Feedback-responsive cell factories for dynamic modulation of the unfolded protein response

Abstract

Engineering cell factories that support the production of large quantities of protein therapeutics remains a significant biomanufacturing challenge. The overexpression of secretory proteins causes proteotoxic stress, affecting cell viability and protein productivity. Proteotoxic stress leads to the activation of the Unfolded Protein Response (UPR), a series of signal transduction pathways regulating protein quality control mechanisms aimed at restoring homeostasis. Sustained UPR activation culminates with the induction of apoptosis. Current strategies for enhancing the production of therapeutic proteins have focused on the deregulated modulation of key components of the UPR. These strategies have resulted in limited and often protein-specific improvements as they may lead to adaptation and cell toxicity and do not account for natural population heterogeneities. We report here feedback-responsive cell factories that sense proteotoxic stress and, in response, modulate the UPR to enhance stress attenuation and delay cell death, addressing the limitations of current strategies. We demonstrate that our cell engineering approach enables dynamic UPR modulation upon proteotoxic stress. The sense-and-respond systems that mediate dynamic UPR modulation enhance the production of the therapeutic enzyme tissue plasminogen activator and the bispecific antibody blinatumomab. Our feedback-responsive cell factories provide an innovative strategy for dynamically adjusting the innate cellular stress response and enhancing therapeutic protein manufacturing.

Engineering cell factories that support the production of large quantities of protein therapeutics remains a significant biomanufacturing challenge. Here the authors engineered cells to sense an early marker of the Unfolded Protein Response (UPR) and, in response, modulate specific UPR signaling pathways mediating stress attenuation and/or apoptosis.

Introduction

Protein therapeutics have revolutionized modern medicine and represent a critical fraction of the healthcare industry. Due to their fundamental role in diverse biological processes, protein therapeutics are used to treat many diseases, from acute ischemic strokes to rheumatoid arthritis and cancer¹. As a result, the therapeutic protein market has grown dramatically in the past decade, with more than a dozen recombinant proteins being approved annually for clinical use². Such market growth is limited by the tremendous demand for affordable drugs surpassing the global production capacity. The introduction of biosimilars is expected to expand the therapeutic protein market, further enhancing the global demand³. This increasing demand has created a pressing need for more efficient production methods to meet the market demand and generate economic supplies of large quantities of therapeutic proteins for cost-sensitive indications.

Current strategies for enhancing recombinant protein production focus on optimizing DNA vector design, building and isolating stable cell lines, and engineering pathways mediating protein folding and secretion, cell growth, and cell viability^4–9. Recent advances in genetic engineering have led to dramatic improvements in protein yields and strain optimization. Protein overexpression, however, results in the accumulation of off-pathway, misfolded intermediates and proteotoxic stress, which leads to the activation of a global response aimed at restoring homeostasis that may, ultimately, inhibit protein synthesis and reduce cell viability. Increasing the transgene copy number or mRNA concentration does not result in a corresponding increase in protein productivity, suggesting translational and post-translational bottlenecks affect the overall protein yields^7,10. Modulation of post-translational processes in high-expressing cells has resulted in limited and often protein-specific improvements in productivity. Such cell engineering efforts, including the overexpression of folding chaperones and modulation of apoptotic genes, support the notion that post-translational processing is a bottleneck of recombinant protein production^4,7,11–15.

Most protein therapeutics are large proteins with complex folding and post-translational modifications secreted extracellularly and processed through the secretory pathway. The folding and quality control of secretory proteins occurs in the endoplasmic reticulum (ER). Protein overexpression overloads the ER protein processing capacity, which triggers activation of the unfolded protein response (UPR), a series of signal transduction pathways that mediate the expansion of the ER membrane, upregulation of protein folding and quality control pathways, and attenuation of the ER protein load (Fig. 1). Upon sustained ER stress, the UPR induces apoptosis. The UPR comprises a series of integrated signaling pathways controlled by three ER membrane proteins that function as stress sensors: inositol-requiring kinase 1 (IRE1), dsRNA-induced protein kinase-like ER kinase (PERK), and activating transcription factor 6 (ATF6). Under basal conditions, the three ER stress sensors are bond to BiP. Upon accumulation of misfolded protein, BiP dissociates ER stress sensors to bind the misfolded proteins, and the ER sensors activate. IRE1 activation mediates the non-conventional splicing of X-box binding protein 1 (XBP1) mRNA, resulting in the translation of the XBP1s transcription factor. XBP1s controls the expression of genes involved in protein folding, ER-associated protein degradation, and ER expansion. PERK activation mediates phosphorylation of eukaryotic translation initiator factor 2α (eIF2α), which results in stress attenuation signaling through inhibition of global protein translation and selective translation of activating transcription factor 4 (ATF4). Specifically, ATF4 controls the expression of genes involved in protein folding, amino acid metabolism, and oxidative stress resistance. Upon prolonged ER stress, however, ATF4 upregulates C/EBP-homologous protein (CHOP), which mediates apoptosis. ER stress also leads to the translocation and release of the active transcription factor ATF6, which upregulates ER quality control proteins. The relative dynamics of the different UPR signaling pathways (i.e., the timing and duration of the stress attenuation response and apoptosis induction) dictate cell fate decisions and whether homeostasis is restored or stress culminates with the induction of apoptosis. Specifically, cells surviving proteotoxic stress show an early activation of stress attenuation through the IRE1 pathway and a delay in UPR-induced apoptosis mediated by PERK¹⁶. IRE1-mediated splicing of XBP1 results in the expression of pro-survival genes^17,18. Moreover, cells surviving ER stress and restoring homeostasis display a faster rate of XBP1s production than dying cells and faster decay in XBP1 splicing, suggesting stress attenuation¹⁶. On the other hand, apoptosis is associated with reduced IRE1-XBP1 signaling (among other pathways) and activation of ATF4-mediated upregulation of CHOP^19–21. These studies suggest that not only the absolute expression of key UPR component but also the relative activation times of the IRE1-XBP1 and PERK-CHOP pathways are critical determinants of the shift from stress attenuation to apoptosis.

Fig. 1. Schematic overview of the UPR.

Under basal conditions, the three ER stress sensors are bound to BiP. Upon proteotoxic stress and accumulation of misfolded proteins, BiP dissociates from the ER stress sensors to bind to misfolded intermediates. The release of BiP results in the activation of the ER membrane sensors IRE1, PERK, and ATF6. The active form of IRE1 mediates the non-conventional splicing of the XBP1 mRNA, resulting in the translation of the active transcription factor, XBP1s, which controls the expression of genes involved in protein folding and protein quality control, ER-associated degradation, and phospholipid synthesis. The active form of PERK mediates phosphorylation of eIF2α, leading to inhibition of global protein translation and selective translation of ATF4. ATF4 initially upregulates genes involved in amino acid metabolism, resistance to oxidative stress, and autophagy, thus contributing to stress attenuation, and upon prolonged ER stress, regulates the expression of the pro-apoptotic transcription factor, CHOP. The activation of ATF6 results in translocation to the Golgi apparatus, cleavage, and release of active ATF6, which controls the expression of genes involved in protein folding and ER expansion.

Current strategies for overcoming protein folding and secretion bottlenecks are mainly based on the deregulated modulation of specific UPR genes. This approach has repeatedly resulted in limited improvements in recombinant protein yields as it may lead to metabolic burden, disruption of homeostatic systems, and cell adaptation^15,22,23. In addition, the deregulated modulation of protein folding pathways does not account for the population heterogeneity that characterizes cells engineered for high protein expression^24,25. These observations indicate the need for engineering cells that dynamically modulate UPR genes in response to fluctuations in proteotoxic stress and provide control at the single-cell level to address cell heterogeneity.

The design of cell factories that modulate the UPR to meet fluctuations in protein folding requirements and account for the natural heterogeneity of highly producing clones requires control systems that sense and respond to ER stress. Progress in synthetic biology has provided the tools to develop sophisticated feedback-controlled genetic circuits that can interface with innate signal transduction systems. Applying synthetic biology tools for engineering cells with sense-and-respond capabilities can overcome the limitations of current methods based on deregulated modulation of the UPR and ultimately generate dynamic cell factories with sustained production of therapeutic proteins. Here, we report the development of feedback-responsive cell factories that can sense proteotoxic stress and respond by modulating the IRE1 and PERK signaling pathways. We first built a cellular sensor of proteotoxic stress based on detecting IRE1 activation by linking the IRE1 target XBP1 to a user-defined output. This sensor was then deployed to build sense-and-respond systems that sense XBP1s levels and, in response, enhance stress attenuation singling through XBP1s amplification or delay apoptosis induction through downregulation of the pro-apoptotic regulator CHOP. We report the dynamic behavior of these sense-and-respond systems upon induction of ER stress. Our results demonstrate that combining enhancement of stress attenuation signaling and delay in apoptosis induction in cells overexpressing the model therapeutic protein tissue plasminogen activator (tPA) increases cell viability and the level of functional, secreted tPA. To evaluate the potential of this strategy as a platform to support the production of therapeutic proteins, we adapted our sense-and-respond system to enhance the production of the bispecific antibody blinatumomab. This innovative strategy enables the design of next-generation cell factories that dynamically adjust the innate cellular capacity to buffer proteotoxic stress, increasing cell viability and protein productivity and, in turn, improving the manufacturing of therapeutic proteins.

Results

Overexpression of tPA causes upregulation of cytoprotective and pro-apoptotic adaptive responses

To explore strategies for modulating the UPR in response to stress induced by overexpression of secretory proteins, we used the model protein human tPA. Human tPA is a 70 kDa serine protease containing three major N-glycosylation sites and 17 disulfide bonds requiring complex post-translational processing²⁶. Currently the only FDA-approved thrombolytic agent (alteplase) for the critical early treatment of acute ischemic strokes, tPA is used to treat acute thromboembolisms²⁷. Concerns regarding the potential effects of the recent global shortage of recombinant tPA have been recently exacerbated by the urgent need for tPA supplies to address COVID-19-associated acute respiratory distress syndrome, indicating a pressing need to maintain ample economic supplies of this emergency medication²⁸.

