Measurement

Agar plate stress test, horizontal and vertical

Our project mainly discusses whether an engineered Bacillus subtilis can supply the right plant protectant at the right moment and how much that protectant will help plant grow and further leads to better crop yield for farmers in real life application. To know whether a protectant truly has a rescue effect on a plant and whether timing matters, we built a plant measurement on Arabidopsis thaliana seedlings grown on ½ Murashige and Skoog (MS) agar (Murashige & Skoog, 1962). It needed to be a controlled and repeatable system where the stress dose, protectant types, and the timing could each be changed one at a time.

Salinity stress damages plants in readable ways. Exposure of Arabidopsis to NaCl alters stress-response gene expression (Cheng et al., 2013; Rolly et al., 2020) and, at the whole-seedling level, stunts root growth, bleaches leaves through loss of chlorophyll (chlorosis), and turns leaves purple through anthocyanin accumulation. Stress ethylene contributes to these effects as it inhibits root elongation (Růžička et al., 2007) and impairs photosynthesis (Ceusters & Van De Poel, 2018), which is the reason why the bacterial enzyme ACC deaminase, which breaks down the ethylene precursor ACC, is one of our protectants (Glick, 2014). With the damage showing up in root length differences, biomass and leaf pigment, a protectant that works should be able to lengthen roots, raise biomass, and preserve chlorophyll compared with a stressed and untreated plate.

We chose agar for three reasons. Firstly, every seedling receives identical nutrients and an exact stress dose, which is hard to guarantee in soil as it contains more uncontrollable variables such as ununiformed moisture levels, pH, and minerals. Secondly, early physiological changes appear within days of observation, allowing many conditions to be screened quickly. Thirdly, it establishes clear baselines before testing in complex soil systems for salinity concentrations that can easily demonstrate stressed environments, protectant concentrations that has obvious rescue effectiveness on plants, and protocol validations.

The experiments ran in three phases, ordered by how well each protectant was understood, commercially used, and researched so that each phase checked the system before the next relied on it. Phase 1 used trehalose, a disaccharide reported to protect Arabidopsis under salinity stress (Kim et al., 2025), to find a NaCl level that gives a readable signal and to confirm the platform can show a rescue. Phase 2 used a commercial peptide (CH Biotech bio peptide) as a positive control and to test application timing. Phase 3 used ACC deaminase, LEA, and BoPEP4 secreted by our engineered Bacillus subtilis. A heat-stress version of the assay, based on published thermopriming protocols (Lee et al., 2023; Auverlot et al., 2025), was designed for the same protectants.

What the platform is useful for

Seedlings on agar are one of the most widely used formats in plant biology and the platform described here applies beyond our project in several ways.

• Screening for salt-tolerance of plants. Arabidopsis grown on agar supplemented with NaCl is a standard way to test how genes, mutants, or treatments change salt tolerance (Cheng et al., 2013; Rolly et al., 2020). Our dose-response data tested salinity concentrations of 75, 100, 150 and 250 mM NaCl, read against the control group on the same plate. At 75 mM, fresh weight and total chlorophyll stayed near control levels. At 100 mM growth was clearly reduced and the rescue effect was still visible, which is what makes it the working concentration. At 150 mM growth was stunted to about the same degree as at 100 mM but with little rescue, total chlorophyll fell sharply, and anthocyanin accumulated enough to purple the leaves. At 250 mM close to every sample showed chlorosis and severe bleaching and none recovered. The boundaries are approximate percentages of control and not fixed thresholds, because each trial differed. The discovery allows future researchers to skip the pretests and begin in the 75 to 100 mM window where growth is reduced but rescue is still visible.
• Testing plant growth-promoting bacteria and biostimulants. Microbes that lower plant ethylene with ACC deaminase are used to promote plant growth under stress (Glick, 2014). Plate assays are a common way to test bacterial effects on Arabidopsis stress tolerance (Lee et al., 2023). Our design compares bacterial products at defined timings and with equal liquid controls. The standardized protocols allow a reusable screen for any protectant testing, so other teams and researchers can test their own molecule or microbe with the same readouts, plates, and analysis.
• Testing exogenous protective compounds. Sugars such as trehalose have been tested on Arabidopsis thaliana under salinity stress (Kim et al., 2025). Our trehalose sets show how to find the usable dose range of a candidate compound. Too high of concentrations might inhibit and harm plants as well because it triggers the toxic accumulation of an intermediate signaling metabolite, trehalose-6-phosphate (T6P), which disrupts normal carbon allocation and energy signaling, causing growth arrest such as stunted roots and undeveloped leaves (Schluepmann et al., 2004).
• Heat-stress assays. The same plates and readouts can be used with thermopriming protocols (Lee et al., 2023; Auverlot et al., 2025).
• Timing studies for delivery driven by sensors. The protectant is applied before, with, or after stress; therefore, the platform can test whether early delivery matters. This directly connects to the question behind AIoT-based stress prediction and sensing in the field.
• Teaching and iGEM labs. The equipment is inexpensive (plates, a camera in a light box, a balance, and free ImageJ software) and every step is described below in the procedural steps.
• Honest record of failure modes. Especially for contamination issues or failure to see effectiveness in protectant doses, so other teams can avoid repeating them as to also provide improvements and suggestions to protocols for higher quality readouts.

Hydroponic stress test

We transplanted Arabidopsis into a hydroponic box to evaluate the efficacy of protectants on plants. We use hydroponics to test the effectiveness because we can control the environmental factors precisely. First, the agar in the lid and the medium in the hydroponic box ensure a complete nutrient supply for the plants. We also control humidity and temperature in a constant range to fix the outer conditions that may affect the results of experiments.

Agar settles the dose and the timing on a flat surface; hydroponics is where the same protectants meet a root system standing in liquid, which is the condition a field application has to work in. The two platforms share their readouts, so a number taken here can be set beside a number taken on a plate. What hydroponics does not give is root length: the roots grow into the well and cannot be photographed against a scale, so the readouts are fresh weight, chlorophyll and the daily photographs.

