Engineered gamma radiation phytosensors for environmental monitoring

Summary Nuclear energy, already a practical solution for supplying energy on a scale similar to fossil fuels, will likely increase its footprint over the next several decades to meet current climate goals. Gamma radiation is produced during fission in existing nuclear reactors and thus the need to detect leakage from nuclear plants, and effects of such leakage on ecosystems will likely also increase. At present, gamma radiation is detected using mechanical sensors that have several drawbacks, including: (i) limited availability; (ii) reliance on power supply; and (iii) requirement of human presence in dangerous areas. To overcome these limitations, we have developed a plant biosensor (phytosensor) to detect low‐dose ionizing radiation. The system utilizes synthetic biology to engineer a dosimetric switch into potato utilizing the plant's native DNA damage response (DDR) machinery to produce a fluorescent output. In this work, the radiation phytosensor was shown to respond to a wide range of gamma radiation exposure (10–80 Grey) producing a reporter signal that was detectable at >3 m. Further, a pressure test of the top radiation phytosensor in a complex mesocosm demonstrated full function of the system in a ‘real world’ scenario.


Introduction
Globally there is renewed interest in adopting nuclear power as a cleaner alternative to coal power plants to meet increasingly aggressive climate goals. Despite the success of many nuclear reactors, the public still remains wary of failures, such as Chernobyl in 1986 and Fukushima Daiichi in 2011, and the environmental impact of these incidents (Medvedev, 1990;Povinec et al., 2013). To assuage public concerns and increase safety, there is a need for new technologies to monitor for radiation contamination. For highly penetrative ionizing radiation such as gamma radiation, clear risk to human health appears above doses of 0.1 Gy (acute) or 0.3 Gy (chronic), with the LD 50/30 for humans at approximately 4 Gy and certain death at ≥10 Gy acute exposure (Metting, 2017). Current mechanical sensors are not feasible for long-term environmental monitoring of ionizing radiation due to costs associated with maintenance and operation. Further, mechanical sensors do not provide an accurate measure of the biological impact of low doses of exposure over an extended period of time. With the advent of synthetic biology, plant biosensors (phytosensors) are emerging as a feasible option to detect and report the presence of environmental disturbances (Volkov and Markin, 2012). Phytosensors are uniquely tuned to their environment (soil, water and air), and the reporters produced by these biosensors directly reflect a biological impact on the surrounding ecosystem.
Plants have a history of use as radiation phytosensors. After Chernobyl, pine tree forests were used as a visual indicator of radioactivity with the phenotypes observed correlating dosimetrically to both radiation dose received and the radiation emittance rate (Yoschenko et al., 2018). Compared to mammals, plants have a much higher radiotolerance, allowing them to persist and monitor exposures much higher than their animal counterparts (Real et al., 2004). A key component of plants' radiotolerance is their native DNA damage response (DDR) pathway. When organisms are exposed to gamma radiation there is a rapid increase in the reactive oxygen species (ROS), which produce numerous single-and double-strand breaks. These DNA breaks are repaired through a complex network of genes that coordinate the repair and maintain the genome integrity. As with nearly all eukaryotes, in Arabidopsis DNA breaks are sensed primarily by ataxia telangiectasia-mutated (ATM) and ataxia telangiectasia-mutated and Rad3-related (ATR) protein kinases as part of protein complexes at the lesion site (Kimura and Sakaguchi, 2006;Nisa et al., 2019). ATM and ATR then catalyse the phosphorylation of many protein targets to induce cell cycle arrest and initiate DNA repair (Roitinger et al., 2015). Subsequent phosphorylation of a key plant transcription factor, Suppressor of Gamma Response 1 (SOG1), activates the protein enabling binding to a specific DNA motif [CTT(N)7AAG] (Ogita et al., 2018;Yoshiyama et al., 2009) and activation of many downstream DNA repair genes, such as RAD51 (Ogita et al., 2018;Yoshiyama et al., 2017). RAD51 encodes a recombinase that is essential for homologous recombination and repair of DNA strand breaks. In other eukaryotes, RAD51 forms a homodimer and upon activation binds to double-strand breaks and facilitates strand exchange (Sung and Robberson, 1995). In previous work, a promoter consisting of a tetrameric repeat of the SOG1 binding site was used to express a reporter gene in response to a genotoxic stress (Ogita et al., 2018), demonstrating the feasibility of such an approach to sense-and-report gamma radiation.
The overall objective of this work was to develop a fully functional phytosensor for sensing and reporting the presence of gamma radiation in the environment. Potato (Solanum tuberosum) was chosen as the chassis organism due to its tetraploid genome and its role as a true crop plant. Additionally, potato's nuclear genome can be effectively engineered. The mEmerald green fluorescent protein was chosen as the reporter molecule due to the ability to easily detect the signal at a standoff distance of ≥3-m with a low-cost imaging system. The radiationresponsive promoters of the genes PCNA, UVH1, RAD51 and a synthetic 4xRAD51 promoter were selected for initial testing. Utilizing a design-build-test strategy employed in the development of other phytosensors (Liu et al., 2013;Persad et al., 2020), these components were engineered to meet the requirement of a fully functional gamma radiation phytosensor.

