Optogenetic protein production

Abstract

The present disclosure provides a host cell for optogenetic protein production, comprising: a first sequence comprising a first promoter operably linked to a gene of interest; a second sequence comprising a second promoter operably linked to a gene encoding a light-sensitive transcription factor, wherein the light-sensitive transcription factor is configured to induce expression of the gene of interest upon activation by light; and a first selection marker. The gene of interest, first promoter, second gene, second promoter, and first selection marker are either integrated at a first locus, or the gene of interest, first promoter, and first selection marker are integrated at the first locus while the second gene, second promoter, and a second selection marker are integrated at a second locus. The disclosure also provides systems and methods for optogenetic protein production using the host cell.

Description

Princeton- 101576 OPTOGENETIC PROTEIN PRODUCTION CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Application No. 63 / 641,271, filed May 1, 2024, which is hereby incorporated by reference in its entirety. STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] This invention was made with government support under Grant No. MCB-2300239 awarded by the National Science Foundation and Grant No. DE-SC0022155 from the Department of Energy. The government has certain rights in the invention. FIELD OF INVENTION

[0003] The present disclosure relates to optogenetic systems for protein production. BACKGROUND

[0004] Recombinant protein production is a cornerstone of modern biotechnology, enabling the synthesis of valuable proteins for diverse applications in medicine, industry, and research. Yeast cells, such as Pichia pastoris, have emerged as powerful host organisms for recombinant protein expression due to their ability to grow to high cell densities, perform eukaryotic post-translational modifications, and efficiently secrete proteins.

[0005] A commonly used approach for controlling recombinant protein production in P. pastoris involves the methanol-inducible AOX1 promoter system. This system allows for tight regulation of gene expression, with protein production initiated by the addition of methanol to the growth medium. While effective, the use of methanol as an inducer presents several challenges. Methanol is derived from non-renewable resources, poses safety hazards due to its flammability and toxicity, and its metabolism generates substantial heat and requires high oxygen demand, which can limit scalability of production processes.

[0006] These limitations have motivated research into alternative induction systems for P. pastoris that avoid the use of methanol. Some approaches have focused on modifying the regulation of the AOX1 promoter to respond to other carbon sources. Other strategies have explored different native P. pastoris promoters or heterologous promoter systems. However, many of these alternatives have not matched the tight regulation and strong induction capabilities of the methanol-inducible AOX1 system.Princeton- 101576

[0007] Optogenetic tools, which use light to control biological processes, have shown promise for regulating gene expression in various organisms. These systems typically employ photosensitive proteins that change conformation in response to specific wavelengths of light, allowing for precise temporal and spatial control of cellular functions. The application of optogenetics to control recombinant protein production in industrial microorganisms is an emerging area of research that may offer new capabilities for bioprocess control.

[0008] Developing improved methods for controlling recombinant protein production in P. pastoris remains an active area of investigation. Strategies that provide tight regulation, strong induction, and avoid the use of methanol could potentially address current limitations and expand the utility of this expression system for biotechnology applications. SUMMARY

[0009] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0010] According to an aspect of the present disclosure, a host cell for optogenetic protein production is provided. The host cell includes a first sequence comprising a first promoter operably linked to a gene of interest, a second sequence comprising a second promoter operably linked to a gene encoding a light-sensitive transcription factor, wherein the light-sensitive transcription factor is configured to induce expression of the gene of interest upon activation by light, and a first selection marker. Either the gene of interest, the first promoter, the second gene, the second promoter, and the first selection marker are integrated at a first locus, or the gene of interest, the first promoter, and the first selection marker are integrated at the first locus, and the second gene, the second promoter, and a second selection marker are integrated at a second locus.

[0011] According to other aspects of the present disclosure, the host cell may include one or more of the following features. The light-sensitive transcription factor may comprise a LOV domain. The light-sensitive transcription factor may be EL222. The first promoter may be PC120. The host cell may be a yeast cell. The yeast cell may be a methylotrophic yeast cell. The methylotrophic yeast cell may be Komagataella phaffii. The gene of interest may encode a secreted protein. The secreted protein may be selected from the group consisting of yEGFP, β- lactoglobulin, and SR18 nanobody. The first selection marker may be zeocin. The secondPrinceton- 101576 selection marker may be G418. The gene encoding a light-sensitive transcription factor may have a predetermined copy number at the first locus or the second locus.

[0012] According to another aspect of the present disclosure, a system for optogenetic protein production is provided. The system includes a plurality of host cells, each host cell being a host cell as described above, and a light source configured to activate the light-sensitive transcription factor, wherein activation of the light-sensitive transcription factor induces expression of the gene of interest.

[0013] According to other aspects of the present disclosure, the system may include one or more of the following features. The plurality of host cells may include a first host cell and a second host cell, wherein the first host cell and the second host cell have different copy numbers of the light-sensitive transcription factor. Each host cell may have a copy number of the light- sensitive transcription factor that is 1-10. Each host cell may have a copy number of the light- sensitive transcription factor that is 2-10.

[0014] According to another aspect of the present disclosure, a method for optogenetic protein production is provided. The method includes providing a host cell as described above, culturing the host cell under conditions suitable for growth, and exposing the host cell to light to activate the light-sensitive transcription factor, wherein activation of the light-sensitive transcription factor induces expression of the gene of interest.

[0015] According to other aspects of the present disclosure, the method may include one or more of the following features. The method may further include selecting a light-sensitive transcription factor-containing parent strain based on an illuminating capability of a bioreactor system and protein production needs, cloning a plasmid and placing a gene of interest under control of a PC120promoter with a zeocin marker, and forming the optogenetically-controlled strain by transforming the light-sensitive transcription factor-containing parent strain with the plasmid. The method may further include determining illuminating capability of a bioreactor system.

[0016] According to another aspect of the present disclosure, a hybrid machine-learning- supported method for light-induced recombinant protein production is provided. The method includes formulating differential equations to represent relevant system states, such as biomass, a protein of interest, glucose, and copy number, where the copy number is treated as a dynamic state and remains constant with a right-hand side of its differential equation set to zero, determining a target initial condition of the copy number, performing parameter estimation based on dynamic data from specific experiments conducted under predetermined light intensity regimes and light-sensitive transcription factor copy numbers, where performingPrinceton- 101576 parameter estimation includes using a trained multi-input, single-output Gaussian process to predict each parameter as a function of a light intensity regime and light-sensitive transcription factor copy number, and incorporating predicted means of the trained multi-input, single-output Gaussian process as model parameters into mechanistic kinetic functions.

[0017] The foregoing general description of the illustrative embodiments and the following detailed description thereof are merely exemplary aspects of the teachings of this disclosure and are not restrictive. BRIEF DESCRIPTION OF FIGURES

[0018] Non-limiting and non-exhaustive examples are described with reference to the following figures.

[0019] FIGS. 1A-1B illustrate block diagrams of optogenetic systems for protein production, according to aspects of the present disclosure.

[0020] FIG. 2A illustrates a system in a dark state with no activation, according to an embodiment.

[0021] FIG. 2B illustrates the system of FIG. 2A in an activated state, according to an embodiment.

[0022] FIG. 3 illustrates a block diagram of a bioreactor system for optogenetic protein production, according to aspects of the present disclosure.

[0023] FIGS.4A-4C are graphs showing intracellular yEGFP production with the coupled optogenetic system. Light-induced production of yEGFP in 1% glucose (BMD1) and 1% glycerol (BMG1) is compared to methanol induction of PAOX1 at a single copy and ~ 8 copies in 70 ^^^^^^^^ / ^^2^^ (4A), 5 ^^^^^^^^ / ^^2^^ (4B), and 70 ^^^^^^^^ / ^^2^^ (4C) light intensities. Single copy strains are ySMH3-100-3 and ySMH5-100-5. Multicopy strains are ySMH3-1000-3 and ySMH5-1000-3.

[0024] FIGS.5A-5F are graphs characterizing the effect of EL222 copy number and light intensity on growth and production. yEGFP-producing strains with different levels of EL222 (ySMH193-11, 6, and 9) were tested at several light intensities to identify a suitable condition for production (5A, 5C, 5E) and growth (5B, 5D, 5F) in BMD1 medium. Strains containing one (5A, 5B), three (5C, 5D), or eight copies (5E, 5F) of EL222 were examined, with the higher copy strains exhibiting greater sensitivity to light. Mean values are depicted as data points. Lines illustrate the predicted dynamics generated by the hybrid model with Gaussian-process- predicted parameters. The predicted glucose concentrations are shown in Supplementary Figure S21.Princeton- 101576

[0025] FIG. 5G shows a graph relating to optimizing light dosage based on EL222 copy number. The total yEGFP (not normalized by cell density) after 25 hours is shown for strains containing 1, 3, and 8 copies of EL222 to depict the optimal light dosage for each strain. Data is derived from the same experiment as FIGS.5A-5F.

[0026] Figure 6 relates to intracellular production of M. californianus Mfp5 with optogenetics and PAOX1. Production of Mfp5 is compared between PAOX1and the decoupled optogenetic system using 1% glucose or glycerol for the optogenetic strain (ySMH238-9) and 1% methanol for the PAOX1strain (ySMH186-9). Cell cultures were equalized to initial OD600= 10 from overnight cultures prior to 48h-induction with methanol or light. Optogenetic cultures were exposed to light at 50 ^^^^^^^^ / ^^2^^ intensity and fermentations were carried out for 48 h. Gel was loaded with equal amounts of final biomass as prescribed by the TCA protein extraction method. Pixel intensity in the western blot bands was analyzed using ImageJ to generate quantitative values for Mfp5 production. Mean values after 48 h are shown. Coomassie staining of the blotted membrane confirmed that comparable amounts of samples were loaded on the gel

[0027] Figures 7A and 7B relate to phototoxic stress of EL222 inhibiting secretion even at low light levels. The phototoxic effect on secretion was verified by co-expressing EL222 in a strain secreting yEGFP from PAOX1. In 7A, secretion by a non-optogenetic PAOX1-driven strain without EL222 (ySMH154-1) and with EL222 (ySMH185-1) at a range of light intensities is shown by dot blot, in which each dot represents a biological replicate induced for 48 h in BMM1 medium. In 7B, this effect was quantified using ImageJ analysis of the dot blot, in which mean values are shown.