Two vessels

Transparent tip box. After testing several prototypes of the hydroponic box, we settled on a 1000 µL tip box with a customized 3D printed board. The box is transparent, which lets us watch the plants and holds thirteen Arabidopsis seedlings. We used the tip box for the pretests of salinity and heat stress rescue by the CH peptide.

Tube rack. We first used the tube rack only to grow seedlings before an experiment began, then began running experiments in it directly. One rack holds forty-eight plants, enough to test ACCD and BoPep4 against both salinity and heat stress on a single rack. We moved to the rack because of time, and the plants in it grew at least as well as those in the tip box with less contamination. The build history of both vessels is on Plants; what follows is the measurement they carry.

What the platform is useful for

A reproducible hydroponic cultivation system. Hydroponics provides a stable environment for plant research by fixing nutrient availability, water purity and stress intensity. We tested two systems, the tip box and the tube rack, and record the materials, the assembly, the nutrient medium and the environmental conditions for both, so another team can rebuild either one.

A salinity-stress protocol. We took the salt concentration from Cortés-Jiménez et al. (2014). After several pretests, 100 mM NaCl is the concentration at which plants are damaged and can still be rescued by a protectant.

A heat-stress protocol. We followed the heat-stress design of Lee et al. (2023) and exposed plants at 37 °C and 42 °C. The groups at 42 °C did not survive the day, while the groups at 37 °C were rescued by the CH peptide. A team setting up a heat assay can start from 37 °C and treat 42 °C as past the point of recovery.

A working dilution range for a bacterial supernatant. We tested 1/40, 1/60, 1/80 and 1/100 dilutions of both protectants. The plants tolerate all four, and 1/40, 1/60 and 1/80 are the three the team reads as giving the best rescue, ACCD under heat and BoPep4 under both heat and salinity. The finding another team can take without qualification is the ceiling: the carrier medium, and not the protectant, sets the strongest dose a plant will tolerate.

A contamination finding. Plants grown in the tube rack were contaminated less often than plants in the tip box. Our working explanation is the transfer step: tip box plants were moved from the rack at four days old, and although a group would usually survive three to four days afterwards, most ended in severe contamination. Plants left in the rack never have their roots handled.

A record of four boxes that did not work. We troubleshot four hydroponic boxes before the fifth, a 1000 µL tip box with a 3D printed board and a 3D printed black cover to keep light off the medium. The tip box is less stable than the tube rack, and we still offer the fifth design as a prototype other teams can modify.

Finding a dilution the plants survive

TB at full strength kills the plants, so the first run was a pretest on TB medium carrying B. subtilis and no protectant, to separate the carrier from anything the protectant might do. We tested 1, 1/2, 1/4, 1/8 and 1/16, with 0 as the control, at pH 5.7 and pH 7. The optimum for Arabidopsis is 5.7 and ACCD works best between 8.0 and 8.5, which Arabidopsis will not tolerate, so 7 is the compromise we tested against.

Every TB concentration in the pretest was too strong for the plants to survive, including 1/16. That is a failed rescue window, and it is still the reading that set the next run: the damage falls off steadily with dilution, so the usable range had to be below 1/16.

The salinity run that followed tested ACCD at 1/10, 1/20, 1/40 and 1/60 against 0 ACCD, each with and without 100 mM NaCl, and carried a separate group of 1/20 TB alone at both salt levels. We had expected 1/20 to be dilute enough for the plants to live and 1/40 and 1/60 to be almost without effect.

The expectation was wrong in one place and right in the other. 1/20 TB was still too much for the plants, and ACCD at 1/10 and 1/20 was not dilute enough for them to survive; 1/40 and 1/60 were tolerable. We therefore used 1/40 TB as the strongest dose in the heat runs, and these two runs together narrowed the working range from undiluted TB to 1/40 through 1/100.

The two heat runs

The ACCD and BoPep4 heat tests were set up together and differ only in the protectant, so the two panels below can be read against each other. In both figures the blue rack is the heat control at 22 °C and the white rack is the heat treatment at 37 °C.

At 22 °C both racks carry full green rosettes at every dilution, which says the carrier at 1/40 and below is no longer doing the damage the pretest showed. At 37 °C every group is pale and stunted, the dosed wells along with the two controls, so neither protectant separates from its control on these photographs. The run is reported here as a protocol that worked and a rescue that did not appear; what the same plants gave as fresh weight and chlorophyll is on Plants.

What went wrong, and what we changed

Hydroponics was easier to control than a field and harder to control than we expected. Repeating the same trial does not reliably give the same result, even holding 37 °C and 100 mM NaCl after Cortés-Jiménez et al. (2014) and Lee et al. (2023). What the platform does establish is a working window: co-treatment is the timing under which both protectants rescued plants at 100 mM salinity, and dilutions of 1/80 and weaker are where the carrier stops interfering with the protectant's own effect. Those two findings set the arms of the Phase 2 runs.

Chlorophyll extraction

To assess the physiological response of Arabidopsis thaliana seedlings under stress, and evaluate protectant treatment rescue effectiveness in our project, chlorophyll extraction is introduced as a precise tool of measurement to strengthen the analysis of each experiment's results.

Chlorophyll is an essential pigment in the photosynthetic machinery of plants that plays a pivotal role in converting light energy into glucose. Ultimately, varying chlorophyll levels impact plant growth (Roberts, 2023). The metabolic pathways that govern the chlorophyll biosynthesis are complex, which abiotic stress such as salinity stress, drought, extreme temperature variations can substantially perturb these processes (Li et al., 2024).

