and instead support targeted strategies: for multidomain antigens like P1_2A, silencing or inhibiting specific SBTs and PLCPs [6,26] may be more effective than broad-spectrum approaches. The 4D-DIA approach enabled the identification of 9,495 protein groups across all samples, achieving a deeper proteome coverage than previous DDA-based studies of agroinfiltrated N. benthamiana. The incorporation of ion mobility as a fourth separation dimension enhanced peptide identification confidence and quantification accuracy, particularly for low-abundance regulatory proteins such as proteases and signaling kinases. The annotation of 1,505 putative proteases, representing a 21% increase over the 1,243 reported by Grosse-Holz et al. [20], can be attributed to three factors: (i) the use of the updated HZ version 1 genome assembly [14], which provides more complete gene models; (ii) the superior sensitivity of 4D-DIA in detecting low-abundance proteins; and (iii) systematic hmmer-based domain scanning against the MEROPS/Pfam reference database with default E-value thresholds. However, some of these annotations remain putative and would benefit from experimental validation of proteolytic activity, for instance through activity-based protein profiling as described by Jutras et al. [22]. This study has several limitations. First, our analysis was conducted using a single N. benthamiana accession RA4 under controlled growth conditions. Agroinfiltration responses differ between accessions [7], so these results may not transfer to other accessions or to field-grown plants. Second, we sampled at three time points (1, 4, and 7 dpi), which captures the major temporal dynamics but may miss early signaling events occurring within hours of infiltration. Third, we used a single Agrobacterium density (OD₆₀₀ = 0.6); dose-dependent effects on the host proteome were not explored. Fourth, while our RT-qPCR validation targeted ten protease genes, confirmation of the broader proteomic findings would benefit from orthogonal approaches such as Western blotting or targeted proteomics (PRM/SRM). Finally, the mechanistic basis of the cargo-specific protease induction, whether driven by ER stress sensing, T-DNA delivery efficiency, or direct protein–protein interactions, is not yet clear. Future studies combining genetic perturbation (e.g., CRISPR-mediated knockout of key SBTs and PLCPs) with time-resolved proteomics will be essential to translate these observations into actionable strategies for improving recombinant protein yields in plant molecular farming.

The observation that recombinant-protein levels peaked at 4 dpi and declined toward baseline by 7 dpi should not be interpreted as evidence of a suboptimal expression system. In the binary T-DNA system used here the transgene is expressed from T-DNA that is delivered in a single wave of infection and is predominantly maintained as non-integrated episomal DNA during transient expression [27]. Neither the DNA template nor the transcript is autonomous amplified in planta, so the available pool of transcriptionally competent T-DNA is established during delivery and subsequently decreases through dilution, degradation, chromatinization, or transcriptional silencing. Product accumulation is expected to rise to an early maximum and then fall as that pool is depleted and the infiltrated tissue mounts its defense response. Extrachromosomal T-DNA can still be recovered from agroinfiltrated leaves at 4–8 dpi [28], so the template is physically present across the window we sampled, giving a persistence on the order of a week. Whether it remains transcriptionally active throughout that period has not been established, so the 4 dpi maximum and near-baseline 7 dpi values we report cannot be attributed to template loss alone. This contrasts with viral-replicon (deconstructed-virus) vectors, in which continuous replication of the transgene transcript sustains and amplifies accumulation over a longer window; reports of sustained recombinant-protein accumulation over a week therefore largely reflect the replicative nature of those systems rather than a general property of agroinfiltration. The 4 dpi peak we observe is thus consistent with the intrinsic kinetics of non-amplified transient expression, and the cargo-independent decline reflects this shared kinetic ceiling rather than a distinct, cargo-specific suppression mechanism.

Our results also point to practical routes for improving plant-based recombinant-protein production. Because the cargo-specific losses we observe track particular host proteases rather than a generalized stress program, suppressing the specific subtilases and papain-like cysteine proteases induced during expression—rather than co-expressing broad-spectrum protease inhibitors—may be a more effective way to stabilize multidomain antigens such as P1_2A. Reducing the early, infiltration-associated mechanical stress and the accumulation-linked UPR, for instance through gentler delivery, optimized infiltration media, or co-expression of stress-buffering factors, may extend the productive window beyond the 4 dpi peak seen here. The question of how agroinfiltration itself might be replaced is also pertinent: stable transgenic lines avoid the acute wound-and-immune response but need much longer development times; viral-replicon vectors sustain higher per-cell expression through transcript amplification but raise biosafety and genetic-stability concerns; and cell-free or suspension-culture platforms trade scale and post-translational capacity for greater control. A host-proteome map of the kind presented here offers a rational starting point for choosing and engineering among these options.

A broader consideration when interpreting our results is the subcellular destination of the recombinant cargoes. GFP and the FMDV P1_2A capsid precursor are cytosolic proteins that are neither translocated into the endoplasmic reticulum nor routed through the secretory pathway, whereas much of the molecular-farming literature against which we benchmark our data concerns secreted or ER-resident products, for example the secreted antibody trastuzumab studied by Prudhomme et al. [19], the three secreted recombinant proteins profiled by Grosse-Holz et al. [20], and the secretory glycoprotein VLPs such as influenza hemagglutinin and coronavirus-like particles that elicit a strong UPR [8,9]. Secreted and ER-client proteins place folding, glycosylation and quality-control demands directly on the ER, so the ER-processing and UPR responses described in those systems may not operate identically for the cytosolic cargoes used here. This difference bounds how far our observations can be compared with that body of work, and it is consistent with our finding that the cargo-specific protease differences are decoupled from a classical, secretion-linked UPR. It also implies that engineering strategies aimed at relieving ER stress during secreted-protein production may transfer only partially to cytosolic antigens such as P1_2A.

Conclusion

In conclusion, this study provides three main insights. First, buffer infiltration is not a neutral control but actively primes wound and oxidative stress responses, confounding the interpretation of bacterial and product-specific effects. Second, the core Agrobacterium-responsive proteome is defined by a sustained growth-defense trade-off, explaining the metabolic cost of transient expression. Third, host proteolytic responses vary with the recombinant cargo: the multidomain capsid precursor P1_2A was associated with a distinct and more pronounced protease signature—enriching subtilases and cysteine proteases—that was not mirrored by proportional cargo-specific differences at the transcript level. These findings argue against a uniform, broad-spectrum protease-inhibitor strategy and instead support approaches tailored to the target product.

Supporting information