To evaluate the activation of the UPR and, more specifically, UPR signaling pathways mediating stress attenuation and apoptosis, we profiled the expression of three UPR marker genes, ERdj4, EIF4, and CHOP, in cells engineered to express tPA. ERdj4 is controlled by the transcription factor XBP1s and has been validated as a reporter of the IRE1 pathway, providing a measurement of stress attenuation²⁹. EIF4 and CHOP are controlled by the transcription factor ATF4 and are commonly monitored as reporters of the PERK pathway to quantify UPR-mediated translational control and apoptosis, respectively. Importantly, ERdj4 and EIF4 are specific targets of XBP1s and ATF4, respectively, and do not respond to crosstalk between the UPR branches²⁹. CHOP is a master regulator of UPR-induced apoptosis during prolonged ER stress and provides a specific readout of cytotoxicity associated with recombinant protein overexpression^30,31. To monitor ERdj4, EIF4, and CHOP expression, we used a set of reporter cell lines previously developed to quantify gene activity from the chromosomal context²⁹. In these reporter cell lines, the expression of a UPR target gene is linked to that of the green fluorescence protein (GFP) through an orthogonal genetic circuit that enables target gene profiling with high sensitivity and dynamic resolution of UPR induction (Fig. 2a)²⁹. Specifically, the expression of the UPR target gene is linked to that of a transcriptional regulator (the tetracycline-dependent transactivator, tTA) through an internal ribose entry site (IRES) to maintain a constant expression ratio between the UPR target gene and tTA^32,33. GFP expression is activated by tTA, mediating signal output amplification. GFP expression is also negatively regulated transcriptionally by the erythromycin-dependent Krüppel-associated box repressor (EKRAB) and post-translationally via proteasomal degradation mediated by the NanoDeg³⁴. EKRAB and the NanoDeg are repressed by tTA, and thus activated under basal conditions to mediate GFP repression and rapid GFP protein depletion, respectively, ultimately enhancing the output dynamic range and the dynamic resolution of the target gene expression. To evaluate the effect of tPA expression on ERdj4, EIF4, and CHOP activity, we produced lentiviral particles expressing the gene encoding tPA (hPLAT) linked to the gene encoding the near-infrared fluorescent protein (iRFP) through an IRES. Lentiviral transductions were conducted by varying the number of viral particles expected to affect transgene delivery and, ultimately, tPA expression levels (low, medium, and high). Given that the probability of infection can be estimated using a Poisson distribution, transductions were conducted using a multiplicity of infection (MOI) expected to minimize the probability of cells with multiple infections (MOI = 0.35, low), maximize the probability of single infections (MOI = 1, mid), and maximize the probability of multiple infections (MOI = 35, high). To verify tPA expression levels and assess the correlation between hPLAT expression and tPA protein production, we transduced the UPR reporters with viral particles encoding hPLAT and iRFP using different MOIs. We monitored the iRFP signal as an indication of transgene expression levels and tPA activity as an indication of functional tPA production. The iRFP mean fluorescence of transduced (iRFP-positive) cells was measured using flow cytometry, and tPA activity was monitored by quantifying the cleavage of a fluorogenic substrate as reported^35,36. We observed a correlation between the iRFP signal and transgene delivery. Specifically, the iRFP mean fluorescence increased by 32% and 60% after 8 days of transduction with medium and high MOI compared to low MOI (Fig. 2b). We observed a correlation between iRFP and hPLAT mRNA levels, the iRFP signal, and transduction conditions (i.e., MOI), indicating that transgene expression depends on the transgene delivery (Supplementary Fig. 1a-c). Interestingly, cells transduced with a medium MOI displayed an 84% increase in tPA activity levels compared to cells transduced with a low MOI (Fig. 2c). However, tPA activity levels did not increase further upon cell transduction with a high MOI. These results suggest that increasing transgene delivery results in an increase in hPLAT expression but does not parallel the expected increase in functional tPA production, as previously reported^7,10.

Fig. 2. Cell factories of tPA present UPR activation.

a Schematic of the UPR gene signal amplifier. IRES-tTA is integrated 3’ of the target gene to link tTA expression to the UPR target. GFP is controlled by a hybrid promoter comprising the tTA-specific operator (TRE), the minimal CMV promoter, and the EKRAB-specific operator (ETR). EKRAB and the NanoDeg are repressed by tTA (CMV-TO). The NanoDeg mediates GFP degradation. b iRFP signal of HEK293 cells with the gene signal amplifier described in a for monitoring the ERdj4 gene, transduced with virus expressing tPA and iRFP at MOIs of 0.35 (low), 1 (mid), and 35 (high), eight days post-transduction. (c) tPA activity of cells transduced as in b. d–f Mean iRFP fluorescence of iRFP-positive HEK293 cells with the gene signal amplifier linked to ERdj4 (d), EIF4 (e), and CHOP (f) and transduced with virus expressing tPA and iRFP at MOIs of 0.35 (low), 1 (mid), and 35 (high). g–i UPR induction index of HEK293 cells engineered with the gene signal amplifier linked to ERdj4 (g), EIF4 (h), and CHOP (i) and transduced with virus expressing tPA and iRFP at MOIs of 0.35 (low), 1 (mid), and 35 (high). The UPR induction index was obtained by multiplying the mean GFP fluorescence by the fraction of cells expressing GFP. j Representative flow cytometry analyses of HEK293 cells engineered with the gene signal amplifier linked to EIF4 transduced with virus expressing tPA and iRFP at an MOI of 35 (high) depicting gating boundaries used to define the “low” and “high” GFP subpopulations within iRFP-positive cells. k–m iRFP signal of HEK293 cells engineered with the gene signal amplifier linked to ERdj4 (k), EIF4 (l), and CHOP (m) and transduced with virus expressing tPA and iRFP at an MOI of 35 (high) four days post-transduction. The “low” and “high” GFP subpopulations were defined as shown in j. Data are reported as mean ± s.e.m. (n = 3 biological replicates). P-values were calculated using an unpaired Student’s t-test with a two-tailed analysis. Source data are provided as a Source Data file.

To monitor ERdj4, EIF4, and CHOP activity upon expression of tPA, ERdj4, EIF4, and CHOP reporter cells were transduced with lentiviral particles expressing tPA/iRFP at different MOIs and analyzed by flow cytometry. Flow cytometry measurements verified the correlation between the iRFP signal and the transduction MOI over time (Fig. 2d-f), suggesting low, intermediate, and high levels of tPA expression. To evaluate the expression of UPR target genes in cells expressing tPA, we quantified the GFP signal in iRFP-positive cells as a function of time. Specifically, we evaluated the UPR induction index of iRFP-positive cells, which accounts for the fraction of the cell population with UPR target gene upregulation and the extent of UPR target gene expression. We observed a correlation between ERdj4 expression and transgene delivery. ERdj4 induction was found to peak after 4-6 days of transduction and display a 2- and 6-fold increase in cells transduced with medium and high MOI, respectively, compared to cells transduced with a low MOI (Fig. 2g), indicating higher activation of the IRE1 pathway in cells with higher transgene expression. After prolonged tPA expression, the GFP signal of ERdj4 reporter cells transduced with a low MOI returned to levels comparable to those observed at initial times (compare 10 days to 2 days post-transduction). The GFP signal of cells transduced with medium or high MOIs was significantly higher than the corresponding measurements at initial transduction times, suggesting a prolonged upregulation of the IRE1 pathway at higher levels of tPA expression. The EIF4 reporter cells presented a similar behavior: EIF4 expression peaked 4 days post-transduction in an MOI-dependent fashion (Fig. 2h), suggesting that the activation of the PERK pathway depends on the tPA expression levels and that it is sustained at high tPA expression levels. Measurements of CHOP reporter cells revealed minimal CHOP expression in cells transduced with low or medium MOIs (Fig. 2i), suggesting low UPR-induced apoptosis. Transductions with high MOI caused an increase in GFP signal that peaked 6 days post-transduction, resulting in a 7-fold increase in the UPR induction index compared to cells transduced with a low MOI, suggesting activation of UPR-mediated apoptosis in response to high levels of tPA expression. In summary, these results suggest that the UPR is induced in cells expressing tPA to an extent that depends on tPA expression level. Interestingly, we found the fraction of cells displaying UPR induction to decrease with the time of transgene expression (Supplementary Fig. 2a-c), suggesting that the cell population presenting UPR activation is heterogenous and that a fraction of this population recovers from proteotoxic stress. We also found the fraction of cells displaying UPR induction to maintain constant expression levels of the UPR marker gene monitored in this study (Supplementary Fig. 2d-f).

To further evaluate the correlation between tPA expression and UPR induction, we quantified transgene expression levels in subpopulations of cells with different extents of UPR induction. To this end, we quantified the iRFP mean fluorescence of subpopulations of the UPR reporter cells transduced with a high MOI 4 days post-transduction. Transduced (iRFP-positive) cells were gated to select GFP-positive cells (Fig. 2j). High and low GFP signals were defined by gating subpopulations corresponding to the lowest 20% and highest 20% of the GFP-positive distribution. We observed a correlation between transgene expression, evaluated by quantifying the iRFP signal, and UPR induction, evaluated by quantifying the GFP signal, suggesting that the extent of tPA expression affects the upregulation of the UPR target genes ERdj4, EIF4, and CHOP. Specifically, we found the “high GFP” subpopulations of the ERdj4, EIF4, and CHOP reporter cell lines to present, respectively, 27%, 39%, and 27% higher transgene expression compared to the corresponding “low GFP” subpopulations (Fig. 2k-m).