As connected to our project, we mainly focus on two abiotic stress types: salinity and heat. Given this category, these abiotic stressors trigger the overproduction of reactive oxygen species, leading to lipid peroxidation and chlorosis. Hence, extracting the chlorophyll as a form of data readout and measurement allows quantification on the degree of chlorosis, and permits accurate evaluation of plant physiological response in our experiments that cannot be visually and directly assessed through bare eyes.

Results

Chlorophyll content maps out the data trend and visualizes seedlings' internal stress regulatory response among experiments.

Trehalose treatment at 0.25, 0.5 and 1.0 mM has an upward trajectory of rescue effectiveness combatting 100 mM NaCl. In phase 1 of trehalose and NaCl agar testing, we integrated chlorophyll content since conducting experiment set #4; through chlorophyll extraction, we were able to visualize the chlorophyll content among different groups. The relationship of trehalose dose between control and 100 mM NaCl conditions shows a conversely contrasting pattern: under baseline conditions, higher trehalose doses slightly depress baseline chlorophyll levels, whereas under 100 mM NaCl stress, they progressively rescue and enhance chlorophyll retention in a dose-dependent manner. Although 0.5 mM trehalose still outperforms other trehalose doses in terms of rescue effect, the overall trend demonstrates a progressive upward trajectory that indicates the trehalose' protective efficacy in mitigating salinity stress.

Read Figure 7 with the caution the agar section ends on: the values are per gram of fresh weight, and stressed seedlings weigh very little, so a group can read high on this chart while the plate photograph shows smaller plants. The heat pretest in Figure 8 is a single set and its control group is the only group without the peptide, so it sets the baseline; it does not test the peptide against a stressed, untreated plate.

The physical measurements of Arabidopsis thaliana

Alongside the chlorophyll extraction data collection, both fresh weight and root length data have been collected for selected experimental sets.

Fresh weight is most closely used to accompany chlorophyll extraction because the fresh weight of each experimental set is used to normalize the chlorophyll accordingly. On the other hand, primary length data was collected to observe the effect of salinity and respective protectant application on root behaviour. The expected behaviour of the primary roots was conducting halotropism, where roots pave their way through the agar as usual but specifically bending away from the salty patches to search for areas without osmotic stress (Zou et al., 2021). However, considering our entire agar plate was mixed with NaCl of different concentrations, the root length of respective agar plates was significantly affected and demonstrated intriguing differences between the application of NaCl and without.

What the procedure above fixes is the part that decides whether the numbers mean anything: the 120 mm plate width as the scale, the baseline the root is measured from, and the blotting step before the balance.

Results

Root length, heat stress with ACCD, trial 1. Averaged over the whole trial, unstressed Arabidopsis reached 116.4 mm of primary root against 43.5 mm under heat stress, so thermopriming at 37 °C costs the root system roughly two thirds of its length. Within the heat-stressed groups, ACCD applied before the stress gave the longest average root at 46.9 mm, and the co-treatment group performed worst.

Fresh weight, the same trial. Unstressed plants averaged 0.5069 g against 0.0237 g under heat stress, which is the water loss thermopriming causes stated as a number. Among the heat-stressed groups the post-treatment plants held the most water at 0.0318 g, with pre-treatment next at 0.0258 g; the post-treatment group also had three days of recovery under normal conditions, which the comparison cannot separate from the treatment itself.

Both readouts agree on the part that matters for the project: stressed plants given ACCD do better than stressed plants given none. The timing arms do not rank consistently from one experimental set to the next, so the pre-treatment advantage seen here is a repeated observation and not a settled result. It is the observation the delivery argument rests on, because a protectant that works best before the stress arrives is what makes stress prediction worth doing at all.

What went wrong, and what we changed

Plates leaning back in the light box shortened the roots. A plate resting against the back of the box is photographed at an angle, which foreshortens everything on it. We 3D printed a rack that holds a 12 to 13 cm square plate upright, so the plate is square to the camera and the 120 mm calibration in ImageJ is the distance it claims to be.

Reflections obscured the measurement. Turning off the room lights and covering the opening of the light box with black cloth was not enough, because the box's own lights reflect off the plate. Moving the rack within the box and changing its distance from the camera left a reflection only at the extreme edges, clear of the roots.

Roots grew over each other in the denser groups. In the control groups especially, the primary root could not be told from its neighbours. For the last two or three sets we lifted the plants onto a thin sheet of black foam, laid parallel; pulled from the agar they clump, and the primary root is the longest, so it can be followed. Those photographs were taken from directly overhead with a 15 cm ruler beside the plants for calibration.

Bacillus subtilis electroporation

To successfully engineer Bacillus subtilis to produce and secrete bioprotectants (such as ACC deaminase, LEA, and BoPEP4), foreign plasmid constructs carrying these genetic sequences must be efficiently introduced into the host cells.

Electroporation is the application of a high-voltage electrical pulse across the cell membrane which creates transient physical pores that permit plasmid DNA uptake. While standard transformation methods in Bacillus subtilis often suffer from strain-dependent variable efficiency, electroporation provides a more controllable, high-efficiency technique critical for expressing recombinant stress-mitigating proteins (Duckert et al., 2022).

In our project, the experimental protocol uses electrocompetent B. subtilis cells and evaluates transformation under different electroporation conditions. Because electroporation efficacy can be influenced by the electrical parameters used, different voltages were tested while maintaining the designated resistance and capacitance for each strain. Following electroporation, cells were allowed to recover for 3 hours before being plated on selective LB agar. Successful transformants were identified based on colony formation under antibiotic selection, and colony abundance was used as an indicator for comparing transformation efficiency between experimental conditions.

Efficiency in B. subtilis is governed by strain physiology as much as by the applied field, so optimization means tailoring both the cell preparation and the pulse. The two strains differ in the density at which they are harvested, the osmoprotectant buffer and the wash steps: 168 after the DTU-Denmark iGEM team (2015) and WB800 after Lu et al. (2012). Evaluating transformant recovery across a range of peak voltages identifies the field strengths that maximise plasmid uptake without lysing the cells.