[0028] Figures 8A-8F are graphs relating to optogenetic secretion of diverse proteins compared to PAOX1. Secretion of yEGFP (8A, 8B), β-lactoglobulin dairy protein (8C, 8D), and the SR18 nanobody (8E, 8F) was tested in triplicate with the best optogenetic and methanol- induced strains. For the optogenetic strains, the best producers were tested from a low copy EL222 background at 50 ^^^^^^^^ / ^^2^^ light intensity (8A, 8C, and 8E) and a high copy EL222 background at 5 ^^^^^^^^ / ^^2^^ intensity (8B, 8D, 8F) in both BMD1 and BMG1 media. PAOX1- driven strains were induced in BMM1 medium. Western blots were quantified using ImageJ, and bar graphs depict mean values.

[0029] Figures 9A and 9B are graphs relating to optogenetic secretion (9A) and specific productivity (9B) of the SR18 nanobody in a bioreactor, to test optogenetic induction in a larger volume and high cell density. The optogenetic system is compared to a methanol-inducedPrinceton- 101576 bioprocess in which equal amounts of the carbon sources were fed on a mass basis. Samples taken at the indicated time points were analyzed by Bradford analysis to measure the secreted protein titers. DETAILED DESCRIPTION

[0030] The following description sets forth exemplary aspects of the present disclosure. It should be recognized, however, that such description is not intended as a limitation on the scope of the present disclosure. Rather, the description also encompasses combinations and modifications to those exemplary aspects described herein.

[0031] The present disclosure relates to systems and methods for optogenetic protein production in host cells. In some cases, the systems and methods utilize light-sensitive transcription factors to control protein expression. The light-sensitive transcription factors may be activated by exposure to light, which induces expression of a gene of interest.

[0032] In some cases, the host cell may be a yeast cell. The yeast cell may be a methylotrophic yeast cell. In some implementations, the methylotrophic yeast cell may be Komagataella phaffii, also known as Pichia pastoris. In some implementations, the methylotrophic yeast cell may be Ogataea polymorpha, Candida boidinii, and / or Pichia methanolica. In some implementations, the yeast cell may be Saccharomyces cerevisiae, which has been widely used for recombinant protein production and genetic engineering due to its well-characterized genetics and metabolism.

[0033] The use of light-sensitive transcription factors for controlling protein expression provides several advantages over traditional induction methods. Light activation allows for precise temporal and spatial control of gene expression. Additionally, light induction avoids the need for chemical inducers, which can be costly or potentially toxic to cells. The optogenetic approach also enables tunable expression levels by modulating light intensity or duration.

[0034] In methylotrophic yeasts like Komagataella phaffii, the optogenetic system may provide an alternative to methanol-induced expression systems commonly used for recombinant protein production. This light-inducible system may allow for protein production without the safety and scaling challenges associated with methanol induction.

[0035] The optogenetic protein production system described herein may be applied to produce a wide variety of recombinant proteins for research, industrial, or therapeutic applications. The system provides a flexible platform that can be adapted to different host organisms and protein targets.Princeton- 101576

[0036] The present disclosure relates to a host cell for optogenetic protein production. FIG. 1A and FIG.1B illustrate block diagrams of optogenetic systems for protein production.

[0037] In some cases, a host cell may comprise a coupled system 100. The coupled system 100 may include a first sequence 120 and a second sequence 110. The first sequence 120 may comprise a first promoter 122 operably linked to a first gene of interest 124. The second sequence 110 may comprise a second promoter 112 operably linked to a gene for light-sensitive protein with a binding domain 114. The coupled system 100 may also include a first selection marker 130.

[0038] In coupled implementations, the gene of interest 124, first promoter 122, gene for light-sensitive protein with a binding domain 114, second promoter 112, and first selection marker 130 may be integrated at a first locus in the host cell genome.

[0039] The first promoter 122 may be any appropriate promoter capable of being activated by the light-sensitive protein with a binding domain. Preferably, the promoter is engineered specifically to be bound by the light-sensitive protein with a binding domain. For example, in some embodiments, the first promoter 122 may be, e.g., PC120, known to be engineered to be bound by the EL222 transcription factor under blue light. In particular, EL222 has a helix- turn-helix (HTH) DNA-binding domain that recognizes a specific palindromic motif in the C120 site. Other non-limiting example promoter / transcription factor pairs that may be utilized include, e.g., UAS / GAL4 bound by CRY2 / CIB1 (blue light), VVD (blue light), or PhyB / PIF (red light), PcpcG2 bound by CcaR (green light), OmpC bound by OmpR (red light), or a T7 promoter bound by an engineered Opto-T7 RNA polymerase (various light conditions specific to engineered Opto-T7 RNA polymerase).

[0040] The second promoter 112 may be a constitutive promoter. Non-limiting examples of constitutive promoters in yeast cells include, e.g., TEF1, PGK1, ADH1, and GPD1 promoters.

[0041] The light-sensitive protein with a binding domain may be any appropriate light- sensitive protein that can activate the first promoter. Non-limiting examples includes:

[0042] - Blue light (~450 nm) activated systems: EL222 (from Erythrobacter litoralis) that has a LOV domain + HTH DNA-binding domain and activates transcription when exposed to blue light; CRY2 / CIB1 (from Arabidopsis thaliana) where CRY2 dimerizes with CIB1 under blue light and fuses to transcription activators or DNA-binding domains; VVD-based systems (from Neurospora crassa), where a VVD LOV domain fused to GAL4 forms dimers under blue light to drive transcription; or AsLOV2-based switches (from Avena sativa) where the LOV2 domain undergoes conformational change under blue lightPrinceton- 101576

[0043] - Red / Far-Red Light (~650-750 nm) activated systems: PhyB / PIF (from Arabidopsis thaliana) where Phytochrome B (PhyB) binds to PIF in red light and dissociates in far-red light; used for reversible control of transcription; or Cph1 / EnvZ / OmpR(Cph8) (engineered from Synechocystis) which is a red-light-sensitive two-component system typically engineered into E. coli to regulate transcription.

[0044] - Green light (~520 nm) activated systems: CarH-based systems (from Thermus thermophilus) that may bind DNA in the dark using AdoB12 cofactor and dissociates in light; or CcaS / CcaR (from Synechocystis) which is a two-component system that responds to green / red light.

[0045] The gene for light-sensitive protein with a binding domain 114 may encode a light- sensitive transcription factor. In some cases, the light-sensitive transcription factor may comprise a LOV domain. The light-sensitive transcription factor may be configured to induce expression of the gene of interest 124 upon activation by light (e.g., by interacting with the first promoter). In some implementations, the light-sensitive transcription factor may be an EL222 protein.

[0046] The gene of interest 124 may encode any desired protein. In some implementations, the protein may comprise a tag domain. The tag domain may include, e.g., a fluorescent tag, a purification tag, an epitope tag, etc. Non-limiting examples of fluorescent tags include green fluorescent proteins (e.g., GFP, sfGFP, EGFP, yEGFP), yellow fluorescent proteins (e.g., YFP, EYFP, Citrine), blue fluorescent proteins (e.g., EBFP, EBFP2, Azurite), cyan fluorescent proteins (e.g., ECFP, Cerulean), red fluorescent proteins (e.g., mKate, mKate2, mPlum, DsRed monomer, mCherry), and orange fluorescent proteins (mOrange, mKO, mTangerine) or any other suitable fluorescent protein. Non-limiting examples of purification and / or epitope tags includes, e.g., His-tag, FLAG-tag, Strep-tag / Strep-tag II, HA-tag, GST-tag, Myc-tag, VG- tag, T7-tag.

[0047] The disclosed systems preferably utilize at least one selection marker. As used herein, the term "selection marker" may refer to a gene or genetic element that confers a selectable trait to cells containing it, allowing for the identification and isolation of cells that have successfully incorporated a desired genetic construct. Selection markers typically encode proteins that provide resistance to specific antibiotics or enable growth under particular nutrient conditions. Examples of selection markers may include, but are not limited to:

[0048] 1. Antibiotic resistance genes: Zeocin resistance gene, G418 (geneticin) resistance gene, hygromycin resistance gene, ampicillin resistance gene, kanamycin resistance genePrinceton- 101576

[0049] 2. Auxotrophic markers: HIS3 (histidine biosynthesis), LEU2 (leucine biosynthesis), URA3 (uracil biosynthesis), TRP1 (tryptophan biosynthesis)

[0050] 3. Metabolic markers: ADE2 (adenine biosynthesis), LYS2 (lysine biosynthesis)

[0051] 4. Fluorescent proteins: Green Fluorescent Protein (GFP), Yellow Fluorescent Protein (YFP), Red Fluorescent Protein (RFP)

[0052] 5. Chromogenic markers: LacZ (β-galactosidase)

[0053] 6. Toxic gene resistance: Blasticidin S resistance gene, puromycin resistance gene

[0054] These selection markers may be used individually or in combination to facilitate the selection and maintenance of cells carrying specific genetic constructs in various experimental and industrial applications.

[0055] The first selection marker 130 may be a zeocin resistance marker in some cases. In implementations using the decoupled system 105, the second selection marker 140 should be different from the first selection maker; in some embodiments, the second selection marker may be a G418 resistance marker.

[0056] The gene for light-sensitive protein with a binding domain 114 may have a predetermined copy number at the first locus or the second locus. In some cases, copy numbers of 1, 3, 5, and 8 may be specifically tested. The decoupled system 105 may allow for different copy numbers of the gene for light-sensitive protein with a binding domain 114 and the gene of interest 124.

[0057] In other cases, a host cell may comprise a decoupled system 105. The decoupled system 105 may separate the components into two distinct sequences. The first sequence 120 may comprise the first promoter 122 operably linked to the first gene of interest 124, along with the first selection marker 130. The second sequence 110 may include the second promoter 112 operably linked to the gene for light-sensitive protein with a binding domain 114, along with a second selection marker 140.

[0058] In the decoupled system 105, the gene of interest 124, first promoter 122, and first selection marker 130 may be integrated at a first locus, while the gene for light-sensitive protein with a binding domain 114, second promoter 112, and second selection marker 140 may be integrated at a second locus in the host cell genome.

[0059] The present disclosure relates to a system for optogenetic protein production. In some cases, the system may comprise a plurality of host cells and a light source. The plurality of host cells may include host cells as described previously, such as those comprising the coupled system 100 or the decoupled system 105.Princeton- 101576

[0060] FIG. 2A and FIG. 2B illustrate activation states of a light-sensitive transcription factor in the system. As shown in FIG. 2A, in dark conditions, an EL222 protein 210 may remain inactive. FIG.2B depicts the system in an activated state, where exposure to light may cause the EL222 protein 210 to bind to the second promoter 112 and the gene for light-sensitive protein with a binding domain 114, enabling transcription activation.