A cellular sensor for detection of the IRE1 pathway activation

To engineer host cell lines with enhanced tPA productivity, we sought to modulate the UPR to improve the attenuation of proteotoxic stress. To this end, we explored feedback-controlled systems for modulating the expression of key URP target genes in response to the detection of ER stress. Such an approach would enable fluctuations in folding requirements to be addressed and the UPR controlled at the single-cell level, thereby addressing population heterogeneities. To develop sense-and-response systems that modulate the UPR upon detecting ER stress, we first generated a cellular sensor of IRE1, a master regulator of the UPR, activated upon accumulation of misfolded proteins in the ER lumen. Activated IRE1 mediates the non-conventional splicing of XBP1 mRNA, resulting in the translation of an active transcription factor (XBP1s) that controls the expression of a network of genes, including ERdj4. To create an IRE1 sensor, we engineered HEK293 cells to link XBP1 splicing to the expression of the master regulator, tTA. Such a strategy allows interfacing the IRE1 pathway with user-controlled orthogonal genetic networks designed to respond to tTA control. Specifically, we integrated a cassette containing the tTA gene preceded by the coding sequence of a 2 A self-cleaving peptide (P2A) from porcine teschovirus downstream of the XBP1 ER stress-specific intron using CRISPR-Cas9 mediated editing (Fig. 3a)³⁷. Upon activation of IRE1, splicing of a 26 bp intron from the XBP1_P2A_tTA mRNA leads to a frameshift and results in the translation of XBP1s fused to tTA through the 2 A peptide³⁸. Self-cleavage of the 2A peptide results in equal expression levels of XBP1s and tTA. The resulting cells (HEK293/tTA) were analyzed by genomic PCR to verify the knock-in mutation at the desired target site (Supplementary Fig. 3a, b). To verify that HEK293/tTA cells are engineered to link the expression of tTA to that of spliced XBP1, we evaluated tTA activity upon induction of ER stress. HEK293/tTA cells were transiently transfected with a plasmid encoding GFP under the control of the tetracycline-responsive element (TRE) promoter that responds to tTA activation. ER stress was induced using tunicamycin, a canonical UPR inducer, which affects glycoprotein processing^39,40. Flow cytometry analyses of transfected cells treated with tunicamycin (1 μg/ml) revealed a 2-fold increase in GFP fluorescence compared to untreated cells (Fig. 3b). The addition of the tTA inhibitor tetracycline (1 μg/ml) to cells treated with tunicamycin resulted in a 74% reduction in GFP fluorescence, suggesting that the GFP signal induced by ER stress is mediated by tTA. Interestingly, adding tetracycline at a dose expected to mediate complete inhibition of tTA resulted in a 50% decrease in the GFP output compared to untreated cells (Fig. 3b), indicative of XBP1 splicing under resting conditions, as previously reported⁴¹. To verify that the integration of tTA downstream of XBP1 does not affect the activity of the IRE1 pathway, we monitored the mRNA levels of spliced XBP1 and its downstream target gene ERdj4 in HEK293 and HEK293/tTA cells induced with tunicamycin (1 μg/ml) for 24 hours by using quantitative RT-PCR. HEK293/tTA and parental HEK293 cells treated with tunicamycin displayed comparable XBP1s and ERdj4 mRNA levels (p > 0.1, Student’s t-test) (Supplementary Fig. 3c, d), indicating that chromosomal integration of tTA does not affect the native IRE1-XBP1s signaling.

Fig. 3. Engineering a cellular sensor of the IRE1 pathway.

a Schematic representation of the IRE1 sensor (HEK293/tTA cells) built by integrating the P2A_tTA cassette downstream of the XBP1 splice site in the genome of HEK293 cells. b GFP signal of HEK293/tTA cells transfected with a plasmid encoding GFP under the control of tTA, treated with tunicamycin (1 μg/mL) and tetracycline (1 μg/mL), and measured by flow cytometry. Data are reported as mean ± s.e.m. (n = 3 biological replicates). P-values were calculated using an unpaired Student’s t-test with a two-tailed analysis. Source data are provided as a Source Data file.

Feedback-regulated control of XBP1s expression amplifies XBP1s signaling

To generate a sense-and-response system that amplifies the cytoprotective effect of the UPR upon ER stress, we engineered the IRE1 sensor (HEK293/tTA cells) with a positive feedback loop resulting in overexpression of the IRE1 pathway master regulator, XBP1s, upon detection of IRE1 activation. We integrated a cassette encoding spliced XBP1 (XBP1s) and a fluorescent reporter (iRFP) linked by an IRES and under the control of the TRE promoter into HEK293/tTA cells using lentivirus transduction (Fig. 4a). In the resulting cells, the splicing of endogenous XBP1 and the expression of recombinant XBP1s are linked by tTA, which interfaces the XBP1s amplification loop with IRE1 activation. Transduced cells were selected and diluted to isolate single clones (HEK293/tTA_XBP1s). The resulting monoclonal populations were treated with tunicamycin (1 μg/ml, 48 hours), and iRFP fluorescence was measured using flow cytometry (Supplementary Fig. 4). Two representative clones (HEK293/tTA_XBP1s clones #1 and #2) that present different extents of XBP1s self-amplification upon treatment with tunicamycin, as quantified by monitoring the iRFP fold change, were selected for further studies (Supplementary Fig. 4, blue bars). Treatment of HEK293/tTA_XBP1s #1 and #2 with tunicamycin (10 μg/ml, 48 hours) resulted in a 2.4- and 7.6-fold increase in iRFP fluorescence, respectively, compared to untreated cells (Fig. 4b, c), which is indicative of the expression of the XBP1s_IRES_iRFP cassette in response to UPR induction. The addition of tetracycline (1 μg/ml) to tunicamycin-treated cells resulted in more than 60% reduction in iRFP fluorescence in both HEK293/tTA_XBP1s #1 and #2, suggesting that the iRFP signal induced by ER stress is mediated by tTA. As previously observed in the parental HEK293/tTA cell line, the addition of tetracycline at a concentration expected to mediate complete inhibition of tTA resulted in a 24% and 39% decrease in the iRFP fluorescence in HEK293/tTA_XBP1s #1 and #2 cells, respectively, compared to untreated cells, which is indicative of splicing of XBP1 at basal conditions⁴¹.

Fig. 4. Feedback-regulated amplification of the IRE1 pathway master regulator XBP1s.

a Schematic representation of the sense-and-respond system for amplifying XBP1s expression in response to XBP1 splicing. The sense-and-respond system (HEK293/tTA_XBP1s cells) was built by integrating a cassette encoding XBP1s and iRFP linked by an IRES, under the control of tTA in the genome of the IRE1 sensor (HEK293/tTA cells). b, c iRFP signal of HEK293/tTA_XBP1s cells clones #1 (b) and #2 (c) treated with tunicamycin (10 μg/mL) and tetracycline (1 μg/mL) for 48 hours and measured by flow cytometry. iRFP fold change values were obtained by normalizing the iRFP fluorescence intensity of treated cells to that of untreated cells. d, e Quantitative RT-PCR analyses of XBP1s (d) and ERdj4 (e) mRNA levels of HEK293 and HEK293/tTA_XBP1s cells treated with tunicamycin (10 μg/mL). Relative mRNA expression levels were obtained by normalizing the mRNA expression levels of treated cells to those of untreated cells. f MTS absorbance of HEK293 and HEK293/tTA_XBP1s cells treated with tunicamycin (10 μg/mL) relative to untreated cells. Data are reported as mean ± s.e.m. (n = 3 biological replicates). P values were calculated using an unpaired Student’s t-test with a two-tailed analysis. Source data are provided as a Source Data file.

To characterize the behavior of the XBP1s sense-and-respond system upon ER stress, we monitored XBP1s mRNA levels in HEK293/tTA_XBP1s cells induced with tunicamycin. HEK293/tTA_XBP1s #1 and #2 and parental HEK293 cells were treated with tunicamycin (10 μg/ml), and the mRNA levels of XBP1s were measured using quantitative RT-PCR. HEK293 cells induced with tunicamycin, used here as a control, displayed a steady increase in XBP1s mRNA, from a 5-fold change after 12 hours of treatment to a 14-fold change after 48 hours of treatment compared to untreated cells (Fig. 4d). HEK293/tTA_XBP1s #1 cells displayed a 22-fold increase in XBP1s mRNA levels after 12 hours of treatment, which decreased by 78% after 36 hours and remained at constant low levels for the next 12 hours. HEK293/tTA_XBP1s #2 displayed a 35-fold increase in XBP1s mRNA levels after 12 hours of treatment, which peaked after 24 hours, reaching a 59-fold increase, and decreased to levels comparable to those observed in HEK293/tTA_XBP1s #1 after 48 hours. The dramatic increase in XBP1s mRNA levels of HEK293/tTA_XBP1s clones compared to parental HEK293 cells after 12-24 hours of tunicamycin treatment is indicative of self-amplification of XBP1s expression upon ER stress. This increase in XBP1s expression correlates with the fold change in iRFP signal (Fig. 4b, c), confirming that XBP1s expression depends on the feedback amplification loop. Notably, the XBP1s mRNA levels of both HEK293/tTA_XBP1s clones undergo a more dramatic decrease than the XBP1s mRNA levels of the parental HEK293 exposed to the same treatment (compare relative XBP1s mRNA levels at 48 hours, Fig. 4d), suggesting a feedback mechanism in which the XBP1s amplification loop leads to attenuation of the stress response.

To further characterize the XBP1s self-amplification response, we monitored the mRNA levels of ERdj4, an XBP1s downstream target, in HEK293/tTA_XBP1s cells induced with tunicamycin. HEK293 cells treated with tunicamycin (10 μg/ml) displayed an increase in ERdj4 mRNA levels that reached a 6-fold change after 48 hours of treatment compared to untreated cells (Fig. 4e). HEK293/tTA_XBP1s #1 displayed a peak 5-fold increase in ERdj4 mRNA levels after 12 hours of treatment, which decreased by 51% after 36 hours and remained constant for the next 12 hours. HEK293/tTA_XBP1s #2 displayed a 7-fold increase in ERdj4 mRNA levels after 24 hours of treatment, which decreased by 42% after 48 hours. Interestingly, the extent of upregulation of XBP1s and ERdj4 expression in the HEK293/tTA_XBP1s clones correlates with the iRFP fold change, with the clone presenting a larger amplification of XBP1s as quantified by the iRFP signal (#2) displaying a larger increase in XBP1s and ERdj4 mRNA levels. The correlation between ERdj4 and XBP1s mRNA levels in the HEK293/tTA_XBP1s clones suggests that the XBP1s self-amplification loop mediates the upregulation of ERdj4 observed in these cells.

Given the previously reported cytoprotective effects of IRE1 signaling and, particularly, XBP1s overexpression, we also evaluated the impact of the amplification loop on HEK293/tTA_XBP1s cell viability. HEK293/tTA_XBP1s #1 and #2 and parental HEK293 cells were treated with tunicamycin (10 μg/ml, 48 hours), and cell viability was monitored using the MTS assay. HEK293/tTA_XBP1s clones #1 and #2 displayed a 42% and 55% increase in viability, respectively, compared to HEK293 cells (Fig. 4f). The higher cell viability in the HEK293/tTA_XBP1s clones subjected to ER stress than in parental cells indicates that amplifying the IRE1 response attenuates ER stress.