The purpose of this experiment was therefore to evaluate bacterial growth following electroporation and identify voltage conditions that produced comparatively high numbers of colonies. The results provide a basis for selecting suitable electroporation conditions for future plasmid transformation and genetic engineering of B. subtilis.

What the measurement contributes

Strain-specific electrocompetent cell protocols. We wrote up separate preparation workflows for B. subtilis 168, after the DTU-Denmark iGEM team (2015), and for WB800, after Lu et al. (2012). They differ in the OD₆₀₀ at harvest, the washing buffer, the number of washes and the resuspension volume, and those differences are the reason a protocol written for one strain does not transfer to the other.

An optimal voltage for each strain and construct. Comparing voltages at a fixed resistance and capacitance identified the settings that gave the highest colony yields. For strain 168, 2.1 kV returned more colonies than 2.3 kV. For WB800 carrying the CcaSR construct, 2.3 kV gave the highest yield at 370 colonies, and 2.5 kV reduced it.

Evidence of where the voltage starts killing cells. Colony counts fall at the top of each range, which is what excessive field strength and cell lysis look like, and sets an upper limit for these strains. A team can use it to avoid spending cells and plasmid on a pulse that destroys both.

Several constructs through the same protocol. We introduced AcdS-based plasmids, both wild type and v2t1, and the CcaSR light-sensor circuit at over 10 kbp, into B. subtilis. The protocol holds across plasmid sizes and selection markers.

A documented, reproducible procedure. The steps below cover cell preparation, the pulse, three hours of recovery, selective plating and colony counting, which is the baseline for transforming any plasmid expressing a stress-mitigating protein.

Results

Plates were incubated under the same conditions, around 37 °C, and colony growth was assessed by visual observation and colony counting. The photographs below are representative of each strain, construct and voltage; the counted plates are tabulated after them.

One plate is shown per condition. Further repeat plates of the B. subtilis WB800 v2t2 conditions were taken, and the original plate lettering repeats, so the figures here are keyed to strain, construct and voltage instead of to those letters. The two strains are not close: strain 168 returns single colonies you can circle by hand at both voltages, while WB800 returns plates that have to be counted.

Read the table down the plating column before reading it down the colony column. The 434 colonies at 2.1 kV and the 16 at the same voltage are the same strain, construct and pulse plated two different ways, so the gap between them is the dilution and the antibiotic, not the electroporation. Only rows sharing a plating condition can be compared.

On that basis the CcaSR series is the clean one: one construct, one plate type, three voltages, giving 268 colonies at 2.1 kV, 370 at 2.3 kV and 233 at 2.5 kV. The curve has a maximum in the middle, which is what a field strength that helps up to a point and then lyses cells should produce, and 2.3 kV is the setting we take forward for WB800 with this construct. Strain 168 points the other way across its two settings, 21 colonies at 2.1 kV against 8 at 2.3 kV, and both counts are small enough that the useful conclusion is about the strain and not the voltage. The BoPep4 series was pulsed in a 0.1 cm cuvette, where the same voltage is twice the field strength, and there 1.1 kV returned 15 colonies against 6 at 1.5 kV. Every count is a single plate.

Western blot

To verify whether the engineered B. subtilis could successfully produce and secrete protein or peptide products, we performed a Western blot to detect the secreted protein via the His tag we added at the C-terminus.

Western blotting separates proteins by molecular weight using polyacrylamide gel electrophoresis, transfers them onto a PVDF membrane, and uses a specific antibody to visualize the target. Standard quantification assays return a total; they cannot tell our protein from the host's. The blot answers a narrower question that the total cannot: is a band present, and is it the size the construct predicts.

What the measurement contributes

Sample preparation from a bioreactor. We established a workflow for separating and concentrating protein from bioreactor culture supernatant, which arrives dilute and carrying medium components.

Verification of recombinant protein expression. Using an anti-His antibody against our ACCD construct, we confirmed the presence and the molecular weight of the expressed bioprotectant. We also recorded where the expressed protein ran larger than its amino acid sequence predicts, which fusion tags or incomplete denaturation can both cause, so a team reading an unexpected band has a documented case to compare against.

Semi-quantification across a run. We tracked ACCD across bioreactor sampling points, collecting liquid every two hours, to find the phase of peak production.

Results

Panel A is the detection this measurement exists to make. A single band appears in the supernatant lane at about 37 kDa, which is the size the secreted ACCD construct predicts, so the protein is both expressed and outside the cell. Panel B reads the His-tag purification: the 25 mM lane is clean, and the band appears from 10 mM onwards and is strongest at 250 mM. The two elution series were loaded in the order 25, 10, 150, 250 mM, which is not monotonic, so the comparison that can be made from this gel is which buffers carried protein, not how elution scales with imidazole.

The third panel of the same blot follows ACCD through a bioreactor run, sampling the lumen cell lysate, the lumen medium and the shell medium at 14 to 44 hours. That panel is a reactor result and is reported with the rest of the run on Results, where the band is in the cell lysate lanes and faint to absent in the two medium fractions.

Competitive His-tag ELISA

To evaluate whether our Bacillus subtilis successfully secretes bioprotectants, the amount of protein in culture samples must be accurately measured. Total protein assays cannot distinguish the target from host proteins, and gel-based methods are only semi-quantitative.

All three of our proteins carry a polyhistidine tag, so an immunoassay against the tag gives a quantitative and independent measure of expression. It is sensitive, and it processes many samples in parallel, which is what comparing strains, constructs and culture conditions needs.