[0061] In some implementations, the plurality of host cells may include a first host cell and a second host cell with different copy numbers of the light-sensitive transcription factor. For example, the first host cell may have 1 copy of the gene encoding the light-sensitive transcription factor, while the second host cell may have 3, 5, or 8 copies. In some cases, each host cell may have a copy number of the light-sensitive transcription factor that is 1-10.

[0062] The system may include a light source configured to activate the light-sensitive transcription factor. Activation of the light-sensitive transcription factor may induce expression of the first gene of interest 124. In some cases, the light source may be configured to provide light intensities of 5, 10, 30, 50, and 70 μmol / m2 / s. These specific light intensities may be tested to determine optimal conditions for protein production.

[0063] In some implementations, the system may utilize pulsed light schedules to tune expression levels. For example, light pulses may be applied in patterns such as 10 seconds ON / 100 seconds OFF or 1 second ON / 100 seconds OFF. These pulsed light schedules may allow for fine control of protein expression levels.

[0064] The system may be used with various carbon sources for cell growth. In some cases, glycerol may be used as a carbon source for the host cells.

[0065] To mitigate potential phototoxic effects, the system may incorporate co-expression of reactive oxygen species (ROS) scavengers with the light-sensitive transcription factor. In some implementations, ROS scavengers such as CTA1, SOD1, or PMP20 may be co-expressed with EL222 in the host cells.

[0066] FIG. 3 illustrates a block diagram of a bioreactor system that may be used for optogenetic protein production. The system may include a bioreactor vessel 310 connected to various control and monitoring components. A controller 320 may be operatively connected to multiple components of the system through control and data lines.

[0067] The bioreactor system may include several pumps: a pH pump 330 for pH control, a feed pump 332 for media addition, and an air pump 334 for aeration. An agitator 336 may provide mixing within the bioreactor vessel 310. A heat jacket 340 may surround the vessel for temperature control.Princeton- 101576

[0068] At the top of the bioreactor vessel 310, a condenser 350 may be installed, and a sample port 352 may allow for sample collection. The system may monitor various parameters including dissolved oxygen data 354, pH data 356, and temperature data 358, which may be transmitted to the controller 320.

[0069] For optogenetic control, a light panel 360 may be positioned beneath the bioreactor vessel 310. The light panel 360 may emit light 362 upward into the vessel to activate light- sensitive transcription factors within the culture. This specific bioreactor setup with bottom illumination may allow for efficient activation of the light-sensitive transcription factors in the host cells.

[0070] The controller 320 may receive input from the monitoring systems and control the operation of the pumps, agitator 336, and heat jacket 340 to maintain desired cultivation conditions. The system may allow for precise control of environmental parameters to optimize protein production using the optogenetic host cells.

[0071] The present disclosure relates to a method for optogenetic protein production. The method may involve providing a host cell with specific genetic components for light-controlled protein expression. In some cases, the host cell may comprise a first sequence including a first promoter operably linked to a gene of interest, and a second sequence including a second promoter operably linked to a gene encoding a light-sensitive transcription factor. The host cell may also include a selection marker.

[0072] In some implementations, the method may involve culturing the host cell under conditions suitable for growth. The growth conditions may be optimized based on the specific host organism and protein production requirements. For example, the host cell may be cultured in a medium containing an appropriate carbon source, such as glucose or glycerol.

[0073] The method may include exposing the host cell to light to activate the light- sensitive transcription factor. Activation of the light-sensitive transcription factor may induce expression of the gene of interest. In some cases, the light exposure may be controlled to optimize protein production. For example, the light intensity, duration, or pulsing pattern may be adjusted based on the specific light-sensitive transcription factor and host cell characteristics.

[0074] In some implementations, the method may involve determining the illuminating capability of a bioreactor system. This step may be important for selecting appropriate light exposure conditions and optimizing protein production. The illuminating capability may be assessed based on factors such as light intensity, penetration depth, and uniformity of illumination within the bioreactor.Princeton- 101576

[0075] The method may include selecting a light-sensitive transcription factor-containing parent strain based on the illuminating capability of a bioreactor system and protein production needs. In some cases, different parent strains may have varying copy numbers of the light- sensitive transcription factor gene, which may affect their sensitivity to light and protein production capabilities.

[0076] In some implementations, the method may involve cloning a plasmid and placing a gene of interest under control of a PC120 promoter with a zeocin marker. This step may allow for the introduction of a specific protein-coding gene into the host cell for optogenetic production.

[0077] The method may include forming the optogenetically-controlled strain by transforming the light-sensitive transcription factor-containing parent strain with the plasmid. This transformation step may integrate the gene of interest and associated regulatory elements into the host cell genome.

[0078] In some cases, the method may be used for the production of intracellular proteins. For example, the method may be applied to produce intracellular Mfp5 protein. The production of intracellular proteins may require optimization of cell lysis and protein extraction procedures following the light-induced expression period.

[0079] The method for optogenetic protein production may offer several advantages over traditional induction methods. Light-controlled expression may allow for precise temporal control of protein production without the need for chemical inducers. Additionally, the method may be adaptable to various host organisms and protein targets, providing a flexible platform for recombinant protein production.

[0080] The present disclosure relates to a hybrid machine-learning-supported method for light-induced recombinant protein production. The method may combine mechanistic modeling with machine learning techniques to predict and optimize protein production in optogenetic systems.

[0081] In some cases, the method may involve formulating differential equations to represent relevant system states. These states may include biomass, a protein of interest, glucose, and copy number. The copy number may be treated as a dynamic state that remains constant throughout the process. To achieve this, the right-hand side of the differential equation for copy number may be set to zero.

[0082] The method may include determining a target initial condition of the copy number. This step may allow for optimization of the initial copy number, which may correspond toPrinceton- 101576 selecting the most suitable strain exhibiting a specific copy number of the light-sensitive transcription factor.

[0083] In some implementations, the method may involve performing parameter estimation based on dynamic data from specific experiments. These experiments may be conducted under predetermined light intensity regimes and light-sensitive transcription factor copy numbers. The parameter estimation process may utilize a trained multi-input, single- output Gaussian process to predict each parameter as a function of the light intensity regime and light-sensitive transcription factor copy number.

[0084] The method may incorporate a hybrid Gaussian-process-supported mathematical model to predict system dynamics. This model may combine mechanistic knowledge of the biological system with machine learning components to enhance prediction accuracy and adaptability.

[0085] In some cases, the method may include incorporating predicted means of the trained multi-input, single-output Gaussian process as model parameters into mechanistic kinetic functions. This step may integrate the machine learning predictions into the mechanistic model, creating a hybrid approach that leverages both data-driven and knowledge-based modeling techniques.

[0086] The hybrid machine-learning-supported method may offer several advantages for optimizing light-induced recombinant protein production. By combining mechanistic modeling with Gaussian process regression, the method may provide more accurate predictions of system behavior under various conditions. This approach may enable more efficient optimization of protein production processes and facilitate the development of digital twins for optogenetically assisted recombinant protein production.

[0087] In some implementations, the method may be applied to predict and optimize the production of various recombinant proteins in different host organisms. The flexibility of the approach may allow for adaptation to different optogenetic systems and protein targets.

[0088] The method may also enable the implementation of model-based optimization and control strategies. For example, the hybrid model may be used to identify optimal light intensity regimes and copy numbers for maximizing protein production efficiency. Additionally, the method may support the development of closed-loop control systems for real-time optimization of protein production processes.

[0089] In some cases, the hybrid machine-learning-supported method may be extended to incorporate additional system states or parameters. For example, the method may be adaptedPrinceton- 101576 to include models of light penetration in high-density cultures or oxygen transfer limitations in large-scale bioreactors.

[0090] The method may also support continuous model improvement through adaptive modeling techniques. As new experimental data becomes available, the Gaussian process components of the model may be updated, potentially improving prediction accuracy over time.

[0091] In some implementations, the hybrid machine-learning-supported method may be integrated with other computational tools for protein production optimization. For example, the method may be combined with protein structure prediction algorithms or metabolic modeling approaches to provide a more comprehensive optimization framework.

[0092] The method may also be applied to study the relationship between light intensity, copy number, and protein production efficiency. This analysis may provide insights into the optimal design of optogenetic systems for different applications and scale-up scenarios.

[0093] In some cases, the hybrid machine-learning-supported method may be used to develop specialized software tools for modeling and optimizing optogenetic protein production processes. These tools may facilitate the adoption of optogenetic approaches in industrial biotechnology applications.

[0094] EXAMPLE

[0095] In this example, tools and methods for methanol-free induction of recombinant protein production in K. phaffii with light are discussed. The two general strategies developed, the coupled or decoupled systems, have different advantages: simplicity (coupled) versus versatility in strain design for varying light sensitivities to protein production (decoupled). Both systems are strongly activated in glucose or glycerol, eliminating the need to switch carbon sources mid-process as is the case with the PC120promoter. Furthermore, because K. phaffii generally metabolizes glucose and glycerol faster and can, under some conditions, reach higher biomass with these carbon sources than methanol, protein production can be higher using optogenetic induction under the tested conditions. Together, the disclosed optogenetic methods are promising for inducing protein production and should be considered as potential alternatives to PC120 or other media-based approaches.

[0096] In these examples, all plasmids were cloned into DH5α E. coli cells prepared using the Inoue method (Zhang et al., 2018). These constructs were assembled using Gibson assembly or restriction ligation with enzymes purchased from NEB. Primers were purchased from Integrated DNA Technologies, and fragments were PCR amplified using CloneAmp Hifi PCR premix from Takara Bio. The Mfp5, SR18, and β-lactoglobulin genes were codon- optimized for expression in K. phaffii and synthesized by Twist Bioscience, while otherPrinceton- 101576 sequences, including those of the PC120 promoter and EL222 were obtained from lab plasmids or genomic K. phaffii DNA. Plasmids and fragments were extracted and purified using Epoch DNA Miniprep and Omega E.Z.N.A. Gel Extraction kits. Final constructs were sequenced using Sanger sequencing by Genewiz or nanopore sequencing through Plasmidsaurus. All yeast strains were derived from NRRL Y-11430, which was acquired from ATCC. Transformations were carried out using the standard condensed electroporation method. Transformants were selected with either zeocin or G418 at concentrations ranging from 100- 1000 μg / mL and incubated for 2-3 days at 30 °C until colonies formed. All experimental growth was performed at 30 ºC.