Feedback-regulated silencing of CHOP response delays CHOP signaling and apoptosis

To generate a sense-and-response system that delays the activation of UPR-induced apoptosis, typically observed upon sustained ER stress, we engineered the IRE1 sensor (HEK293/tTA cells) to control a feedback loop mediating CHOP silencing. We designed a short hairpin RNA against CHOP (shCHOP) to mediate CHOP knockdown. To verify that the expression of shCHOP leads to CHOP silencing, we evaluated CHOP expression using the CHOP reporter cell line previously developed to link the expression of CHOP to GFP production. CHOP reporter cells were transiently transfected with a plasmid encoding shCHOP and treated with tunicamycin (10 μg/ml, 1 hour). Flow cytometry analysis of cells transfected with shCHOP revealed a 69% decrease in GFP fluorescence compared to cells transfected with a nontargeting control (Supplementary Fig. 5), indicating that shCHOP mediates knockdown of CHOP expression. To link the expression of shCHOP to activation of the IRE1 sensor, we generated an expression cassette containing shCHOP flanked by the genomic microRNA miR-30 and inserted within the 3’UTR of iRFP under the control of the TRE promoter^42–46. HEK293/tTA cells were transduced with viral particles encoding the iRFP-shCHOP cassette expected to generate a tTA-inducible system expressing constant ratios of shCHOP and iRFP (Fig. 5a). In the resulting cells, the splicing of endogenous XBP1 and the expression of shCHOP are linked by tTA, which interfaces IRE1 activation with CHOP silencing. Stable cell lines were expanded and diluted to isolate monoclonal populations (HEK293/tTA_shCHOP). The resulting monoclonal populations were treated with tunicamycin (1 μg/ml, 48 hours), and the iRFP fluorescence was measured using flow cytometry (Supplementary Fig. 6). The monoclonal population presenting the highest fold change in iRFP fluorescence upon treatment with tunicamycin was selected for further studies (Supplementary Fig. 6, green bar).

Fig. 5. Feedback-regulated silencing of the CHOP response.

a Schematic representation of the sense-and-respond system for silencing CHOP expression in response to XBP1 activation. The sense-and-respond system (HEK293/tTA_shCHOP cells) was built by integrating a cassette encoding iRFP followed by the miR30-based anti-CHOP short hairpin RNA (shCHOP) under the control of tTA into the genome of the IRE1 sensor (HEK293/tTA cells). b iRFP signal of HEK293/tTA_shCHOP cells treated with tunicamycin (10 μg/mL) and tetracycline (1 μg/mL) for 48 hours and measured by flow cytometry. iRFP fold change values were obtained by normalizing the iRFP fluorescence intensity of treated cells to that of untreated cells. c Quantitative RT-PCR analyses of iRFP-shCHOP RNA levels of HEK293/tTA_shCHOP cells treated with tunicamycin (10 μg/mL) and tetracycline (10 μg/mL). Relative RNA expression levels were obtained by normalizing the RNA expression levels of treated cells to that of untreated cells at initial conditions (0 h). d, e Quantitative RT-PCR analyses of CHOP (d) and ATF5 (e) mRNA levels of HEK293 and HEK293/tTA_shCHOP cells treated with tunicamycin (10 μg/mL). Relative mRNA expression levels were obtained by normalizing the mRNA expression levels of treated cells to that of untreated HEK293 cells at initial conditions (0 h). f Representative flow cytometry histograms of Annexin V-FITC-stained HEK293 and HEK293/tTA_shCHOP cells untreated or treated with tunicamycin (10 μg/mL) for 48 hours. Data are reported as mean ± s.e.m. (n = 3 biological replicates). P-values were calculated using an unpaired Student’s t-test with a two-tailed analysis. Source data are provided as a Source Data file.

To verify that activation of the CHOP silencing response depends on UPR induction and is controlled by tTA activity, we monitored the iRFP fluorescence of HEK293/tTA_shCHOP cells upon tunicamycin and tetracycline treatment. Tunicamycin treatment (10 μg/ml, 48 hours) resulted in a 2.7-fold increase in iRFP fluorescence compared to untreated cells (Fig. 5b), which indicates expression of the iRFP-shCHOP cassette in response to UPR induction. Upon the addition of tetracycline (1 μg/ml), tunicamycin-treated cells displayed a 70% decrease in iRFP fluorescence levels compared to tunicamycin-only treated cells and a 19% decrease in iRFP fluorescence levels compared to untreated cells, indicating that the increase in iRFP expression upon UPR induction depends on tTA activity and suggesting basal expression of tTA at resting condition likely due to basal levels of XBP1 splicing. To confirm expression of the iRFP-shCHOP transcript, we monitored the RNA levels of iRFP-shCHOP in HEK293/tTA_shCHOP cells treated with tunicamycin (10 μg/ml) and tetracycline (10 μg/ml) by quantitative RT-PCR using primers designed to amplify from the 3’ end of iRFP to the 3’ end of miR-30 (Supplementary Table 1). We observed an 8-fold increase in the iRFP-shCHOP RNA levels after 12 hours of treatment with tunicamycin compared to untreated cells, which decreased by 43% after 48 hours (Fig. 5c). iRFP-shCHOP RNA levels in cells treated with tetracycline were 74% lower than in untreated cells, confirming tTA control and basal activation of XBP1 splicing, as seen before (Fig. 5b). These results demonstrate that these cells induce CHOP silencing upon ER stress and that shCHOP expression is mediated by tTA, confirming that tTA interfaces the IRE1 pathway with silencing of the pro-apoptotic pathway.

To verify that the activation of the CHOP silencing circuit in response to XBP1 activation leads to the knockdown of CHOP mRNA, we monitored CHOP mRNA levels upon cell treatment with tunicamycin using quantitative RT-PCR. HEK293 cells displayed a steady increase in CHOP mRNA levels upon tunicamycin treatment (10 μg/ml), specifically a 15-fold increase after 12 hours and a 43-fold increase after 48 hours (Fig. 5d). HEK293/tTA_shCHOP cells displayed a similar increase in CHOP mRNA levels after the first 12 hours of tunicamycin treatment but remained constant after 24 hours and 36 hours of treatment (p > 0.1) and displayed only a mild increase after 48 hours. The limited change in CHOP mRNA levels observed in HEK293/tTA_shCHOP cells treated with tunicamycin over time, which reached less than half of the CHOP mRNA levels of parental HEK293 cells, indicates that the expression of shCHOP upon UPR induction results in the knockdown of CHOP expression.

To further characterize the extent to which activation of the CHOP silencing circuit inhibits CHOP-mediated signaling, we also monitored the mRNA levels of the CHOP downstream target ATF5 upon cell exposure to tunicamycin. HEK293 cells treated with tunicamycin (10 μg/ml) displayed a 2.5-fold increase in ATF5 mRNA levels after 12 hours and reached a 5.5-fold increase after 48 hours compared to untreated cells (Fig. 5e). HEK293/tTA_shCHOP cells displayed a 2.2-fold increase in ATF5 mRNA levels after 12 hours of tunicamycin treatment and reached a maximum 3-fold increase in ATF5 mRNA levels after 48 hours (compared to the 5.5-fold increase in HEK293 cells). Similar to CHOP mRNA levels, ATF5 mRNA levels in HEK293/tTA_shCHOP cells were found to be about 50% lower than in parental HEK293 cells, indicating that activation of CHOP silencing in response to the detection of XBP1 activation leads to the suppression of CHOP downstream signaling.

To evaluate the effect of the CHOP silencing circuit on cellular apoptosis, we monitored the membrane rearrangement characteristic of early apoptosis in HEK293 cells and cells engineered with the CHOP silencing circuit upon ER stress. HEK293 and HEK293/tTA_shCHOP cells were treated with tunicamycin (10 μg/ml, 48 hours), and early apoptosis was measured using flow cytometry by quantifying the binding of Annexin V to phosphatidylserine, a residue located in the inner leaflet of the plasma membrane that becomes exposed to the outer cell surface during early apoptosis. Treatment of HEK293 cells with tunicamycin increased the Annexin V-positive population compared to untreated cells (Fig. 5f and Supplementary Fig. 7). The Annexin V-positive population of HEK293/tTA_shCHOP cells was found not to increase upon tunicamycin treatment and, notably, to be comparable to that of parental HEK293 cells. These results suggest that activation of the CHOP silencing circuit reduces the activation of UPR-induced apoptosis observed upon ER stress.

Feedback-regulated control of XBP1s expression and CHOP silencing amplifies XBP1s signaling and reduces CHOP response

We asked whether combining the amplification of XBP1s signaling with inhibition of the CHOP-mediated pro-apoptotic response would provide a more robust cytoprotective response to ER stress, eventually enabling cells to adapt to the demands associated with high expression of recombinant proteins. To address this question, we engineered the IRE1 sensor (HEK293/tTA cells) with a feedback response mediating XBP1s overexpression and CHOP silencing. We designed an expression cassette containing XBP1s followed by an IRES and iRFP, with shCHOP inserted at the 3’UTR of iRFP. The expression cassette was placed under the control of the TRE promoter to enable activation by tTA (Fig. 6a). The resulting construct, XBP1s_IRES_iRFP-shCHOP, was integrated into the IRE1 sensor using lentivirus transduction. Stable cell lines were expanded and diluted to isolate monoclonal populations (HEK293/tTA_XBP1s_shCHOP). Single HEK293/tTA_XBP1s_shCHOP clones were treated with tunicamycin (1 μg/ml, 48 hours), and the iRFP fluorescence was measured using flow cytometry (Supplementary Fig. 8). Two clones (#1.1 and #2.1) were selected that display a change in iRFP signal upon tunicamycin treatment similar to the HEK293/tTA_XBP1s clones #1 and #2, respectively (Supplementary Fig. 8, orange bars). To verify that HEK293/tTA_XBP1s_shCHOP cells respond to UPR activation and that tTA mediates this response, we monitored the iRFP fluorescence of clones #1.1 and #2.1 upon tunicamycin and tetracycline treatment. Tunicamycin treatment (10 μg/ml, 48 hours) resulted in a 2.8- and 6.9-fold increase in iRFP fluorescence of HEK293/tTA_XBP1s_shCHOP #1.1 and #2.1, respectively (Fig. 6b, c). Adding tetracycline (1 μg/ml) to tunicamycin-treated cells decreased the iRFP fluorescence to values lower than those of untreated cells, as observed before (Figs. 4b, c, 5b). These results indicate that expression of the XBP1s_IRES_iRFP-shCHOP cassette is activated upon UPR induction and is mediated by tTA.