ELISA uses antigen–antibody interactions to detect or quantify a molecule, and comes in four main formats. A direct ELISA uses an enzyme-labelled primary antibody to detect an antigen. An indirect ELISA uses an enzyme-labelled secondary antibody to detect a bound primary antibody, which is the usual way to detect antibodies. A sandwich ELISA captures an antigen between two antibodies that recognise different epitopes, giving high specificity at the cost of needing a compatible antibody pair for each target. A competitive ELISA measures an antigen by letting it compete with a labelled tracer for a limited number of antibody binding sites, so the signal falls as the target concentration rises.

We used the competitive format because our recombinant proteins share a His tag and can therefore be quantified by competition for one anti-His antibody, with no separate antibody pair per protein. His-tagged proteins in the sample compete with an alkaline phosphatase-labelled His tracer for the anti-His monoclonal antibody, which is itself captured on a goat anti-mouse IgG-coated plate. Bound tracer converts pNPP substrate to an absorbance signal read at 405 to 420 nm. Because the signal decreases as concentration rises, a standard curve converts absorbance back into recombinant protein concentration.

What the measurement contributes

The assay gives a concentration for ACC deaminase, LEA14 and BoPep4 in bacterial culture samples, which lets production be compared across experimental conditions. We ran it with a six-point standard curve, duplicate measurements and binding controls, so the assay's own performance can be judged from the same plate as the samples. Alongside the Western blot and the activity assays it separates two questions that are easy to confuse: how much protein is present, and how much of it is working. The approach transfers to any microbial system expressing His-tagged recombinant proteins.

Plate controls

Non-specific binding was 0.005 after blank correction, so almost nothing bound the plate without the antibody. Corrected maximum binding was 1.109, inside the 0.3 to 1.5 A.U. window the kit asks for, which is what makes every %B/B₀ below meaningful. The two B₀ wells read 1.311 and 1.181, an 11% spread on the one value every sample is divided by, and that spread propagates into every concentration on this plate.

The curve had to be refitted

The instrument's own analysis is not usable. The twelve standard wells were entered as twelve separate concentrations instead of six concentrations in duplicate, so the FlexA fitted a linear regression to mislabelled standards, returned R² = 0.181, and reported sample concentrations from −18.1 to +22.2, most of them negative.

We reanalyzed the same absorbances against the correct concentrations. An unconstrained four-parameter logistic fit did not converge usefully, because the measured standards span only 31.6 to 95.1% B/B₀ and leave both asymptotes undefined. Fixing the top and bottom at 100% and 0% gives the fit the concentrations below are read from: IC₅₀ = 1.21 µg/mL, slope 0.88, R² = 0.98.

%B/B₀ falls as concentration rises, which is the behaviour a competitive assay must show, and every duplicate agrees to within 6%. The limit is the span: the six standards cover 31.6 to 95.1% B/B₀, so any sample outside that band is extrapolated and the number attached to it is weaker than the curve's R² makes it look.

Results

The two LEA samples are the ones this plate measures properly: both dilutions sit inside the 20 to 80% band and agree to within 10%, giving about 5 µg/mL. The four ACCD wells all read below 20% B/B₀, which is off the bottom of the standards, so 13.5 to 19.1 µg/mL is an extrapolation and should be read as "higher than LEA and BoPep4" and not as a figure. One BoPep4 pair disagrees by 43% between its dilutions, which is the kit's own signal for interference and means that sample needs purifying and re-assaying.

Both sides peak at 16 hours and fall afterwards. That is a trend and not a result: each time point is one well with no technical replicate, and half the readings sit below the standard curve. The later anti-His runs on the October reactor batches, which do carry replicates, are reported on Results.

One thing this assay cannot do, whatever its precision: anti-His ELISA counts protein carrying a His tag. It says nothing about whether that protein folds or turns substrate over. Activity is what the Berthelot and DNPH assays in this section are for, and the two numbers should never be substituted for one another.

Berthelot assay for ACC deaminase

In our project, 1-aminocyclopropane-1-carboxylate (ACC) deaminase derived from Pseudomonas putida UW4 was selected as a bioprotectant to mitigate abiotic stress in plants.

Exposure to environmental stresses, such as drought, salinity, and extreme temperatures, triggers the accumulation of the stress hormone ethylene (Dubois et al., 2018), which accelerates leaf senescence and abscission, inhibits primary and lateral root elongation (Růžička et al., 2007), induces programmed cell death (Drew et al., 2000), and impairs photosynthesis (Ceusters & Van De Poel, 2018), ultimately precipitating rapid plant decline. Because ACC serves as the precursor to ethylene biosynthesis, ACC deaminase produced by rhizosphere microorganisms degrades ACC into α-ketobutyrate and ammonia. This enzymatic cleavage depletes rhizosphere ACC levels, establishing a concentration gradient that drives root exudation of ACC and thereby reduces endogenous ethylene synthesis (Glick, 2014). To exploit this mechanism, Bacillus subtilis was engineered to secrete recombinant ACC deaminase and cultured in a bioreactor for mass production. To assess bioreactor yield and confirm the enzymatic activity of the secreted protein, we evaluated the degradation efficiency of ACC in crude bioreactor extracts using two biochemical assays targeting ACC degraded products: a 2,4-dinitrophenylhydrazine (DNPH) assay to quantify α-ketobutyrate accumulation (Penrose & Glick, 2003), and a Berthelot (indophenol blue) assay to determine ammonia concentrations.