[0097] For all microplate experiments, blue (465 nm) light was applied using LED panels (HQRP New Square 12” Grow Light Blue LED 14W) placed above the culture plate. The distance was adjusted for each panel (between 40-70 cm) to achieve the desired light intensity for each experiment (between 5 and 70 ^^^^^^^^ / ^^2^^ ), which was measured using a Quantum meter (Apogee Instruments, Model MQ-510) and equivalent to intensities between ~1.3 and ~18.2 W / m2(according to Ephoton=hc / λ, and a measured panel wavelength of ~462 nm). Cultures grown in the dark were shielded from light using aluminum foil.

[0098] Since the caps and holding clamps for Erlenmeyer flasks used in the examples would obstruct light and complicate application of a consistent intensity, light was applied using a different approach for flask experiments. Instead of illuminating from above, an apparatus was constructed to allow illumination by an LED panel from below. For this setup, a light panel was secured to the shaker with four brackets, facilitating the exchange of panels to achieve different light intensities. The panel was covered with a 12” x 12” x ⅛” frosted acrylic sheet (AZM Displays APSheet1 / 8Milky) with a reusable transparent adhesive pad on top (Chemglass CLS-4010). As with the microplate experiments, intensity was measured at 465 nm with a Quantum meter (Apogee Instruments, Model MQ-510).

[0099] To establish optogenetic control of gene expression in K. phaffii, EL222 was constitutively expressed using the native PADH2 promoter with the gene of interest downstream of the PC120promoter. The PADH2promoter was selected to express EL222 due to its previously reported moderate constitutive behavior in both glucose and glycerol. While PADH2 is strongest in ethanol, it is still active in other carbon sources, including methanol, glucose, and glycerol (at approximately 23%, 10% and 5%, of the ethanol levels respectively). Ethanol produced when growing in glucose is therefore expected to upregulate PADH2, especially as glucose is depleted. Using this promoter, we developed two light-activated systems that allow for either coupled (see FIG.1A) or decoupled (see FIG.1B) integration of EL222 and the gene of interest.Princeton- 101576 The coupled system, which benefits from simplicity, employs a single vector with a zeocin resistance selection marker, and thus integrates an equal number of copies of EL222 and the gene of interest. The decoupled system integrates these genes separately, which allows for independently selected copy numbers of EL222 and the gene of interest using G418 and zeocin, respectively. Both systems offer different advantages in the development of strains for two- phase processes characterized by a dark growth phase and light-induced production phase (see FIGS.2A-2B).

[0100] Under the tested conditions, the simpler coupled system (FIG. 1A) achieves stronger induction of intracellular protein production than the methanol-induced PC120 promoter. Comparing the normalized expression of yEGFP (^^^^^^^^^^ / ^^^^600) in strains containing a single copy of the expression cassette, using the coupled optogenetic system or PC120shows that blue light induction of EL222 yields approximately twice as much yEGFP than methanol induction (see FIG. 4A). For this test, an intensity of 70 ^^^^^^^^ / ^^2^^ was used, which falls within the typical range for microbial optogenetic experiments. In this case, light induction is also faster than methanol induction and shows similar strength in both glucose and glycerol. Because cells grow faster and to a higher density in these carbon sources than in methanol, light induction also leads to substantially higher overall intracellular protein production levels. Total production from the optogenetic strain continues to increase at 15 h, whereas PC120plateaus by that time, suggesting the coupled system could produce even more protein in longer cultivations. Additionally, optogenetic induction is tunable to a wide range of expression levels by adjusting the light dosage, offering flexibility for intermediate expression strengths. Having found it to be effective at a single integrated copy, it was next explored how this system performs at the higher copy numbers, more typical of production processes. When integrated at a higher copy number, the coupled optogenetic system remains competitive with PC120induction levels. Hypothesizing that the higher levels of EL222 expression may increase the light response sensitivity, strains containing approximately 8 copies of either the coupled optogenetic system or PC120were identified, and tested expression at both a low (5 ^^^^^^^^ / ^^2^^) and high light intensity (70 ^^^^^^^^ / ^^2^^). In both conditions, the expression strength when normalized by cell density (^^^^^^^^^^ / ^^^^600) is comparable to PC120, but light induction is substantially faster (see FIGS. 4B-4C). Interestingly, it was found that the optogenetic strain containing a high-copy number of EL222 (and yEGFP) is more productive at the lower light intensity (see FIGS. 4B-4C), suggesting possible phototoxicity at 70 ^^^^^^^^ / ^^2^^ consistent with reduced cell growth at this light intensity. Finally, though production remains strongPrinceton- 101576 compared to PC120 across copy number and light conditions, PC120 is slightly leakier, likely due to the lack of a repression mechanism of EL222. Having confirmed that the high-copy optogenetic strain excels at a low light intensity, it was sought to clarify the connection between EL222 copy number and light sensitivity.

[0101] To test the hypothesis that high EL222 expression increases photosensitivity, the copy number of EL222 and recombinant protein were decoupled (see FIG.1B). This approach was used to analyze growth and production at a diverse range of EL222 copy numbers and light intensities. To ensure that EL222 copy number was the only difference between strains, the PADH2-EL222 cassette was integrated into a parent strain already containing 12 copies of PC120- yEGFP. Transformants were selected and screened for a range of EL222 copy numbers (1, 3, or 8 copies) using qPCR.

[0102] Growth and fluorescence of each strain at different light intensities was then measured (see FIGS.5A-5F), compared to a control strain lacking EL222. As the goal was to characterize the relationship between EL222 copy number and light intensity, methanol- induced strains were not tested in this experiment. Consistent with the earlier observations, it was found that the strain with one copy of EL222 expresses better at high light intensities (50- 70 ^^^^^^^^ / ^^2^^ ), while the strain containing 8 copies is better at the lowest intensities tested, 5-10 ^^^^^^^^ / ^^2^^ (see FIGS.5A, 5E, 5G). Consistent with this trend, an intermediate 3 copies of EL222 reaches its highest activation at a moderate light intensity (10-30 ^^^^^^^^ / ^^2^^ ) (see FIG. 5C, 5G). In addition to inhibiting protein production, high light dosages dramatically reduce cell growth in the strain with 8 copies of EL222 (see FIGS.5B, 5D, 5F). In other words, high light intensities become detrimental only when EL222 is expressed, especially at high levels.

[0103] Regardless of EL222 copy number or light intensity, there is generally a decrease in fluorescence at later times after cell growth plateaus. This is likely due yEGFP gradually being degraded after the carbon source is fully consumed. Together, these observations highlight the importance of selecting the light dosage based on the EL222 expression level and argue against the tenet that higher light doses always lead to higher activation. To capture this complex interaction, a hybrid Gaussian-process-supported mathematical model was developed to fit the dynamics of the yEGFP-producing system described above, under varying conditions of EL222 copy number and light intensity.

[0104] Gaussian-process-supported modeling of yEGFP production.Princeton- 101576

[0105] To model the dynamics of protein production under varying EL222 copy number and light intensity, four differential equations were considered, as stated below. As proof of concept, it focused on yEGFP production, although the model can be readily adapted to other types of recombinant proteins. The model follows: ^^^^^^^^ ൌ ^^^^^^ ∙ ^^^^^^^, ^^^^^^^^ ൌ ^^^^, (3)^^^^^^^^ ൌ ^^^^^^^ െ ^^^ௗ ^ ^^^^^^^ ∙ ^^^^^^^, ^^^^^^^^ ൌ ^^^^, (4)^^^^^^^^ ^^^^^^^ ∙ ^^^^^^^^ (5)(6)

[0106] , theintracellular protein concentration (in arbitrary units of yEGFP), ^^^ ∈ ^^ is the glucoseconcentration (in g / L), and ^^ ∈ ^^ is the EL222 copy number. Despite being denoted as adynamic state, the EL222 copy number is a natural number and remains constant throughout the process, as it depends only on the initial condition (i.e., the selected strain). The process time is represented by ^^ and the initial process time by ^^0. Henceforth, the time dependency of the variables will be omitted.

[0107] Here, ^^, ^^^^, ^^^^, and ^^^^ are appropriate macro-kinetic kinetic functions: ^^(7) (8)(9)^^^ ൌ ^^^^ ∙ ^^, (10)

[0108] with parameters ^^^, ^^^, ^^, ^^, ^^^, and ^^^^. The specific growth rate, ^^, follows a hyperbolic function of ^^^^; the intracellular protein production rate, ^^^^, follows a hyperbolic function of ^^; the intracellular protein degradation rate, ^^^^, is a constant; and the specificglycerol uptake rate, ^^^^, is linearly coupled to ^^. The model parameters, collected in ^^ ∈ ^^^ഇ,are a function of the light intensity ^^ ∈ ^^ and the EL222 copy number. Therefore, ^^ ≔^^^^, ^^^, α, β, ^^^, ^^் ^^൧ and ^^=^^^^(^^,^^), where ^^ ^ఏ: ^^ ൈ ^^ → ^^ ഇ is a vector-valued functionparameter ^^^ ∈ ^^ is modeledas the mean function of a multi-input single-output Gaussian process ^^^^^^, thereby linking light intensity and EL222 copy number to the dynamic behavior of the system.

[0109] Denote ^^ ∈ ^^^ൈ^^ and ^^ ∈ ^^ଶൈ^^ as the training datasets for the output label (^^^^)and input features (^^ and ^^) of ^^^^^^, respectively. To populate these training data sets, one canPrinceton- 101576 conduct a parameter estimation routine for all batch experiments, each corresponding to a specific EL222 copy number and light intensity condition. The parameter estimation was performed using the particle swarm algorithm in COPASI. The model was fitted using the experimental data of the initial glucose concentration and the dynamic profiles of OD600and protein production under different conditions of light intensity and copy numbers.