Fig. 6. Feedback-regulated amplification of XBP1s expression and CHOP silencing.

a Schematic representation of the sense-and-respond system for amplifying XBP1s and silencing CHOP expression in response to XBP1 splicing. The sense-and-respond system (HEK293/tTA_XBP1s_shCHOP cells) was built by integrating a cassette encoding XBP1s, an IRES, and iRFP followed by the miR30-based anti-CHOP shRNA (shCHOP) under the control of tTA in the genome of the IRE1 sensor (HEK293/tTA cells). b,c iRFP signal of HEK293/tTA_XBP1s_shCHOP clones #1.1 (b) and #2.1 (c) treated with tunicamycin (10 μg/mL) and tetracycline (1 μg/mL) for 48 hours and measured by flow cytometry. iRFP fold change values were obtained by normalizing the iRFP fluorescence values of treated cells to that of untreated cells. d,e Quantitative RT-PCR analyses of XBP1s (d) and ERdj4 (e) mRNA levels of HEK293 and HEK293/tTA_ XBP1s_shCHOP cells treated with tunicamycin (10 μg/mL). Relative mRNA expression levels were obtained by normalizing the mRNA levels of treated cells to the expression levels of untreated cells. f Quantitative RT-PCR analyses of iRFP-shCHOP RNA levels of HEK293/tTA_XBP1s_shCHOP cells treated with tunicamycin (10 μg/mL). Relative RNA expression levels were obtained by normalizing the RNA levels of treated cells to the expression levels of untreated HEK293/tTA_XBP1s_shCHOP #1.1 cells at initial conditions (0 h). g,h Quantitative RT-PCR analyses of CHOP (g) and ATF5 (h) mRNA levels of HEK293 and HEK293/tTA _XBP1s_shCHOP cells treated with tunicamycin (10 μg/mL). Relative mRNA expression levels were obtained by normalizing the mRNA levels of treated cells to that of untreated HEK293 cells at initial conditions (0 h). f MTS absorbance of HEK293 and HEK293/tTA_XBP1s_shCHOP cells treated with tunicamycin (10 μg/mL) relative to untreated cells. Data are reported as mean ± s.e.m. (n = 3 biological replicates). P values were calculated using an unpaired Student’s t-test with a two-tailed analysis. Source data are provided as a Source Data file.

To characterize the behavior of the sense-and-response system designed to amplify XBP1s and shCHOP expression upon ER stress, we first monitored XBP1s and ERdj4 mRNA levels. Specifically, HEK293/tTA_XBP1s_shCHOP #1.1 and #2.1 and parental HEK293 cells were treated with tunicamycin (10 μg/ml), and the mRNA levels were measured using quantitative RT-PCR. Parental HEK293 cells displayed a steady increase in XBP1s mRNA levels that reached a 7-fold change after 48 hours of tunicamycin treatment (Fig. 6d). HEK293/tTA_XBP1s_shCHOP #1.1 displayed a 6.5-fold increase in XBP1s mRNA levels after 12 hours of treatment, which decreased by 43% after 36 hours and remained constant for the next 12 hours. HEK293/tTA_XBP1s_shCHOP #2.1 displayed a 24-fold increase in XBP1s mRNA levels after 24 hours of treatment and decreased to levels comparable to untreated cells after 48 hours. Similar to XBP1s mRNA levels, HEK293 cells displayed a constant increase in ERdj4 mRNA levels that reached a 4.1-fold change after 48 hours of tunicamycin treatment (Fig. 6e). HEK293/tTA_XBP1s_shCHOP #1.1 displayed a peak 4.4-fold increase in ERdj4 mRNA levels after 12 hours of treatment that decreased by 19% after 48 hours. HEK293/tTA_XBP1s_shCHOP #2.1 displayed a 13-fold increase in ERdj4 mRNA levels after 24 hours of treatment, which decreased by 87% after 48 hours, reaching ERdj4 mRNA levels lower than those of parental HEK293 cells. The increase in XBP1s and ERdj4 mRNA levels of HEK293/tTA_XBP1s_shCHOP clones compared to parental HEK293 cells after 12-24 hours of tunicamycin treatment is indicative of self-amplification of XBP1s expression upon ER stress, and the extent of upregulation of XBP1s and ERdj4 expression in the HEK293/tTA_XBP1s_shCHOP clones reflects the change in iRFP signal, with the clone presenting a larger iRFP fold change (#2.1) displaying a larger increase in XBP1s and ERdj4 mRNA levels.

To evaluate the amplification of the CHOP silencing response in HEK293/tTA_XBP1s_shCHOP, we first monitored iRFP-shCHOP RNA levels by quantitative RT-PCR using primers designed to amplify from the 3’ end of iRFP to the 3’ end of miR-30 (Supplementary Table 1). HEK293/tTA_XBP1s_shCHOP #1.1 displayed 55-fold increase in iRFP-shCHOP RNA levels after 12 hours of tunicamycin treatment (10 μg/ml), which decreased mildly at later times (Fig. 6f). HEK293/tTA_XBP1s_shCHOP #2.1 displayed a 3.2-fold increase in the iRFP-shCHOP RNA levels compared to clone #1.1 at basal conditions (Fig. 6f, 0 hours). Clone #2.1 displayed a 174-fold increase in iRFP-shCHOP RNA levels after 12 hours of treatment, which decreased by 89% after 48 hours. Notably, the extent of iRFP-shCHOP RNA expression in the HEK293/tTA_XBP1s_shCHOP clones reflects the change in iRFP signal, with the clone presenting a larger amplification of the feedback response (#2.1) displaying a larger increase in iRFP-shCHOP RNA levels.

To verify that the expression of shCHOP in response to ER stress leads to the knockdown of CHOP mRNA, we monitored CHOP and ATF5 mRNA levels upon cell treatment with tunicamycin (10 μg/ml). HEK293 cells displayed a steady increase in CHOP mRNA levels upon tunicamycin treatment, from a 16-fold increase after 12 hours of treatment to a 40-fold increase after 48 hours (Fig. 6g). HEK293/tTA_XBP1s_shCHOP #1.1 displayed CHOP mRNA levels comparable to parental cells at early time points and mildly lower at later time points. HEK293/tTA_XBP1s_shCHOP #2.1 displayed a 25-fold increase in CHOP mRNA levels after 12 hours of treatment, which decreased by 58% after 36 hours of treatment and remained constant for the next 12 hours, about 3-fold lower than in parental cells. Similar to CHOP mRNA levels, HEK293 cells displayed a 2.1-fold increase in ATF5 mRNA levels after 12 hours of treatment and reached a 4.5-fold increase after 48 hours (Fig. 6h). HEK293/tTA_XBP1s_shCHOP #1.1 displayed comparable ATF5 mRNA levels (p > 0.1) to HEK293 cells upon tunicamycin treatment. HEK293/tTA_XBP1s_shCHOP #2.1 did not display an increase in ATF5 mRNA levels upon tunicamycin treatment, which remained dramatically lower than in parental cells. The significantly lower CHOP and ATF5 mRNA levels observed after 48 hours of tunicamycin treatment in the clone presenting a larger amplification of iRFP-shCHOP transcript levels (#2.1) indicates that the expression of high levels of shCHOP upon UPR induction results in the knockdown of CHOP expression and suppression of its downstream signaling.

To evaluate the cytoprotective effect of the sense-and-respond system mediating XBP1s upregulation and CHOP silencing, we monitored the viability of HEK293/tTA_XBP1s_shCHOP cells. HEK293/tTA_XBP1s_shCHOP #1.1 and #2.1 and parental HEK293 cells were treated with tunicamycin (10 μg/ml, 48 hours), and cell viability was monitored using the MTS assay. HEK293/tTA_XBP1s_shCHOP #2.1 displayed a 42% increase in viability compared to HEK293 cells, while HEK293/tTA_XBP1s_shCHOP #1.1 displayed levels comparable to parental cells (Fig. 6i). The higher cell viability in the HEK293/tTA_XBP1s_shCHOP clone with high expression of the feedback response compared to parental cells indicates that amplification of XBP1s combined with CHOP silencing attenuates ER stress.

Feedback-regulated control of XBP1s expression and CHOP silencing enhances tPA production

We asked whether the feedback control systems mediating stress attenuation through self-amplification of XBP1s signaling and apoptosis delay through CHOP silencing affect the production of the secretory protein tPA. We transduced HEK293 cells and the three derivative cell lines engineered to amplify XBP1s expression (HEK293/tTA_XBP1s), inhibit CHOP expression (HEK293/tTA_shCHOP), and combine the two responses (HEK293/tTA_XBP1s_shCHOP), with viral particles encoding hPLAT linked to GFP by an IRES (Fig. 7a) using a high MOI (Fig. 2). To assess the extent of activation of the response cassettes upon transgene expression, we measured the iRFP fluorescence of transduced (GFP-positive) cells using flow cytometry. Notably, the iRFP signal of cells selected based on high feedback response amplification (HEK293/tTA_XBP1s #2 and HEK293/tTA_XBP1s_shCHOP #2.1) was higher than that of cells selected based on low feedback response amplification (HEK293/tTA_XBP1s #1 and HEK293/tTA_XBP1s_shCHOP #1.1) (Supplementary Fig. 9a), indicating that the relative extents of feedback response amplification resulting from UPR induction via chemical treatment correlate with those observed upon UPR induction via secretory protein overexpression. To evaluate the correlation between transgene expression and the activation of the feedback-controlled genetic circuits, we quantified the iRFP signal (indicative of the extent of activation of the response cassettes) in subpopulations of transduced cells with high and low GFP signal (indicative of transgene expression). High and low GFP signals were defined by gating subpopulations of transduced (GFP-positive) cells corresponding to the lower and upper 15% of the GFP-positive distribution. The higher iRFP mean fluorescence in the high GFP subpopulations compared to the low GFP subpopulations confirms the correlation between transgene expression and UPR response (Supplementary Fig. 9b-f). We then evaluated the production of secreted functional tPA in transduced (GFP-positive) cells by measuring tPA activity of cell supernatants. Nine days after cell transduction, we observed a 57% increase in tPA production in cells with high expression of the feedback response combining XBP1s amplification and CHOP silencing (HEK293/tTA_XBP1s_shCHOP clone #2.1) compared to parental HEK293 cells (Fig. 7b). Measurements conducted 23 days post-transduction revealed sustained tPA production in cells engineered with high feedback response; specifically, we observed a 23% increase in cells with XBP1s amplification compared to HEK293 cells and a 46% increase in cells combining XBP1s amplification and CHOP silencing (Fig. 7c). These results suggest that amplification of the IRE1 pathway increases the secretion of functional tPA and that combining amplification of IRE1 signaling with delay of CHOP-mediated apoptosis further enhances tPA production.