The Berthelot readout applies to the assays in this section. The reactor permeate assays of 7 and 8 October were read by the DNPH method as α-ketobutyrate (Digital Twin). Due to the unavailability of 2,4-dinitrophenylhydrazine (DNPH), the catalytic activity of the recombinant ACC deaminase was determined using the Berthelot assay, targeting the released ammonium (NH₄⁺/NH₃) as the analytical readout. The Berthelot reaction is based on the reaction between ammonia and hypochlorite (OCl⁻) to generate monochloramine, which subsequently reacts with sodium salicylate in the presence of sodium nitroprusside as a catalyst to produce an indophenol dye quantifiable by absorbance at 660 nm (A660) (Krom, 1980). However, several interfering factors were encountered under our specific assay conditions. First, rich medium components, notably yeast extract and free amino acids present in the Terrific Broth (TB) medium, exhibit non-specific chromogenic interference in the Berthelot reaction (Ngo et al., 1982). Furthermore, unreacted ACC substrate also reacts with the reagents to form an indophenol-like colored complex. To accurately quantify the ammonium released by enzymatic ACC cleavage, two preliminary optimization steps were undertaken: determining an optimal dilution factor for the TB medium or bioreactor crude extract to suppress matrix background signals sufficiently below the target analyte response, and constructing a substrate-compensated calibration curve in which ACC and NH₄⁺ were combined at complementary molar ratios summing to the fixed initial substrate concentration, allowing A660 to be calibrated directly against fractional ACC conversion.

What the assay is useful for

The Berthelot assay is predominantly employed in water quality and environmental monitoring, encompassing the determination of ammoniacal nitrogen in surface water, seawater, and industrial wastewater, as well as the quantification of Kjeldahl nitrogen in digested samples (Searle, 1984). In clinical diagnostics, coupling urease activity with ammonia production has established this reaction as a classical method for assessing blood urea nitrogen.

Soil and agricultural science represent another domain historically reliant on this methodology. Key applications include quantification of soil inorganic nitrogen pools, wherein exchangeable ammonium ions are extracted with 1 to 2 M KCl and the supernatant directly subjected to indophenol chromogenesis, a standardized procedure for evaluating available nitrogen and soil fertility; total nitrogen analysis of plant and soil digests, where the sodium salicylate-based indophenol modification is compatible with both 2 M KCl soil extracts and acid digests, offering superior sensitivity over legacy phenol formulations while substantially mitigating operational hazards (Kempers & Zweers, 1986); measurement of net nitrogen mineralization and nitrification rates, fundamentally indexing the soil nitrogen cycle by tracking net shifts in ammoniacal nitrogen across incubation periods; determination of soil urease activity, quantifying ammonium liberated from soil incubated with urea substrate via modified indophenol chromometry, an approach widely adopted in soil enzyme profiling (Kandeler & Gerber, 1988); and evaluation of fertilizer efficacy and urease inhibitors, monitoring urea hydrolysis kinetics to assess the capacity of inhibitors to delay ammonia release and curb volatilization losses. Additionally, in screening plant growth-promoting rhizobacteria and biostimulants, microbial ammonia production itself serves as a hallmark trait of plant-beneficial capacity.

Within biochemistry and enzymology, the Berthelot assay acts as a universal coupled detection system for any enzymatic reaction releasing free ammonia. Consequently, it is broadly applicable for profiling the catalytic activities of ureases, glutaminases, asparaginases, and diverse amino acid or nucleoside deaminases (Searle, 1984).

In our application, ACC deaminase catalyzes the cleavage of ACC to liberate ammonium, which theoretically permits activity assays via the Berthelot method. In practice, however, severe analytical interferences arising from both the rich culture medium and the substrate itself preclude direct quantification. Here, we address and systematically resolve these matrix and substrate interferences, establishing the Berthelot assay as a viable analytical alternative to monitor ACC deaminase activity when DNPH is unavailable.

Results

ACC forms an indophenol-like colored complex in the Berthelot assay. Although chromogenic interference by ACC in the Berthelot reaction has not been previously reported, ACC contains a primary amine moiety that is theoretically susceptible to indophenol dye formation. To evaluate this interference, NH₄⁺ standard calibration curves were generated across varying background concentrations of ACC. As demonstrated below, robust quantification of NH₄⁺ was achieved even against a 2 mM ACC background. Therefore, despite the intrinsic chromogenic contribution of ACC in the Berthelot assay, accurate quantification of ACC deaminase catalytic activity via NH₄⁺ detection remains feasible, provided that the initial ACC substrate concentration is strictly controlled.

The 20 mM series is the result that sets the working limit. Its slope is negative, so at that substrate load the dye is no longer reporting ammonium at all and no conversion number taken there would mean anything. Everything this assay returns therefore has to sit at or below the 2 mM background, which is what the compensated standard above is for. The mechanism figure from Yu et al. (2021) is not republished here; the reaction is in the reference.

DNPH assay for ACC deaminase activity

To determine whether engineered Bacillus subtilis WB800 can produce and secrete active ACC deaminase, we measure the enzymatic cleavage of ACC into α-ketobutyrate and ammonia. The Berthelot assay detects the ammonium; the 2,4-dinitrophenylhydrazine (DNPH) assay reads the other product directly.

Under acidic conditions DNPH reacts with the carbonyl group of α-ketobutyrate to form a hydrazone, which is made visible with alkali and quantified at 540 nm (Penrose & Glick, 2003; Kozaeva et al., 2022). The method answers the three objections to reading ammonium in our samples: the ACC substrate does not interfere, the background in a rich culture medium is far lower, and the assay is nineteen times more sensitive. That sensitivity is what makes initial velocities measurable and so puts true kinetic parameters, K_m and k_cat, within reach. Paired with growth and expression readouts, it tells us whether the enzyme leaving the cell is catalytically intact and not merely present.

What the measurement contributes

We scaled the conventional protocol of Penrose and Glick (2003) down twenty-fold while holding the reaction proportions, which moves the assay onto a 96-well microplate and makes screening a strain library practical. With it we compared ACC deaminase activity across constructs, tested what pyridoxal 5'-phosphate supplementation does to that activity, and checked that recombinant ACCD secreted by our engineered B. subtilis still functions outside the cell. We also set up the standard curve and the arithmetic that converts an absorbance into an α-ketobutyrate concentration, a total product yield and an activity, so the comparison between strains rests on a stated calculation.