[0110] The training of the Gaussian processes and the numerical integration of the hybrid model, with embedded Gaussian-process-predicted parameters, were performed using theHILO-MPC Python´s library. Overall, ^^^^^^ follows a normal distribution ^^^^^^^^^ ∼^^^^^^^^^, ^^^^^, ^^^^ with a mean function ^^: ^^^ೡ → ^^ and a kernel (covariance) function^^: ^^^ೡ ൈ ^^^ೡ → ^^. The vector ^^ ≔ ^^^, ^^^் ∈ ^^ଶ represents the features of ^^^^^. The Gaussianprocess models a true function under normally distributed measurement noise ^^ ∼ ^^^0, ^^ଶ^^,with zero mean and noise variance ^^ଶ ^^ ∈ ^^. For ^^^^ data entries, the Kernel matrix ^^ ∈ ^^ ^ൈ^^captures the neighborhood of data points in the feature space. Each element (^^,^^) of the Kernelmatrix, given two inputs ^^ ଶ ଶ^ ∈ ^^ and ^^^ ∈ ^^ , corresponds to the kernel function ^^^^^^, ^^^^. Inthis example, the ଶkernel function ^^ ^^^^, ^^^|^^^ , ^^^ was used, with signal variance^^ଶ ∈ ^ ଶ^ ^ and length-scale ^^ ∈ ^^ (given automatic relevance determination for the two features).The hyperparameters of the Gaussian process, ^^ ≔ ^^^ଶ^ , ^^, ^^ଶ் ^൧, were determined bymaximizing the log marginal likelihood ^^∗ ൌ^^^. The conditional posteriorprobability for a test input ^^∗, considering a Gaussian process with optimal hyperparameters, is normally distributed ^^(^^,^^)∼ ^^(^^^^,^^^^), where the predicted mean is the value of a modelparameter ^^^^ , and ^^^ ∈ ^^ represents the variance of the prediction. For simplicity of training,the Gaussianin this example were trained using the HILO-MPC’s default prior mean of zero. However, the prior mean can be reconsidered to enhance model generalizability on unseen data.

[0111] The hybrid model satisfactorily fits the dynamic behavior of cell density and protein concentration.

[0112] The phototoxicity of EL222 was investigated. One hypothesis is that activation of EL222 generates damaging reactive oxygen species (ROS), which could inhibit cell growth and protein production. Thus, higher EL222 levels and light intensities would increase ROS concentrations and exacerbate these effects. Though not previously shown with EL222, other light-responsive proteins containing a LOV domain are known to emit ROS, particularly singlet oxygen, when activated by light (Endres et al., 2018; Hernández-Candia et al., 2018).Princeton- 101576

[0113] To test whether the same is true for EL222, the endogenous superoxide dismutase (SOD1), peroxisomal membrane protein 20 (PMP20), and catalase (CTA1) genes, which code for the main ROS-scavenging enzymes in K. phaffii, were overexpressed. Specifically, co- expression of EL222 was tested with a total of three ROS scavengers native to K. phaffii (CTA1 catalase, SOD1 superoxide dismutase, and PMP20 peroxisomal membrane protein). The sequences for all three genes were identified in previous studies and amplified directly from the K. phaffii genome when cloning. In the case of CTA1, the last five amino acids were trimmed to remove the putative peroxisomal targeting tag and permit cytosolic expression. To construct strains co-expressing these scavenger enzymes with EL222, HIS4 was first knocked out from the wild-type NRRL Y-11430 strain using CRISPR-Cas9 (SMH60) to open a third selection marker, beyond zeocin and G418 resistance (ySMH68). This strain was then transformed with a cassette expressing yEGFP under PAOX1(SMH158, linearized with PmeI to target the AOX1 locus) to form ySMH196 (plated on 100 μg / mL zeocin), and subsequently transformed with a construct containing PADH2-EL222 (SMH139, linearized with ApaI to target the ADH2 locus) to form ySMH197 (plated on 1000 μg / mL G418). This ySMH197 strain served as the parent strain for ROS scavenger integration, targeting the HIS4 locus by linearizing SMH219, SMH229, and SMH230 with KasI and yielding ySMH207 (CTA1), ySMH209 (SOD1), and ySMH210 (PMP20), respectively. These ROS scavengers were constitutively expressed from the PGAP promoter.

[0114] To test whether co-expression of the scavengers could offset the phototoxic effects of EL222, four colonies of ySMH207, ySMH209, and ySMH210 were inoculated into 500 μL BMD1 medium supplemented with 0.3 mM L-histidine in a 48-well plate, along with ySMH196 (no-EL222 control) and ySMH197 (no scavenger control). Because the ySMH196 and ySMH197 control strains are auxotrophic for L-histidine, this supplemented media was used for all wells to keep conditions consistent between strains (despite ySMH207, ySMH209, and ySMH210 being prototrophic for L-histidine). This plate was grown overnight for 20 h in the dark at 30 °C with shaking at 200 rpm. The next day, the plate was centrifuged at 234g for 5 min and resuspended in fresh BMD1 medium supplemented with 0.3 mM L-histidine. The strains were diluted in two identical 48-well plates to an OD600 of 0.1 in BMD1 supplemented with 0.3 mM L-histidine as measured by a TECAN plate reader (Infinite F Plex). As previously, the L-histidine was added to all cultures to keep media conditions consistent between the prototrophic scavenger-expressing strains and the auxotrophic control strains. Finally, the two plates were incubated at 30 °C with 200 rpm shaking in their respective light conditions (darkness or 70 ^^^^^^^^ / ^^2^^ light). Plates were briefly removed at the various timepoints (12 h,Princeton- 101576 18 h, 24 h) to measure the OD600, with samples being diluted in a separate 48-well plate for timepoints where the OD600was above 8 and exceeded the linear range of the instrument.

[0115] Unlike the parent strain housing 7 copies of EL222 without overexpressing an ROS scavenging enzyme, the strain overexpressing PMP20 grew equally well with or without light exposure while the strains overexpressing CTA1 or SOD1 did not improve growth in the light. Given that the main ROS generally emitted by phototoxic LOV domain-containing proteins is singlet oxygen, and that Cta1p and Sod1p consume hydrogen peroxide and superoxide respectively, it is not surprising that these enzymes did not protect cell growth from the toxic combination of blue light and EL222. While the ROS substrates for Pmp20p are less characterized than for Sod1p and Cta1p, these results support the hypothesis that ROS emission by EL222 is the cause of its phototoxic effects. This ROS emission, which likely increases with EL222 copy number, offers new insights into improving production by optimizing the levels of EL222 and light. The best strategy to achieve this is to employ the decoupled optogenetic system since it integrates EL222 and the gene of interest independently (see FIG. 1B). The decoupled optogenetic system was utilized to produce proteins beyond yEGFP, including Mfp5 mussel foot protein, β-lactoglobulin, and a nanobody against SARS- CoV-2.

[0116] Optogenetic induction for intracellular production of Mfp5 mussel foot protein.

[0117] To demonstrate the capability of optogenetics to control the intracellular production of proteins besides yEGFP, the decoupled system was used to produce the Mfp5 mussel foot protein from Mytilus californianus, which has applications in the biomaterials industry. A strain containing a single copy of EL222 was transformed with a gene cassette containing PC120-Mfp5 and a zeocin selection marker. To compare light with methanol induction, the wild-type strain was also transformed to express Mfp5 with PAOX1instead of PC120 and tested for production with methanol or blue light induction. After screening 14 optogenetic strains induced with blue light (at 50 ^^^^^^^^ / ^^2^^ intensity) and 14 methanol- induced strains for Mfp5 production, the top producers of each were compared in triplicate. See FIG. 6. Under the tested conditions, it was found that Mfp5 production with the optogenetic system is greater than with PAOX1. Strains were induced at identical cell densities (OD600 = 10) and after 48h of induction grew to comparable biomass in glycerol and methanol media, with slightly lower growth in glucose. Nevertheless, an equal amount of biomass for each sample was loaded on the polyacrylamide gel (see FIG.6) to ensure a fair comparison.

[0118] In particular, to test intracellular production of Mfp5, Mfp5 from Mytilus californianus was codon-optimized for expression in K. phaffii (Twist Bioscience) using a previously reported amino acid sequence. Constructs expressing the gene from both PA0X1 andPrinceton- 101576 PC120 (SMH207 and SMH244) were created. For expression from PAOX1, SMH207 was transformed into NRRL Y-11430 cells to yield ySMH186, which was plated on 500 and 1000 μg / mL zeocin agar plates. For optogenetic production, a strain was first created containing a single copy of PADH2-EL222 by transforming NRRL Y-11430 with the linearized SMH139 plasmid (ySMH203-16). To incorporate the Mfp5 expression cassette, this strain was transformed with SMH244, yielding ySMH238. This transformation was plated on 500 and 1000 μg / mL zeocin agar plates.

[0119] To screen the transformants for strong producers, 14 colonies of each strain were inoculated into minimal media (BMG1 for the ySMH186 PAOX1-driven strains and BMD1 for the optogenetic ySMH238 strains) in a 24-well plate and grown for 24 h at 30 °C with 200 rpm shaking. This growth was performed in the dark for the optogenetic strain to avoid premature induction. Subsequently, the cultures were centrifuged for 5 min at 234g and resuspended in fresh minimal media (BMM1 for ySMH186 and BMD1 for ySMH238). The cell densities were measured using a TECAN plate reader (Infinite F Plex) and equalized to an OD600of 10 in their respective media in 24-well plates. These plates were incubated for 24 h at 30 °C with shaking at 200 rpm, with the optogenetic plate exposed to blue light at an intensity of 50 ^^^^^^^^ / ^^2^^. After this period, cells were resupplied with their respective carbon sources to a final concentration of 1% (using 100% methanol for ySMH186 or 40% glucose for ySMH238) and returned to 30°C for an additional production period of 24 h. The cultures were then lysed at an equal cell mass using a standard trichloroacetic acid (TCA) lysis protocol. For the TCA lysis protocol, the equivalent of 5 mL of cells were harvested at an OD600 of 2 to keep the same amount of cells in each sample. The cells were centrifuged for 2.5 min at 2500 rpm in a microcentrifuge and removed the supernatant. The pellet were resuspended in 1 mL of water and transferred it to a 1.5 mL tube. The cells were pelleted again with the same centrifugation method and the water was removed. Next, the cells were resuspended in 1 mL of 5% TCA and incubated on ice for at least 10 min. Then, the cells were centrifuged for 2 min at maximum speed in a microcentrifuge and the supernatant removed. The pellet was washed once by adding 500 ^^L of 1 M of Tris (without adjusted pH) and inverting the tube. Once more, the cells were centrifuged for 2 min at maximum speed before removing the supernatant. The pellet was then resuspended in 113 ^^L of water by sonicating on low power, which only took a few seconds. Finally, 20 ^^L of 1 M DTT and 67 ^^L of 3X SDS loading buffer was added, to end with ~200 ^^L of sample. Samples were then boiled in a sand bath (at 100°C) for 5 min and thenPrinceton- 101576 centrifuged in a microcentrifuge for 3 min at 12,000 rpm. Finally, 5-20 ml of samples were loaded on the SDS PAGE gel.