Fig. 7. Feedback-regulated modulation of XBP1s and CHOP expression enhances tPA production.

a Schematic representation of the protein expression system for tPA production consisting of the three feedback-regulated engineered cell lines, HEK293/tTA_XBP1s, HEK293/tTA_shCHOP, and HEK293/tTA_XBP1s_shCHOP, transduced with virus expressing tPA and GFP at an MOI of 35 (high). b tPA activity of HEK293, HEK293/tTA_XBP1s, HEK293/tTA_shCHOP, and HEK293/tTA_XBP1s_shCHOP cells transduced with virus expressing tPA at an MOI of 35 (high) measured nine days post-transduction. c tPA activity of HEK293, HEK293/tTA_XBP1s #2, and HEK293/tTA_XBP1s_shCHOP #2.1 cells transduced with virus expressing tPA at an MOI of 35 (high) measured 23 days post-transduction. d tPA activity of HEK293 and HEK293/tTA_XBP1s_shCHOP #2.1 cells transduced with virus expressing tPA at an MOI of 35 (high) measured 48 days post-transduction. e Caspase activity of cells as in d measured 48 days post-transduction. The relative caspase activity was calculated by normalizing the caspase activity of cells transduced to overexpress tPA by that of cells transduced with a control vector. f tPA activity of HEK293 and HEK293/tTA_XBP1s_shCHOP #2.1 cells transduced with virus expressing tPA at an MOI of 35 (high) and treated with tunicamycin 45 days post-transduction for 72 hours. g The ratio of the tPA activity of HEK293/tTA_XBP1s_shCHOP #2.1 to HEK293 cells transduced and treated with tunicamycin as in f. Data are reported as mean ± s.e.m. (n = 3 biological replicates). P values were calculated using an unpaired Student’s t-test with a two-tailed analysis. Source data are provided as a Source Data file.

Given the high protein productivity sustained by the cells combining XBP1s amplification and CHOP silencing with high expression of the feedback response, we evaluated the effect of long-term tPA overexpression (48 days) in parental HEK293 and HEK293/tTA_XBP1s_shCHOP #2.1 cells. We observed a general decrease in the levels of secreted active tPA in both parental and engineered cells compared to what was observed at earlier time points, as expected, due to the sustained proteotoxic stress induced by prolonged protein overexpression. However, tPA productivity in HEK293/tTA_XBP1s_shCHOP #2.1 was 26% higher than in parental HEK293 (Fig. 7d), suggesting that combining amplification of XBP1s signaling with CHOP silencing supports tPA production. Measurements of apoptosis induction quantified by monitoring caspases 3 and 7 activity revealed a 40% increase in apoptotic activity in parental HEK293 cells expressing tPA compared to control cells lacking the hPLAT gene (Fig. 7e), indicating that prolonged hPLAT overexpression induces the activation of pro-apoptotic caspases. HEK293/tTA_XBP1s_shCHOP #2.1 cells expressing tPA displayed caspase activity levels comparable to control cells lacking the hPLAT gene, suggesting that combining the amplification of XBP1s signaling with CHOP silencing reduces the pro-apoptotic effect induced by prolonged hPLAT overexpression.

To further evaluate the feedback control system mediating modulation of XBP1s and CHOP signaling, we assessed the production of functional tPA in cells overexpressing hPLAT and treated with tunicamycin to aggravate ER stress. HEK293 and HEK293/tTA_XBP1s_shCHOP #2.1 cells were treated with tunicamycin (0-1 μg/ml) 45 days post-transduction. The addition of tunicamycin caused a decrease in tPA production in a dose-dependent fashion, as expected (Fig. 7f). However, cells engineered with the feedback circuit mediating XBP1s overexpression and CHOP silencing sustained higher productivity than parental cells. The productivity of HEK293/tTA_XBP1s_shCHOP #2.1 cells increased with increasing stress conditions (i.e., tunicamycin dosage). Notably, the increase in tPA activity in HEK293/tTA_XBP1s_shCHOP #2.1 cells compared to parental HEK293 cells upon treatment with high doses of tunicamycin was 67% higher than in untreated cells (Fig. 7g), suggesting that our sense-and-respond system mediating amplification of XBP1s signaling and CHOP silencing supports higher levels of tPA production upon induction of additional ER stress.

Feedback-regulated control of XBP1s expression and CHOP silencing provides a platform for enhanced therapeutic protein production

To assess the potential use of the sense-and-respond system mediating amplification of XBP1s signaling and inhibition of CHOP-mediated apoptosis as a universal platform to support the production of recombinant proteins, we evaluated the production of the monoclonal antibody blinatumomab. Monoclonal antibodies represent the largest class of biological therapeutics in clinical use. Blinatumomab is a bispecific T-cell engager consisting of two variable antibody fragments that target the CD19 transmembrane protein on B-cells and the CD3 protein complex on T-cells, enabling the immune system to attack targeted leukemic cells⁴⁷. Blinatumomab (Blincyto) is a novel immunotherapy treatment option for relapsed or refractory B-cell precursor acute lymphoblastic leukemia (ALL), historically associated with low patient survival outcomes⁴⁸.

To evaluate the effect of XBP1s amplification and CHOP silencing on blinatumomab production, we transiently transfected cells with high feedback-regulated amplification of XBP1s (HEK293/tTA_XBP1s #2) and combining XBP1s amplification and CHOP silencing (HEK293/tTA_XBP1s_shCHOP #2.1) with a plasmid expressing the gene encoding blinatumomab linked to GFP by an IRES. To examine the extent of activation of the response cassettes upon transgene expression, we measured the iRFP fluorescence of transduced (GFP-positive) cells using flow cytometry. Interestingly, HEK293/tTA_XBP1s #2 and HEK293/tTA_XBP1s_shCHOP #2.1 cells expressing blinatumomab exhibited iRFP fluorescence levels comparable to those of cells producing tPA (Supplementary Figs. 9a, 10). We then evaluated the production of blinatumomab in transduced (GFP-positive) cells by measuring the concentration of blinatumomab in cell supernatants. We observed a 34% increase in blinatumomab production in cells with XBP1s amplification (HEK293/tTA_XBP1s #2) compared to HEK293 cells and a 52% increase in cells combining XBP1s amplification and CHOP silencing (HEK293/tTA_XBP1s_shCHOP clone #2.1) (Fig. 8). These results suggest that our sense-and-respond system mediating amplification of XBP1s signaling and CHOP silencing improves the production of different classes of recombinant proteins, potentially providing a universal platform for cell factories with improved and sustained production of therapeutic proteins.

Fig. 8. Feedback-regulated modulation of XBP1s and CHOP expression enhances blinatumomab production.

Blinatumomab production of HEK293, HEK293/tTA_XBP1s #2, and HEK293/tTA_XBP1s_shCHOP #2.1 cells transfected with a plasmid expressing blinatumomab and GFP for 72 hours and measured by ELISA. Blinatumomab levels are normalized to the blinatumomab levels of HEK293 cells. Data are reported as mean ± s.e.m. (n = 3 biological replicates). P values were calculated using an unpaired Student’s t-test with a two-tailed analysis. Source data are provided as a Source Data file.

Discussion

The protein therapeutic market is expected to see a compound annual growth rate of 7.9% through 2029 (Mordor Intelligence, 2024)⁴⁹. Addressing such high demand for therapeutic proteins that are typically large and “difficult to express” requires better protein production systems that can sustain the manufacturing of ample, economical supplies of proteins. Optimization of medium composition and bioreactor design has led to tremendous improvements in protein production methods. Still, it has inevitably reached a productivity plateau due to cell viability and protein expression limitations. As a result, cell line engineering has been recently at the forefront of technological innovation for improved protein biomanufacturing, reflecting the pressing need to evolve cellular systems that support higher protein production yields. Current strategies are based on deregulated modulation of critical components of pathways mediating the production of secreted proteins and cellular viability. This approach has resulted in limited improvements in protein productivity as it does not account for population heterogeneity, which is characteristic of high protein expression systems, and typically leads to metabolic burden, disruption of homeostatic systems, and cell adaptation^15,22–25. To address these challenges, we explored cell engineering strategies based on feedback-regulated modulation of stress response mechanisms (i.e., the UPR). Our strategy leverages the use of genetic circuits that interface with innate UPR mechanisms to sense proteotoxic stress and, in response, modulate the UPR to enhance stress adaptation and delay cell death, thereby adapting the UPR to fluctuations in proteotoxic stress. Because it relies on detecting UPR marker genes, this strategy provides control at the single-cell level, resolving the natural heterogeneity of cell populations subjected to proteotoxic stress. Specifically, we developed a two-module system comprising a UPR sensor and an actuator component. The UPR sensor was developed by genetically engineering cells to link the activation of an early marker of UPR stress to the expression of a transcription factor (tTA), which translates the detection of UPR induction into activation of an orthogonal genetic circuit mediating user-defined modulation of the UPR. Such strategy allowed us to generate three sense-and-respond systems engineered to (1) enhance protein folding and secretion by amplifying the stress attenuation pathway mediated by XBP1s, (2) delay the PERK-mediated pro-apoptotic response by silencing CHOP, and (3) combine amplification of stress attenuation and delay of apoptosis. We demonstrate that combining stress attenuation via XBP1s amplification and apoptosis inhibition via CHOP silencing results in enhanced therapeutic protein productivity, as determined by monitoring the production of the therapeutic enzyme tPA and the bispecific antibody blinatumomab, and improved cell viability. Our results demonstrate this strategy supports a 57% increase in tPA production and a 52% increase in blinatumomab production compared to parental cells under the conditions of this study. These results hold significant promise in the context of efforts to support the increasing growth of the protein therapeutic market.