PLP is what makes the enzyme work

Without added PLP, the Csn construct reads 0.499 mM against the WB800 host's 0.513 mM: the two are the same number, so whatever ACCD is present is turning over no faster than the host background. Adding the cofactor moves Csn to +13.7 nmol over that baseline, which is the clearest statement this page can make that ACCD does not function without PLP, and it is why PLP is dosed alongside ACCD in every plant run. SamyQ with PLP reaches 89.7 nmol, +30.0 over baseline, which is 2.19 times the net product of Csn with PLP, so the two secretion tags are not equivalent. Every value here is one well on one plate, so the ranking is a first result and not a measured fold difference.

Screening the strain library

In the O group, O3 is the strongest construct on the plate after L24 at 0.550 mM, while O6 falls to 0.141 mM, near the floor. In the L group, L24 is the highest value anywhere in the screen at 0.595 mM and L1 follows at 0.393 mM; L16 sits in between at 0.250 mM. The S group accumulates less than either, with S40 its best at 0.280 mM and S1, S16 and S24 all close to the floor.

Three of the twelve wells carry a negative net absorbance, and L40 converts to a negative concentration of −0.133 mM. A negative α-ketobutyrate concentration is not a measurement; it means the sample absorbed less than its own blank, which substrate depletion, cofactor loss or a degraded colour reaction can each produce. Those wells mark the bottom of what this plate can resolve, and the floor sits somewhere near 0.15 mM rather than at zero. The whole screen is single wells at one time point, so what it supports is picking constructs to repeat, with L24 and O3 first.

This method is also what read the reactor permeate on 7 and 8 October, where it gave the only activity figure the project has for material coming out of the bioreactor, 7.25 mU/mL. Those readings are reported under the runs on Digital Twin and carried into the validation table on Development; like the plates here, each is a single sample.

Dual light path inline photometer

We built an in-line flow-cell photometer that reads OD600 directly from our hollow-fiber bioreactor without ever needing to draw a sample.

OD600 is the field standard for bacterial density but is almost always measured through sampling, which not only interrupts the run it describes but is also a manual action that creates the opportunity for contamination or accidental measured value error, especially when placed in public use for small-scale farmers. So we moved the measurement into the loop itself. A light beam generated by an amber LED, whose peak lies close to 600 nm, passes through a 90/10 beam splitter that diverts part of the path to a reference sensor before the rest crosses our flow cuvette and reaches the main sensor.

A single sensor would be enough if the light source never changed, and over a multi-week run it does. LED output drifts with temperature, current and age, and one sensor cannot tell a dimmer lamp from a denser culture, because both make the beam weaker. The reference sensor reads the source on a path that never touches the medium, so dividing one by the other cancels whatever the source did and leaves what the sample did. That, together with burst sampling and a MAD filter for bubble rejection, is what lets a roughly $150 build compete with professional OD600 instruments such as the BioDrop at around $15,000.

What another team can take from it

Our central contribution is a validated, calibration-free in-line OD600 measurement device. Our raw output already agrees with a $15,000 reference instrument, so another team can rebuild and use it directly. A device that requires the calibration against a benchtop spectrophotometer doesn't remove the pricey spectrophotometer from the workflow, while ours cuts that dependency. 4 main pieces of this photometer can be taken to be learned and reused by others.

• In-line, non-destructive geometry. Our flow cuvette sits in the circulating loop, outside of our tubing. No volume of medium needs to be removed, nor does anyone need to adjust or operate it once it's set up, and the tilt of the photometer ensures almost all bubbles that enter leave within a quarter of a second.
• Reference sensor normalisation. The use of a 90/10 beam splitter sends 10% of our light into a sensor that avoids the path of our medium sample. This allows us to set a calibration for our main sensor in the case of source-intensity variability. Some possible instances could be from inconsistent current (our PN2222 transistor mostly accounts for that) or LED degradation. Over a multi-week run though, precautions have to be set in place.
• Bubble-rejection filter. Each point of measurement from our photometer is determined by a burst of 16 data points. These data points are condensed by median and median absolute deviation (MAD), with a fence at median ± 3×MAD. By using the median twice, as our center and as our spread, we let large bubbles inflate its own fence and pass through as an outlier. The live rejection rate is reported to the user and when flagged as above 50%, triggers a hardware alarm that indicates inaccurate photometer data.
• Optogenetic-compatible wavelength measurement. 600 nm wavelength LEDs keep us on the standard for measuring turbidity, while simultaneously sitting right between the wavelengths of green (~535 nm) and red (~672 nm), the 2 actuation peaks for CcaS/CcaR. This ensures that our photometer measurements don't affect the activation or deactivation of our optogenetic circuit in B. subtilis.

A fifth piece was added on 14 October, after the bioreactor run of 8 to 10 October showed what the first four do not cover. Bubbles and a drifting reference can move the reading while the culture stands still: across that run a sixth of the windows carried more than 20% bubbles, and the reference beam, which never passes through the medium, moved by a factor of 4.8. The acceptance gate now checks each window against the last one accepted; a window that fails holds the previous value and is flagged instead of being published. On the 14 to 16 October run it turned away 31 of 1,083 windows, and the reference held within 17%. The gate also marks the top of the instrument's range: above an optical density of about 2.7 the sample beam falls to a tenth of the reference, and what the photometer reports above that is flagged and not used (Digital Twin).

Results against the reference instrument

The raw, uncalibrated photometer output agrees with the BioDrop to a median paired difference of 3.8% and tracks the TiO₂ mass standard at least as closely as the reference instrument (R² = 0.9999 photometer against R² = 0.9998 BioDrop).

We ran two different iterations of the test against the reference instrument and both agree with the reference independently of each other. We had to exclude two singular points on objective grounds: both were BioDrop readings. We likely had a handling error in the first test and had a monotonicity violation in the second, implying a genuine mistake. Removing both didn't change our photometer readings at all and only made the BioDrop readings more consistent to the other data points.