[0120] Next, SDS-PAGE was performed followed by western blot to identify the best performing colonies of each strain. For any intracellular production strains in this example, samples were lysed using the aforementioned TCA protocol followed by resuspending in a SDS sample buffer. For any secreting strains in this example, the final cultures were centrifuged for 5 min at 234g and the supernatant was mixed with the sample buffer. Samples were boiled for 5 min at 100 °C using a heat block (VWR), loaded onto 12% SDS-PAGE gels, and resolved. A loading volume of 15 μL was used for all SDS-PAGE gels. After resolving, analysis by western blot using established protocols was performed. Proteins were transferred onto a PVDF membrane using a Trans-Blot Turbo Transfer System (Bio-Rad). All proteins analyzed by western blot contained a C-terminal His tag which was assayed using THE™ His Tag Antibody [HRP], mAb (Genscript) with a dilution of 1:10000. Blots were revealed using Clarity ECL substrate (Bio-Rad) and imaged with a Bio-Rad ChemiDoc MP Imaging System with Image Lab software, using the chemiluminescence protocol. After imaging, blots could be quantified by using Fiji ImageJ software to correlate pixel intensity to protein production.

[0121] After identifying the highest producing strains (ySMH186-9 and ySMH238-9), production from these strains was more rigorously compared in triplicate in Erlenmeyer flasks. Each strain was inoculated into 1 mL of minimal media (BMG1 for ySMH186-9 and BMD1 for ySMH238-9) in a 24-well plate and grown for 20 h at 30 °C with 200 rpm shaking. After this growth, 400 μL of each culture was inoculated into 40 mL of fresh media (BMG1 for ySMH186-9 and both BMG1 and BMD1 for ySMH238-9) in 500 mL Erlenmeyer flasks for a second overnight growth. As previously, the optogenetic cells were grown in the dark. The cells were then centrifuged at 234g for 5 min and resuspended in 40 mL fresh minimal media (BMM1 for ySMH186-9 and both BMG1 and BMD1 for ySMH238-9). The cell densities were measured using an Eppendorf spectrophotometer (Eppendorf BioSpectrometer basic) and μCuvette (Eppendorf μCuvette G1.0). To ensure the measured cell densities were within the linear range of the instrument (OD600<10), 0.25 mL samples of each culture were first diluted 1:5 in fresh media before measurement. Then, additional fresh media was added to each culture such that they were all at an OD600 of 10 (BMM1 for ySMH186-9 and both BMG1 and BMD1 for ySMH238-9). This step ensured that the cell density at the time of induction would be equal for all conditions. After equalizing the cell densities to an OD600 of 10, 12.5 mL of each culture was transferred to three identical 125-mL Erlenmeyer flasks (nine flasks total), which were then incubated at 30 °C with shaking at 200 rpm. The flasks with optogenetic strains werePrinceton- 101576 exposed to light at an intensity of 50 ^^^^^^^^ / ^^2^^ using the flask illuminator platform, while the PAOX1-driven strain was exposed to ambient light. After 24 h of incubation, the cells were resupplied with their respective carbon sources to a concentration of 1% (using 100% methanol, 40% glucose, or 60% glycerol) and returned to 30 °C for an additional period of 24 h. Equal cell masses of each culture were lysed using the aforementioned TCA protocol and quantified production using a western blot and ImageJ analysis. To confirm equal loading of samples and avoid bias in the results, the membrane was stained using Coomassie blue R-250 as previously described.

[0122] Optogenetic induction of secreted proteins.

[0123] Next, protein secretion, which often simplifies industrial downstream purification operations, was considered. First, it was investigated whether the inhibitory effects of high EL222 levels at high light intensities also occur when secreting proteins. To test this, two strains that secrete yEGFP from PAOX1 were compared, differing only in whether they co- express EL222. In this part of the experiment, EL222 is not actually controlling protein expression, so strains should perform comparably regardless of light intensity if EL222 is not a factor. However, protein secretion from the EL222-expressing strain was severely impaired by any exposure to light. See FIGS. 7A-7B. In contrast, the control strain is relatively unaffected at all intensities tested.

[0124] These results are consistent with the intracellular experiments (see FIGS. 5A-5F), showing that EL222 activation at suboptimal light conditions affects growth and protein production, whether it is secreted or produced intracellularly. With this insight, the decoupled optogenetic system was applied for protein secretion, carefully selecting light intensities based on the copy number of EL222. As a first target, yEGFP secretion was compared between optogenetic and methanol-induced strains. For the optogenetic system, a set of platform strains was first constructed with various copy numbers of EL222, which were verified by qPCR. Strains with both low (one) and high (five) copies of EL222 were then transformed with PC120- yEGFP and tested for yEGFP secretion at innocuous light intensities: 50 ^^^^^^^^ / ^^2^^ for 1 copy of and 5 ^^^^^^^^ ^^2^^ for 5 copies of EL222.were screened by dot blot to identify suitable producers. The culture plates were centrifuged for 5 min at 234g and filtered 350 μL of supernatant from each well through a nitrocellulose membrane in a Bio-Dot Microfiltration Apparatus (Bio-Rad) using the recommended protocol for the equipment. After filtration, membranes were incubated for 1 h at room temperature in a blocking buffer (1X TBS, 0.1% Tween-20 with 5% w / v nonfatPrinceton- 101576 dry milk) and assayed using THE™ His Tag Antibody [HRP], mAb (Genscript) at a 1:10000 dilution. After incubating with the antibody for 1 h at room temperature, the membranes were washed four times for 10 min with TBST (1X TBS, 0.1% Tween-20). Finally, the blots were imaged with Clarity ECL substrate (Bio-Rad) on a Bio-Rad ChemiDoc MP Imaging System using the chemiluminescence protocol.

[0126] After selecting the top optogenetic and methanol-induced strains, their yEGFP secretion was compared in triplicate, with the optogenetic strains cultivated (and induced) in media containing glucose or glycerol. See FIGS. 8A-8B. Under the tested conditions, quantitative western blots show that the optogenetic strains in glycerol media produce more yEGFP than the methanol-induced strain by approximately two-fold. However, unlike with intracellular production, optogenetic secretion in glucose is weaker than PAOX1. The cause of this discrepancy is not clear but is consistent across the low- and high-EL222 strains. Nevertheless, the strong production in glycerol motivated the testing of secretion of other proteins of higher value.

[0127] To show that optogenetic secretion is effective for proteins beyond yEGFP, it was used to produce β-lactoglobulin dairy protein and the SR18 nanobody which neutralizes the SARS-CoV-2 virus spike protein. As with yEGFP, dot blots were used to screen optogenetic and methanol-induced strains and the best producers were compared in triplicate. Strains were induced at identical cell densities (OD600= 10) and after 48h of induction the cells grew differently in different carbon sources.

[0128] In most cases, the strains grew to comparable biomass in glycerol and methanol media, whereas growth in glucose never exceeded that of strains grown in methanol. For both proteins, under the tested conditions, light induction of PC120achieved higher production in glycerol compared to methanol induction of PAOX1, regardless of EL222 copy number. See FIGS. 8C-8F. As previously observed, secretion in glucose is weaker than in glycerol and generally weaker than PAOX1. Together, these results show that light can be a powerful alternative to methanol induction for diverse applications, particularly in glycerol media.

[0129] In particular, for these secretion examples, the methanol-induced strains were created by transforming NRRL Y-11430 with a construct secreting the gene of interest from PAOX1 using the α-factor secretion tag, yielding ySMH154 (yEGFP), ySMH223 (SR18), and ySMH51 (β-lactoglobulin), all using PmeI to linearize the plasmids. The optogenetic strains were constructed by transforming ySMH203-16 (single copy of EL222) and ySMH203-2 (five copies of EL222) with plasmids linearized with PmeI to secrete the gene of interest from PC120using the same α-factor secretion tag. The resulting strains with one copy of EL222 werePrinceton- 101576 ySMH217 (yEGFP), ySMH230 (SR18), and ySMH224 (β-lactoglobulin), while the strains with five copies of EL222 were ySMH219 (yEGFP), ySMH231 (SR18), and ySMH226 (β- lactoglobulin). All transformations were plated on 500 and 1000 μg / mL of zeocin.

[0130] 94 colonies of each strain were screened in a 96-well format in a deep-well plate (USA Scientific). For the methanol-induced strains, each colony was inoculated into 500 μL of BMG1 medium in the deep-well plate, with the last two wells saved for wild-type and blank negative controls. After one day incubation period at 30 °C and shaking at 300 rpm, the plates were centrifuged at 234g for 5 min and the supernatant was replaced with 500 μL of fresh BMM1 medium. The cells were induced for 48 h at 30 °C with 300 rpm shaking, with methanol added after 24 h to return the final concentration to 1% (assuming all the initial methanol had been consumed in the first 24 h). After the production phase, the plates were again centrifuged for 5 min at 234g and 350 μL of supernatant from each well was analyzed by dot blot. The highest-producing colony of each strain was selected for future experiments at the flask-scale. To screen the optogenetic strains, 94 colonies of each strain were inoculated into 500 μL BMD1 medium in the deep-well plates and grown in the dark for one day at 30 °C with 300 rpm shaking. The plates were then centrifuged for 5 min at 234g and resuspended in fresh BMD1 medium. The plates were placed in their respective suitable light conditions (5 ^^^^^^^^ / ^^2^^ for the 5-copy EL222 strains and 50 ^^^^^^^^ / ^^2^^ for the 1-copy EL222 strains) to produce for 48 h at 30°C with shaking at 300 rpm. Glucose was added to a final concentration of 1% at the 24- h point of this incubation (assuming all the initial glucose had been consumed in the first 24 h). Finally, the plates were centrifuged at 234g for five min and analyzed by dot blots as with the methanol-induced strains. With the best producer of each strain identified, tests in shake flasks were performed to compare production between optogenetics and PC120in triplicate. First, each strain was inoculated into 1 mL of minimal media (BMG1 for ySMH154-C5, ySMH223-G5, and ySMH51-D2 and BMD1 for ySMH217-A2, ySMH219-D7, ySMH230-A6, ySMH231-G2, ySMH224-A11, and ySMH226-C1) in a 24 well plate and grown for 20 h at 30 °C with 200 rpm shaking in the dark. After this growth, 400 μL of each culture was inoculated into 40 mL of fresh media (BMG1 for the methanol-induced strains and both BMG1 and BMD1 for the optogenetic strains) in 500 mL Erlenmeyer flasks for a second overnight growth in the dark. After 20 h, the cultures were centrifuged for 5 min at 234g and resuspended in 40 mL fresh media again (BMM1 for the methanol-induced strains and both BMG1 and BMD1 for the optogenetic strains). The cell densities were measured with an Eppendorf spectrophotometer (BioSpectrometer basic) and equalized to an OD600 of 10 in their respective media. 12.5 mL of each culture was transferred to three identical 125-mLPrinceton- 101576 Erlenmeyer flasks, which were then incubated at 30 °C with shaking at 200 rpm. The optogenetic flasks were exposed to light at an intensity of 50 ^^^^^^^^ / ^^2^^ for the 1-copy EL222 strains and 5 ^^^^^^^^ / ^^2^^ for the 5-copy EL222 strains using an illuminator platform. After 24 h of incubation, the cells were resupplied with their carbon sources to a concentration of 1% and returned to 30 °C for an additional 24 h. The fermentations were harvested by centrifuging the cultures for 5 min at 234g and analyzing the supernatant using quantitative western blots.