In addition to providing a potentially more effective approach than deregulated UPR gene modulation, the feedback-regulated systems for UPR modulation described in this study are also superior to the alternative method for regulating UPR genes based on small molecule regulators^50,51. Such strategies rely on continuously monitoring ER stress levels and require trial-and-error optimization to attain the desired target gene expression levels. In addition, unlike our feedback-regulated systems that enable dynamic UPR modulation at the single-cell level, small molecule treatment conditions do not account for population heterogeneity.

Comparisons of the different sense-and-respond systems reported in this study provide important insights. The feedback-regulated modulation of XBP1s and CHOP affects tPA production more dramatically than only XBP1s. Notably, XBP1s has been considered a target for cell factory engineering due to its role in regulating the expression of genes involved in protein folding and secretion^8,10,22,52. Our results indicate that feedback-regulated amplification of XBP1s does not significantly affect tPA production at earlier times (9 days) but has a dramatic impact upon prolonged tPA expression (23 days). These results suggest that feedback-regulated amplification of XBP1s affects recombinant protein productivity upon sustained proteotoxic stress, which is in agreement with previous reports that XBP1s overexpression enhances protein production upon saturation of the ER protein folding capacity⁷.

Feedback-regulated modulation of IRE1 signaling via XBP1s amplification and PERK-mediated apoptosis via CHOP silencing displayed a highly synergistic effect compared to modulation of either XBP1s or CHOP individually. This feedback control system ensures a delay in pro-apoptotic pathway activation with respect to pro-survival pathway activation, a characteristic signature of cells surviving ER stress¹⁶. CHOP signaling downregulation is induced simultaneously to XBP1s signaling amplification, thus enhancing the cellular folding capacity and maximizing the yield of functional product, which was monitored in this study by quantifying to tPA activity and hence the secretion of functional product. These results support the hypothesis that the relative dynamics of the UPR branches and, specifically, early activation of the pro-survival pathway and delayed induction of the pro-apoptotic pathway affect cell fate and protein productivity¹⁶. It remains to be determined if other crucial mediators of stress attenuation, such as BIP and ATF6, and pro-apoptotic genes, such as BAK and BAX, could be leveraged to achieve feedback-controlled UPR modulation^52–54. Emerging synthetic biology tools for multiplex gene regulation will allow interrogating large arrays of genes individually and in combination to identify cell engineering strategies that maximize protein productivity⁵⁵.

The approach described in this study enables the generation of cell factories with different extents of amplification of the user-defined feedback response. Monoclonal populations were screened to isolate clones with different extents of amplification of the stress attenuation response or combined stress attenuation and apoptosis delay and later assessed to quantify protein productivity. However, the relationship between the extent of amplification of the UPR response and protein productivity remains to be determined, and it is likely to be UPR target- and therapeutic product-specific. Combining this strategy for engineering cell factories that dynamically modulate the cellular response to proteotoxic stress with high-throughput methods for circuit engineering and clone selection will allow uncovering the design rules of cell factories that maximize the production of protein therapeutics.

Methods

Reagents

Tunicamycin (catalog no. T7765-5MG; Sigma-Aldrich) and erythromycin (catalog no. E5389-5G, Sigma-Aldrich) were dissolved in DMSO (catalog no. 472301; Sigma-Aldrich) to prepare a 10 mg/ml stock solution. Tetracycline (catalog no. T7660-5G; Sigma-Aldrich) was dissolved in H₂O to prepare a 10 mg/ml stock solution.

Plasmids and cloning

Plasmids were transformed and maintained in Stbl3 E. coli competent cells (catalog no. 11319019, Thermo Fisher Scientific). The primers used in this study are listed in Supplementary Table 1.

The plasmid pLentiCRISPRv2_XBP1 encoding the SpCas9 gene and the guide RNA targeting the 3’ end of the coding sequence of the XBP1 gene was constructed by cloning the guide RNA into the LentiCRISPRv2 plasmid (Addgene plasmid no. 52961) using BsmBI restriction enzyme sites and appropriate oligos (Supplementary Table 1) as previously descried⁵⁶.

The plasmid pXBP1_P2A_tTA used to generate HEK293/tTA cells via homologous direct repair encodes the P2A_tTA cassette, which contains the gene sequences of the porcine teschovirus 2A peptide (P2A), the tetracycline transactivator (tTA), and the Neomycin (NeoR) selection marker, flanked by XBP1 homology sequences. pXBP1_P2A_tTA was built by cloning the P2A_tTA cassette, the ori_AmpR cassette, and the two ~1-kb XBP1 homologous sequences using four BsaI restriction enzyme sites. Specifically, primer extension PCR was used to amplify the P2A_tTA cassette from pBIP_IRES_tTA²⁹ using appropriate oligos (Supplementary Table 1). The ori_AmpR cassette was amplified from pcDNA 3.1 (Cat. No. V79020; Thermo Fisher Scientific) as previously described²⁹. The two ~1-kb XBP1 homologous sequences were amplified from genomic DNA using oligos recognizing the 3’ of the XBP1 gene, designed to add BsaI enzyme recognition sites (Supplementary Table 1). The stop codon immediately downstream of the coding region of the XBP1 gene was mutated to couple the expression of tTA to the splicing of XBP1 mRNA.

The plasmid pXBP1s_IRES_iRFP used to generate HEK293/tTA_XBP1s cells encodes XBP1s and iRFP and was built by cloning the sequences of XBP1s, IRES, and iRFP into the 7TO_GFP plasmid²⁹ containing the TRE operator site and the minimal CMV promoter. The sequence encoding XBP1s was amplified from pCMV5-Flag-XBP1s (Addgene plasmid no. 63680) with primers designed to add AgeI and SpeI enzyme recognition sites 5’ and 3’, respectively (Supplementary Table 1). The sequence encoding IRES_iRFP was amplified from pLenti_CMV_VHH_IRES_iRFP⁵⁷ with primers designed to add SpeI and SalI enzyme recognition sites 5’ and 3’, respectively (Supplementary Table 1). The XBP1s and IRES_iRFP PCR-amplified fragments were ligated into the 7TO_GFP plasmid using AgeI and SalI restriction enzyme sites, thus replacing the GFP sequence with the XBP1s_IRES_iRFP cassette.

The plasmid piRFP-shCHOP used to generate HEK293/tTA_shCHOP cells encodes iRFP and the short hairpin RNA against CHOP (shCHOP) and was built by cloning the iRFP-shCHOP cassette into the 7TO_GFP plasmid using AgeI and SalI restriction enzyme sites, thus replacing the GFP sequence with the iRFP-shCHOP cassette. The sequence encoding the iRFP-shCHOP cassette was generated from piRFP (Addgene plasmid no. 31857) using primer extension PCR with oligos designed to add the miR-30 loop and the CHOP-specific shRNA at the 3’ of iRFP (Supplementary Table 1).

The plasmid pXBP1s_IRES_iRFP-shCHOP plasmid used to generate HEK293/tTA_XBP1s_shCHOP cells encodes XBP1s, iRFP, and shCHOP and was generated from pXBP1s_IRES_iRFP by replacing the iRFP with iRFP-shCHOP from piRFP-shCHOP at NarI and SalI restriction enzyme sites.

The plasmid phPLAT_IRES_iRFP used to produce lentiviral particles expressing tPA and iRFP was built by cloning the sequences of hPLAT, IRES, and iRFP into the pLenti_CMV_GFP_Blast plasmid (Addgene plasmid no. 17445). The hPLAT gene sequence was amplified from pCAG-hPLAT (Vector Builder) with primers designed to add AgeI and MluI restriction sites at the 5’ and 3’, respectively (Supplementary Table 1). The IRES sequence was PCR-amplified from pBIP_IRES_tTA²⁹ with primers designed to add MluI and SpeI restriction sites at the 5’ and 3’, respectively (Supplementary Table 1). The sequence encoding iRFP was amplified from piRFP using oligos designed to add SpeI and SalI enzyme recognition sites 5’ and 3’, respectively (Supplementary Table 1). The resulting hPLAT, IRES, and iRFP sequences were ligated into the pLenti_CMV_GFP_Blast plasmid using AgeI and SalI restriction enzyme sites, thus replacing the GFP sequence with the hPLAT_IRES_iRFP cassette.

The plasmid phPLAT_IRES_GFP used to produce lentiviral particles expressing tPA linked to GFP was built by replacing the IRES_iRFP cassette from phPLAT_IRES_iRFP with IRES_GFP from piRFP_IRES_GFP²⁹ using BamHI and SalI restriction enzyme sites. The IRES_GFP sequence was amplified from piRFP_IRES_GFP with primers designed to add BamHI and SalI restriction sites at the 5’ and 3’, respectively (Supplementary Table 1).

The plasmid pBlinatumomab_IRES_GFP comprises the gene encoding blinatumomab linked to GFP by an IRES under the control of a CMV promoter. pBlinatumomab_IRES_GFP was built by replacing the etanercept coding sequence of pETN_IRES_GFP (Vector Builder) with the blinatumomab coding sequence using the Gibson Assembly Master Mix (catalog no. E2611S, New England Biolabs) and appropriate oligos (Supplementary Table 1). The blinatumomab coding sequence was amplified from a gene fragment synthesized by Twist Bioscience using appropriate oligos (Supplementary Table 1).

Cell culture and transfections

HEK293 cells (catalog no. CRL-1573, ATCC) and HEK293T cells (catalog no. CRL-3216, ATCC) were cultured in DMEM/High glucose (catalog no. SH30243.01, Hyclone), supplemented with 10% fetal bovine serum (FBS, catalog no. 25-514H, GenClone) and 1% penicillin–streptomycin–glutamine (catalog no. SV30082.01, Hyclone), and maintained at 37 °C and 5% CO₂. Cells were passaged using phosphate-buffered saline (catalog no. 17–516 F, Lonza) and trypsin (TrypLE Express, catalog no. 12605–036, GIBCO).