We didn't calibrate to our BioDrop because the runs showed we didn't need to. Against the mass standard, the BioDrop is slightly less linear than our own photometer, so fitting it would tie our readings to the instrument's own errors when we're already in agreement within a few percent. Therefore, we treat calibration as unnecessary.

That is the TiO₂ comparison. The SiO₂ cross-validation above ran on 10 October 2026, in 2.5 mg steps from 2.5 to 100 mg in 500 mL of 1% SDS. Against the reference the dual-path build reads one to one to OD 2.5 (slope 0.995, intercept −0.07). The numbers, and the comparison with the simplified build, are on the photometer page and under R1 on Digital Twin.

Transmembrane pressure and flowrate meter

Flow and transmembrane pressure were measured together, alongside our pump and specific dials and outputs.

These two numbers decide whether the hollow fiber membrane can run our culture at all. Flow rate tells us how hard the fibers are sweeping the cells; transmembrane pressure tells us whether the protectant the reactor produces can be pushed through the membrane.

Transmembrane pressure is the mean lumen pressure minus the shell, or permeate, pressure, and it has to be positive for anything to pass. A positive TMP drives fluid from the lumen through the fiber wall to the shell side, and permeate flux rises with it, so a TMP at or below zero means nothing leaves with the permeate (Millipore Corporation, 2003). Past a certain point flux stops rising and the fouling layer on the membrane is compressed instead, which only adds resistance (Millipore Corporation, 2003). We therefore read TMP at each pump setting, looking for dials that keep it positive without pushing it into that plateau.

Flow through the fiber lumens sets the wall shear that sweeps the membrane. For laminar flow in a lumen the wall shear rate is

γ̇_w = 4(Q/n) / (πr³)

where Q is the total flow rate, n the number of fibers and r the lumen radius, so shear is fixed by the fiber inner diameter, the fiber count and the flow (Yuen, 2025). Our module has 11 fibers of 1 mm inner diameter, giving r = 0.5 mm. Across the flow meter readings for dials 10 to 24, about 113 to 450 mL/min, the calculated wall shear rate is about 1,750 to 6,950 s⁻¹.

That range is a problem, and it is the reason this measurement exists. In a Couette flow bioreactor B. subtilis grew at shear rates up to 1,482 s⁻¹, but the cells shrank to about half their size and intracellular reactive oxygen species rose 9.3-fold (Sahoo et al., 2003). We therefore treat roughly 1,500 s⁻¹, about 1.5 Pa in a water-like medium, as the upper limit for lumen wall shear, and our whole operating range sits above it. Culture viability after circulation through the module has to be checked before any dial is settled on. At the low end the concern reverses: B. subtilis biofilms and streamers formed at every wall shear stress tested from 0.068 to 0.67 Pa, and biofilm growth slowed as shear rose (Wittig et al., 2025). We found no lower survival limit for planktonic cells. Both studies used a bioreactor and a flow channel, not a hollow-fiber module, so the limits are transferred and not measured here.

Taken together, the two readings choose a pump setting: the flow must keep lumen shear below the survival limit and high enough to slow fouling, and TMP at that flow must stay positive. Both also feed the scale-up question. Larger vessels mix less well, and poor mixing leaves gradients in dissolved gases, pH, substrate and shear that cells meet as they circulate (Lara et al., 2006); mass and heat transfer coefficients both depend on power input and mixing time, and the formulas used to scale them are empirical (Villadsen, 2015). In our system TMP drives the protectant across the membrane, the flow rate sets how well material reaches the fiber wall and how much is swept away from it, and the circulating flow carries heat between the culture and the rest of the loop. Fouling, which shows up as a rising TMP, would reduce mass transfer over a long run.

ReLEAF has a third transfer to keep in mind, because the system is driven by light. As the culture scales the light path lengthens and the cell density rises, so less light crosses the vessel and it eventually becomes too opaque for the light to reach every cell or to pass through our photometer. Optogenetic circuits have given homogeneous induction in bioreactors of at least 5 L, with illumination limited above OD600 40 (Zhao et al., 2021). Flow rate also sets how fast cells move through the lit region and how well they mix within it.

Results

The 16 August series is the one to build on. Taking the background away leaves the module's own contribution, which rises from 0.04 bar at dial 10 to 0.21 bar at dial 24, and the two flow columns agree within a few per cent at every setting, which is what the Hall-effect meter was brought in to fix.

Note what the figure is and is not. Three transducers were fitted, at the feed, the retentate and the permeate ports, so all three pressures were available. The quantity recorded in these tables was computed as (P_feed − P_retentate) / 2, which is a pressure drop along the lumen and not the transmembrane pressure defined in the introduction. TMP needs the permeate reading: the mean of feed and retentate, minus the shell pressure. The permeate channel is not in these tables, so no transmembrane pressure has yet been computed from this rig, and whether TMP stays positive at a working dial is still open.

Hollow fiber membrane breach detector

The hollow fiber is where the culture and harvest stream touch, and a 0.2 µm pore is what keeps B. subtilis on the lumen side. If breached, the cells would be able to come through, so a detector is installed on the shell side to watch for signs that may indicate unwelcomed particles.

Fitted, but no breach test is on record. It currently rests on an offline proxy that was run once. During version 1, 2 mL was drawn from the shell side and spread on LB agar, and no colonies were recovered, so no culturable B. subtilis reached the shell side in that run. It is a single plating, and the replicate count, incubation time and incubation temperature were not recorded.

No detectable B. subtilis colonies were observed in the shell-side sample under the tested conditions. This provides preliminary evidence that the hollow-fiber membrane retained bacterial cells during the tested run.

What this plate can carry is a containment claim for one run; it cannot stand in for the detector, because a plate takes a day to answer and an instrument has to answer while the reactor is running. The build is on Hardware → Bioreactor and the containment argument it belongs to is on Safety.