[0131] Scale up of light induction to bioreactor.

[0132] Having achieved strong light induction of protein production in shake flasks, the disclosed system was next tested in a lab-scale bioreactor. As the SR18 nanobody is the highest value product of the three targets examined, it was selected as a representative example for scale-up. Although the strain with five copies of EL222 requires less light than the single copy strain, this example moved forward with one copy of EL222, as a proof of concept, because it offers a wider range of effective light dosages. See FIG.5A-5F. It also constitutes the strictest option to explore the potential challenge of limited light penetration in (large-scale) bioreactors.

[0133] Production of the SR18 nanobody was scaled to the 2-L bioreactor level using the ySMH223-G5 (methanol-induced) and ySMH230-A6 strains (optogenetic with 1 copy of EL222), which were identified as the best producers by dot blot screens. Both strains were first inoculated into 5 mL of YPD media with 100 μg / mL zeocin and grown in the dark for 20 h with 200 rpm shaking. We used 2 mL of the grown cultures to inoculate 50 mL of YPD in a 500 mL Erlenmeyer flask containing 100 μg / mL zeocin and grew them again for 20 h in the dark. After this growth, the cells were concentrated by centrifuging for 5 min at 234g and removing 25 mL of the supernatant, after which the OD600 was measured using an Eppendorf spectrophotometer (BioSpectrometer basic). To ensure the measured cell densities were within the linear range of the instrument (OD600<10), a 0.1 mL sample of each culture was first diluted 1:10 in fresh media before measurement. The bioreactor experiments followed a pre-defined recipe, arbitrarily chosen and not derived from any design of experiments or optimization procedure. However, it serves the purpose of providing preliminary insights into the process dynamics at this scale.

[0134] A BioFlo120 system was set up with a 2-L bioreactor (see FIG.3) and 1 L of basal salts medium (BSM) supplemented with 4.35 mL PTM1 trace salts (Invitrogen, 2000) and 4% w / v glycerol were added, along with 100 μL of Antifoam 204 (Sigma-Aldrich). The reactor was set to 30 °C with a pH of 5.0, which was maintained using 14% ammonium hydroxide. After reaching these setpoints, the dissolved oxygen probe was calibrated and set to a minimum percentage of 30% by adjusting the agitation speed (200-600 rpm) and air flow rate (0.1-3Princeton- 101576 SPLM). Cell density was measured using a noninvasive biomass sensor with a wide linear range from less than 0.1 to greater than 300 OD units (bug lab BE2100). The reactor was then inoculated with the concentrated culture (either ySMH223-G5 or ySMH230-A6) to an OD600 of 1. The cells were grown for ~24 h in the repressive conditions (darkness in the case of the optogenetic ySMH230-A6 strain) until a spike in the dissolved oxygen was observed, which signaled the end of the batch phase. After the batch phase, a brief semi-batch phase was performed (still in repressive conditions for both the optogenetic and non-optogenetic bioreactor experiments) by feeding 50% w / v glycerol supplemented with 12 mL / L of PTM1 for 4 h at a rate of 6 mL / h. A spike in the dissolved oxygen was observed after stopping the feed, indicating full consumption of glycerol.

[0135] To induce production in the PAOX1-controlled strain, 100% methanol supplemented with 12 mL / L PTM1 was fed to the reactor at an initial rate of 3.6 mL / h. After 4 h at this rate, the feed was turned off until another spike in the dissolved oxygen was observed, after which the flow rate was increased to 7.2 mL / h. The methanol was fed at this higher flow rate for 48 h (the remainder of the fermentation). To prevent excess foaming, 100 μL of Antifoam 204 was added approximately every 12 h after the onset of induction. Samples of 5 mL were taken at each timepoint post-induction for analysis using Coomassie staining of SDS-PAGE gels and Bradford assays.

[0136] To induce production in the optogenetic bioreactor runs, an LED panel positioned under the reactor was illuminated, and the glycerol mixture (50% w / v glycerol supplemented with 12 mL / L PTM1) was fed at a rate of 4.5 mL / h. This feed rate was selected to provide an equivalent amount of carbon source on a mass basis as the methanol-induced experiments. After 4 h, the flow rate was increased to 9 mL / h for the remainder of the fermentation. Antifoam 204 was added, and samples were taken as aforementioned in the methanol-induced bioreactor runs.

[0137] Samples were analyzed using Coomassie staining of SDS-PAGE gels and Bradford assays at the same time and with the same reagents as the methanol-induced cultures. For the Bradford assays, bovine serum albumin (BSA) (Sigma-Aldrich) was used to generate the standard curve.

[0138] Although neither the optogenetic nor methanol-inducible processes were a priori optimized, and therefore cannot be directly compared to determine superiority, these experiments provide a foundation for comparing the dynamics of both induction systems and grant insight towards future process optimization (e.g., induction time, feed rate, DO, etc.). The optogenetic and methanol-induced strains were similarly productive for the first 16 h, withPrinceton- 101576 both secreting approximately 0.5 g / L of SR18; however, under the tested conditions, the optogenetic strain outperforms the methanol-induced process at every subsequent timepoint, ultimately reaching a final titer of 1.5 ± 0.6 g / L by the end of the fermentation, as measured by Bradford assay. See FIG.9A.

[0139] The optogenetic strain likely benefits from the lower oxygen requirement of glycerol metabolism relative to methanol along with the lower toxicity of this substrate, which allows for faster consumption, lower levels of oxygen, and potentially greater production.

[0140] To confirm that most of the secreted protein in these experiments was the intended SR18 nanobody, the Bradford analyses were paired with gel electrophoresis. In both systems, as expected, no protein secretion was detected at the onset of induction (during the glycerol phase for PAOX1 strain, and in the dark for the optogenetic strain). Yet, this nanobody seems to be unstable given its appearance as multiple bands in the stained gels; a trait to consider if this protein is selected for further bioprocessing.

[0141] Nevertheless, since this observation occurs in both the optogenetic and PAOX1- driven processes, the overall titers remain a fair comparison between the two systems, even if the Bradford analysis may overestimate the absolute quantity of functional protein.

[0142] In addition to being two orders of magnitude greater in volume than the shake flask tests, the bioreactor cultures reached exceptionally high cell densities, with the light-induced cells growing more than their methanol-induced counterparts by almost two-fold. See FIG. 9B. Since greater biomass is not necessarily an advantage for secreted proteins due to biomass- product metabolic tradeoff, there is likely an opportunity to further increase the volumetric productivity of the recombinant protein by optimizing the dynamic control strategy to add balance between cell growth and production. Ultimately, the successful production of SR18 by optogenetic induction with only one copy of EL222 (the least light-sensitive strain developed in this example) shows that light penetration is not an obstacle at least at this scale.

[0143] While a previous study (Z. Wang et al., 2022b) showed that EL222 is functional in K. phaffii, it fell short of exploring conditions in which light induction could be comparable to methanol induction. In contrast, the system disclosed herein is carefully compared the performance of different optogenetic strain designs with that of PAOX1, controlling for critical factors like copy number and optimal light intensity. This type of characterization connecting EL222 expression level to optimal light dosage for gene induction had not been previously reported in yeast or any other organism.

[0144] However, adjusting light dosage in the context of EL222 copy number proved essential for optogenetic induction of protein production in K. phaffii at levels that rivalPrinceton- 101576 methanol induction. This suggests that similar characterization of other optogenetic systems involving different light-responsive proteins will likely be valuable to expand the capabilities of those systems.

[0145] A key finding in this example is the importance of adjusting light intensity depending on the EL222 copy number. Blue light at all intensities used in the example did not show any phototoxicity to the wild-type strain. However, the higher light doses impair growth and production of both intracellular and secreted recombinant proteins in strains containing high copies of EL222. See FIGS. 5A-5F, 7A-7B. This observation poses a caveat to the frequently discussed low-toxicity of optogenetics, in that EL222 is only non-toxic below a certain light threshold that depends on EL222 concentration. As disclosed herein, maximal light exposure activates EL222 maximally in S. cerevisiae without any observable phototoxicity, but all strains in those studies contained only one copy of EL222. Consistent with these trends, some studies report poor growth when strongly expressing other optogenetic regulators. Our findings suggest that these systems could be more effective and less toxic if tested at lower light doses. Therefore, the relationship between a suitable light dosage and EL222 copy number contradicts the common perception that maximum light is always associated with maximum induction. In this example, the light dosage was characterized by intensity, but as will be understood in the art, utilizing light pulses is also a suitable strategy for input delivery, as the tunability of the disclosed low copy yEGFP strain suggests.