Transient transfections were conducted by seeding cells onto 12-well or 24-well plates. After 24 hours, upon reaching 70-80% confluency, transfections were performed using the JetPrime DNA transfection reagent (catalog no. 101000046; Polyplus) according to the manufacturer’s protocol. The medium was replaced 8 hours post-transfection.

Lentivirus production and transduction

Third-generation lentiviruses were generated by seeding HEK293T cells onto 100 × 20-mm² tissue culture dishes at a 1 × 10⁶ cells/dish density. Cells were transfected with pXBP1s_IRES_iRFP, piRFP-shCHOP, pXBP1s_IRES_iRFP-shCHOP, phPLAT_IRES_iRFP, or phPLAT_IRES_GFP, and the packaging plasmids pMLg/pRRE (Addgene plasmid no. 12251), pRSV-Rev (Addgene plasmid no. 12253) and pMD2.G (Addgene plasmid no. 12259) in a 2:5:2.5:3 ratio, respectively. Transfections were conducted using 5 μg of total DNA per 100 × 20-mm² tissue culture dish, consisting of 0.8 μg of pXBP1s_IRES_iRFP, piRFP-shCHOP, pXBP1s_IRES_iRFP-shCHOP, phPLAT_IRES_iRFP, or phPLAT_IRES_GFP, 2 μg of pMLg/PRRE, 1 μg of pRSV-Rev and 1.2 μg of pMD2.g, respectively. The medium was replaced with fresh medium 8 hours post-transfection, and the virus-containing medium was collected after 48 hours. The virus was concentrated using a Lenti-X concentrator (catalog no. 631232, Clontech) according to the manufacturer’s protocol. Viruses were titrated by quantitative RT-PCR using Lenti-X qRT-PCR Titration Kit (catalog no. 631235, Takara) according to the manufacturer’s protocol.

Cell transductions were conducted by seeding HEK293 cells onto 12-well or 24-well plates at a 1 × 10⁵ cells/well density. After 24 hours, the medium was replaced with medium supplemented with virus particles and 8 μg/ml polybrene (catalog no. TR-1003, Sigma-Aldrich). The virus-containing medium was replaced with fresh medium 24 hours post-transduction.

Generation of stable cell lines

To generate the HEK293/tTA cell line, HEK293 cells were seeded onto 12-well plates and transfected with pLentiCRISPRv2-XBP1 and pXBP1_P2A_tTA in a 1:2 ratio. Cells were transferred to 100 × 20-mm² tissue culture dishes 48 hours post-transfection and selected for 2 weeks in medium containing 1 mg/ml G418 (catalog no. 345812, EMD Millipore). Control samples for assessing potential off-target integration were generated by transfecting HEK293 cells with a LentiCRISPRv2 plasmid encoding a scrambled gRNA sequence and pXBP1_2A_tTA in a 1:2 ratio. Lack of off-target integration was verified by culturing transfected cells in medium supplemented with G418 (1 mg/ml) and monitoring cell death.

To generate the HEK293/tTA_XBP1s, HEK293/tTA_shCHOP, and HEK293/tTA_XBP1s_shCHOP cell lines, HEK293_tTA cells were seeded onto 12-well plates at a 1 × 10⁵ cells/well density and transduced with pXBP1s_IRES_iRFP, piRFP-shCHOP, and pXBP1s_IRES_iRFP-shCHOP, respectively. Cells were transferred into 100 × 20 mm² tissue culture dishes 48 hours post-transduction and selected for 2 weeks in medium supplemented with 5 μg/ml blasticidin (catalog no. ant-bl-1, InvivoGen).

To screen monoclonal cell populations, sorted cells were seeded onto 96-well plates containing DMEM with 20% FBS at a 0.5 cells/well density, expanded, and treated with 1 μg/ml tunicamycin (catalog no. T7765-5MG; Sigma-Aldrich) for 48 hours.

Genomic PCR

Genomic DNA was extracted using the E.Z.N.A Tissue DNA Kit (catalog no. D3396-02; Omega Bio-tek) according to the manufacturer’s protocol. PCR-mediated amplification of genomic DNA was performed using the Q5 Hot Start High-Fidelity 2X Master Mix (catalog no. M0494S, New England Biolabs) and appropriate primers (Supplementary Table 1). The PCR products were resolved by electrophoresis on a 1% Tris-acetate-EDTA agarose gel, stained with SYBR Safe DNA Gel Stain (catalog no. S33102, Thermo Fisher Scientific) and visualized using a blue light transilluminator.

Flow cytometry analyses

Cells were analyzed with a FACSCanto II flow cytometer (BD Biosciences), using the BD FACSDiva Software (v8.0.2). GFP fluorescence intensity was detected using a 488 nm laser and 530/30 nm emission filter. iRFP fluorescence intensity was detected using a 635 nm laser and 780/60 nm emission filter. At least 10,000 cells were recorded in each sample for analysis. Gating strategy is provided in Supplementary Fig. 11. FlowJo (v10.10.0) and Microsoft Excel (v2501) were used to analyze the data.

Quantitative RT–PCR

RNA was extracted using the RNeasy Plus Mini Kit (catalog no. 74134, Qiagen), and cDNA was synthesized using the qScript cDNA SuperMix (catalog no. 95048-100, Quanta Biosciences) following the manufacturer’s protocol. Quantitative RT-PCR reactions were performed using the PerfeCTa SYBR Green FastMix (catalog no. 95072-012, Quanta Biosciences) in a CFX96 Real-Time PCR Detection System (Bio-Rad) using appropriate primers (Supplemental Table 1). The housekeeping genes RNA18SN1, GAPDH, and ACTB were used as reference points to normalize the data.

tPA activity assay

The culturing media of cells transduced with phPLAT_IRES_iRFP or phPLAT_IRES_GFP was collected 72 hours after cell passage and centrifuged (3000 g for 15 min at 4 °C) to eliminate cell debris. The supernatant was collected, and tPA activity was measured using the SensoLyte AMC tPA Activity Assay Kit (catalog no. AS-72160, AnaSpec) according to the manufacturer’s protocol. Briefly, 50 μl of a synthetic fluorogenic substrate were incubated with 50 μl of diluted cell culture supernatant (1:12) at 37 °C for 60 minutes. Measurements were conducted using a fluorescent plate reader at Ex/Em = 354/442 nm. The tPA activity of the cell supernatant was normalized to the cell number.

Enzyme-linked immunosorbent assay (ELISA)

The culturing media of cells transfected with pBlinatumomab_IRES_GFP was collected 72 hours after transfection and centrifuged (100 g for 8 min at 4 °C) to eliminate cell debris. The supernatant was collected and analyzed by ELISA. Specifically, Nunc-Immuno 96-well flat-bottom plates (catalog no. 442404, Thermo Scientific) were coated with 100 μl of 0.5 µg/ml G4S Linker rabbit antibody (catalog no. 71645, Cell Signaling) for 1 hour at room temperature. Plates were washed with 300 μl of PBS containing Tween-20 (PBST, catalog no. 28352, Thermo Scientific) and blocked with 300 μl of SuperBlock buffer (catalog no. 37515, Thermo Scientific) for 1 hour at room temperature. The buffer was removed, and 160 µl of samples were added to each well, followed by overnight incubation at 4 °C. The following day, the wells were emptied and washed five times with 300 µl of PBST. Then, 100 μl of 0.5 µg/ml 6x-His Tag Monoclonal Antibody, HRP-conjugated (catalog no. MA1-21315-HRP, Invitrogen) were added and incubated for 1 hour at room temperature with shaking. The plates were then washed five times with 300 μl of PBST. For colorimetric detection, 100 µl of TMB substrate (catalog no. 34028, Thermo Scientific) were added to each well, and the reaction was stopped by adding 100 µl of 0.16 M sulfuric acid (catalog no. 339741, Sigma-Aldrich) per well. Measurements were conducted using a colorimetric plate reader at 450 nm with a reference wavelength correction of 630 nm.

MTS assay

Cells were seeded onto 96-well plates at a 1 × 10⁵ cells/ml density and incubated at 37 °C for 24 hours. Cells were treated with 10 μg/ml tunicamycin, and cell viability was determined 48 hours post-treatment using the MTS assay (CellTiter 96 AQueous One Solution cell proliferation assay, catalog no. G3582, Promega) according to the manufacturer’s protocol. Briefly, 20 μl of MTS reagent were added to each well and incubated at 37 °C for 1 hour. Measurements were conducted in triplicate at 490 nm using a colorimetric plate reader.

Annexin-V staining

Cells were seeded onto 12-well plates at a 1 × 10⁵ cells/ml density and incubated at 37 °C for 24 hours. Cells were treated with 10 μg/ml tunicamycin for 48 hours. Cells were collected and stained with Annexin V-FITC (catalog no. A13199, Thermo Fisher Scientific) according to the manufacturer’s instructions and analyzed by flow cytometry.

Caspase-3/7 activity assay

Cells were seeded onto 96-well plates at a 1 × 10⁵ cells/ml density and incubated at 37 °C for 72 hours. Caspase activity was determined using the SensoLyte Homogeneous AMC Caspase-3/7 Assay Kit (catalog no. AS-71118, AnaSpec) according to the manufacturer’s protocol. Briefly, 50 μl of caspase substrate were added to each well and incubated for 12 hours. Measurements were conducted using a fluorescent plate reader at Ex/Em = 354/442 nm.

Statistics and reproducibility

All data are reported as mean, and error bars represent the standard error of at least three independent experiments. The statistical significance of experiments was calculated using a two-tailed Student’s t-test. No statistical method was used to predetermine the sample size. No data were excluded from the analyses, the experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

D.B., B.B., and L.S. conceived the project. D.B., B.B., W.C.A., C.D.L., and W.Z. performed the experiments. D.B. and W.C.A. analyzed the data. D.B. and L.S. wrote the manuscript. All authors reviewed the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data supporting the findings of this study are available within the article, its Supplementary Information, or Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-58994-x.