[0146] The protective effect of PMP20 suggests that the inhibitory effect of light on the growth and protein production of engineered strains stems from the emission of ROS by EL222, which is exacerbated when cells expressing high levels of this transcription factor are exposed to excessive light. This is consistent with previous studies on the generation of ROS by LOV-based proteins. Pmp20p is an endogenous peroxiredoxin involved in degrading hydrogen peroxide and organic hydroperoxides at the peroxisome membrane surface. While it is known to be upregulated in methanol and virtually absent in glycerol, its activity in ROS- generating conditions without methanol has not been examined. ROS emitted by LOV-based proteins are generally singlet oxygen and, to a lesser degree, hydrogen peroxide. This raises the possibility that Pmp20p is also capable of neutralizing singlet oxygen, or that EL222 emits hydrogen peroxide as the cause of its phototoxicity. It is envisioned that optimizing expression levels of PMP20 in engineered strains would increase their tolerance to higher EL222 and blue light levels to further enhance protein production. Another strategy worth exploring in future studies to reduce phototoxicity is to use EL222 mutants with increased light sensitivity, which might be effective at lower copy numbers than the wildtype EL222.Princeton- 101576

[0147] Understanding the link between EL222 concentration and phototoxicity can simultaneously enable optimization of light intensity throughout the process and facilitate the selection of the most suitable system for a given application. For strains requiring low copy numbers of the target protein, the simplicity of the coupled system makes it the preferred choice as it requires only one transformation. This strategy can be implemented by, e.g., replacing yEGFP in the SMH131 plasmid with the gene of interest and integrating into K. phaffii using zeocin selection. However, in situations where high copy numbers of the gene of interest are preferred, the decoupled system is a better option to avoid the toxic effects of high copy numbers of EL222. It is worth noting that, although the activity of PADH2 driving EL222 expression is slightly different in glycerol and glucose, and is expected to vary if ethanol is produced from the latter, these variations are accounted for in the optimization experiments and reflected in the final conditions found to be optimal for the production of each specific protein. To support adoption of the disclosed strategy, a set of useful plasmids and a set of strains with a range of EL222 copy numbers was produced that can be used as parent strains in which to integrate the gene of interest under the control of PC120 at varying and decoupled copy numbers. See Tables 1 and 2 below: Table 1 (Plasmids Used. TT – Terminator, αsec – α-factor secretion tag, * – these plasmids also contain PAOX1 which is solely used for integration at the PAOX1 locus and does not drive expression of any genes, – plasmids belonging to the toolkit to easily implement light induction of protein production in, e.g., K. phaffii) Integration tPrinceton- 101576 HIS4 CRISPPHTX1-Cas9-TTDAS1,SMH60RP -gRNA - Zeocin N / A knockoutHTX1 HIS4TT ) ) )

[0148] Table 2 (K. phaffii strains used. ZeoR – Zeocin resistance, AmpR – Ampicillin resistance, KanR – Kanamycin / G418 resistance, βlacto– β-lactoglobulin, – strains belonging to the toolkit to easily implement light induction of protein production in K. phaffii.) Strain Description Genotype Plasmid usedPrinceton- 101576 NRRL Y- Wild-type parentWTN / A11430 strain Intr lllrPrinceton- 101576 copies of EL222 (used for decouled sstemPrinceton- 101576 β-lactoglobulin ySMH224- secretion with 1 co of EL222 ySMH203-16 PAOX1::(PC120- αsec- SMH240

[0009] s can e ac eve y repac ng y n p asm or p n 244 with the gene of interest, depending on whether the protein is meant to be secreted or produced intracellularly, and using the resulting plasmid to transform the parent strains with zeocin selection. For applications where light penetration may be a concern (for example in large- scale bioreactors), ySMH203-2 and ySMH203-21 (containing 5 and 8 copies of EL222, respectively) may be the best strains to use, as they are more responsive to lower light levels. In cases where light penetration is not a concern, strain ySMH203-16 (housing only 1 copy of EL222) offers a more forgiving range of suitable light intensities, making it easier to avoid phototoxicity.

[0150] The hybrid dynamic model for yEGFP production, influenced by the EL222 copy number and light intensity, outlines a suitable strategy for capturing the dynamic behavior of optogenetic bioprocesses for recombinant protein production. This modeling strategy holds the potential to unlock advanced open-loop and closed-loop model-based optimization schemes in future studies. It is envisioned that optimal control problems could be formulated to maximize bioprocess performance, exploiting the available degrees of freedom while subject to the constraints imposed by some knowledge of the system (in this case, the hybrid dynamic model),Princeton- 101576 as well as potential economic, technical, safety, and environmental considerations. In the context of optogenetic recombinant protein production, degrees of freedom of interest include variables such as inoculum size, initial substrate concentration, strain selection with a specific EL222 copy number, dynamic light intensity trajectories as in two-phase processes, and feed rates as in fed-batch bioreactors.

[0151] When applying the decoupled system to produce secreted proteins, an unexpected finding was that production in glucose is weaker than in glycerol and, in most cases, methanol. This was observed for all three protein targets tested (yEGFP, β-lactoglobulin, SR18) and at both low and high EL222 copy numbers. Though the mechanism causing this difference is not understood, the inventors hypothesize that it is related to some interference between glucose, light, and the regulation of the secretion pathway, as intracellular protein production levels using light induction are similar between glucose and glycerol. Nevertheless, under the tested conditions, it was found that glycerol is the best carbon source for optogenetic control of protein secretion. Thus, glycerol was selected as the substrate when scaling to bioreactors.

[0152] The bioreactor experiments show that light induction of protein secretion can be successfully achieved in 1-L cultures at very high cell densities, even with a strain containing only one copy of EL222. Using this strain provided a wider range of suitable light conditions, which facilitated the scaleup process. However, this strain also shows the lowest light responsiveness of the parent strains developed, leaving open the possibility of examining different light doses at this scale. These results also raise the prospect of obtaining robust light induction in larger pilot- or demonstration-scale bioreactors using strains with higher copy numbers of EL222, which are at least 10 times more sensitive to light (based on observations that strains with one copy need 50 ^^^^^^^^ / ^^2^^ to maximize induction, while strains containing eight copies are fully active at only 5 ^^^^^^^^ / ^^2^^ , and possibly less). See FIGS.5A-5F.

[0153] It is envisioned that further scaleup of light-induced protein production in K. phaffii can be conducted by adapting existing industrial photobioreactors used in microalgae and cyanobacteria processes, by applying known photobioreactor design principles to create new systems. Other possible solutions using existing technologies include industrial-scale transparent plastic bioreactors, submersible LEDs, and transparent illumination chambers through which cultures may be recirculated from large stainless steel bioreactors. Altogether, the example results reduce concerns of light penetration in lab-scale bioreactors and bode well for future development of optogenetically controlled industrial-scale processes.

[0154] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scopePrinceton- 101576 of the disclosure. Accordingly, other implementations are within the scope of the following claims.

Claims

Princeton- 101576 CLAIMS 1. A host cell for optogenetic protein production, comprising: (a) a first sequence comprising a first promoter operably linked to a gene of interest; (b) a second sequence comprising a second promoter operably linked to a gene encoding a light-sensitive transcription factor, wherein the light-sensitive transcription factor is configured to induce expression of the gene of interest upon activation by light; and (c) a first selection marker; wherein either: i. the gene of interest, the first promoter, the second gene, the second promoter, and the first selection marker are integrated at a first locus, or ii. the gene of interest, the first promoter, and the first selection marker are integrated at the first locus, and the second gene, the second promoter, and a second selection marker are integrated at a second locus.

2. The host cell of claim 1, wherein the light-sensitive transcription factor comprises a LOV domain.

3. The host cell of claim 1 or 2, wherein the light-sensitive transcription factor is EL222.

4. The host cell of claim 3, wherein the first promoter is a constitutive promoter.

5. The host cell of claim 4, wherein the constitutive promoter is PC120.

6. The host cell of any of claims 1 to 5, wherein the host cell is a yeast cell.

7. The host cell of claim 6, wherein the yeast cell is a methylotrophic yeast cell.

8. The host cell of claim 7, wherein the methylotrophic yeast cell is Komagataella phaffii.

9. The host cell of claim 6, wherein the gene of interest encodes a protein that comprises a tag domain.

10. The host cell of any of claims 1 to 9, wherein the first selection marker is zeocin.Princeton- 101576 11. The host cell of claim 10, wherein the second selection marker is G418.

12. The host cell of any of claims 1 to 11, wherein the gene encoding a light-sensitive transcription factor has a predetermined copy number at the first locus or the second locus.

13. A system for optogenetic protein production, comprising: a plurality of host cells, each hose cell being a host cell of any of claims 1 to 12; and a light source configured to activate the light-sensitive transcription factor, wherein activation of the light-sensitive transcription factor induces expression of the gene of interest.

14. The system of claim 13, wherein the plurality of host cells includes a first host cell and a second host cell, wherein the first host cell and the second host cell have different copy numbers of the light-sensitive transcription factor.

15. The system of claim 14, wherein each host cell has a copy number of the light-sensitive transcription factor that is 1-10.

16. The system of claim 15, wherein each host cell has a copy number of the light-sensitive transcription factor that is 2-10.

17. A method for optogenetic protein production, comprising: providing a host cell of any of claims 1 to 12; culturing the host cell under conditions suitable for growth; and exposing the host cell to light to activate the light-sensitive transcription factor, wherein activation of the light-sensitive transcription factor induces expression of the gene of interest.

18. The method of claim 17, further comprising: selecting a light-sensitive transcription factor-containing parent strain based on an illuminating capability of a bioreactor system and protein production needs; cloning a plasmid and placing a gene of interest under control of a PC120promoter with a zeocin marker; and forming an optogenetically-controlled strain by transforming the light-sensitive transcription factor-containing parent strain with the plasmid.Princeton- 101576 19. The method of claim 18, further comprising determining the illuminating capability of a bioreactor system prior to selecting the light-sensitive transcription factor-containing parent strain.

20. A hybrid machine-learning-supported method for light-induced recombinant protein production, comprising: formulating differential equations to represent relevant system states, such as biomass, a protein of interest, glucose, and copy number, where the copy number is treated as a dynamic state and remains constant with a right-hand side of its differential equation set to zero; determining a target initial condition of the copy number; performing parameter estimation based on dynamic data from specific experiments conducted under predetermined light intensity regimes and light-sensitive transcription factor copy numbers, where performing parameter estimation includes using a trained multi-input, single-output Gaussian process to predict each parameter as a function of a light intensity regime and light-sensitive transcription factor copy number; and incorporating predicted means of the trained multi-input, single-output Gaussian process as model parameters into mechanistic kinetic functions.