Design and assembly of gamma radiation sensing systems
In order to determine the radiotolerance of potato as a potential phytosensor, an initial dosing experiment was conducted to determine the phenotypic response of four-week-old plants across a range of doses from 0 to 250 Gy ( Figure 1a). All dosing distances and times are indicated in Table S1. After irradiation, the phenotypes of the treated plants were assessed, first at anthesis and then at senescence (Figures S1, S2). At anthesis, plants treated with ≥20 Gy had impaired apical growth and showed a significant growth delay ( Figure S1). Plants treated with 80 Gy had severely reduced plant height which corresponded with an increase in leaf fresh and dry weight per unit area, and chlorophyll content (Figure S1B-E). At the end of the plants' life cycle, potato treated with 20-40 Gy recovered to a similar height and dry biomass production to that of untreated plants by producing lateral stems, with 40 Gy treated plants having the greatest overall stem length and stem biomass ( Figures S2C-F). This translated to a significantly lower stem density compared to 0-20 Gy treated plants ( Figure S2G). Starting at ≥20 Gy a progressive reduction of tuber yield was also observed ( Figure S2H-J). Plants treated with 80 Gy showed a severe phenotype throughout their entire life cycle and did not reach maturity, though they did produce small tubers ( Figure S2). Above 80 Gy the phenotype was so severe that no growth was observed, and thus doses above this range were excluded from further analysis.
After identifying an effective range of doses for which potato could sense-and-report gamma radiation, genes that were strongly upregulated across the selected dose range (0-80 Gy) were identified and mined their promoter sequences. The initial gene targets were the native potato genes StPCNA, StUVH1 and StUVR7. qRT-PCR was performed using leaf samples collected before treatment, along with 1.5, 9 and 24 h after treatment to determine if the genes were upregulated, and the time course of expression. Among the native potato DDR genes analysed, StPCNA expression was the most responsive to gamma radiation, showing a significant induction at ≥5 Gy, while StUVH1 only had a significant increase at 80 Gy. StUVR7 showed no response at any of the doses tested, and thus was excluded from further testing ( Figure 1b). In addition to being the most sensitive, StPCNA was rapidly induced (1.5 h) up to 26.9-fold above basal levels at >40 Gy, and 10-fold at 10 Gy. StUVH1 expression achieved only a 2.2-fold increase above baseline at doses >40 Gy, and no significant induction below 40 Gy at any timepoint tested. While the magnitude of StPCNA expression was high during the first 9 h tested, the expression level fell significantly after 24 h, indicating a burst response to the single dosing scheme. Based on this data, it was hypothesized that the promoters of StPCNA and StUVH1 could be used to develop sensors for gamma radiation.
Promoters for StPCNA and StUVH1 were mined as potential candidates to use as sensors for gamma radiation. Plant promoters are notorious for their variability in length, with many plant promoters being in excess of 1000 bp (Ali and Kim, 2019). Since these genes' promoters have not been characterized, the approximate regions of -1800 to +18 for StPCNA and StUVH1 were extracted from the potato genome and domesticated for use in GoldenGate cloning. This length was selected in order to encompass cis elements within the promoter regions of these genes while not adding additional cloning difficulties due to very long, repetitive sequences. The StPCNA and StUVH1 native promoters along with previously characterized AtSOG1 DNA binding motif [CTT(N) 7 AAG] from the AtRAD51 promoter (Ogita et al., 2018;Yoshiyama et al., 2009) were used to design promoter switches for gamma radiation sensing. Transgene expression cassettes containing the promoter-5'UTR region of either StPCNA, StUVH1 or AtRAD51 (Table S2) were fused to the mEmerald fluorescent reporter gene along with a viral Cowpea Mosaic Virus 3 0 UTR (3CPMV-nos), producing the pPCNA, pUVH1 and pRAD51 plant transformation vectors (Figure 1c), respectively. Additionally, a synthetic promoter including four repeats (4x) of the AtSOG1 binding motif [5 0 -CGAGACTTGTTGAA-GAAGGCCTTT -3 0 ] from AtRAD51's promoter fused to a minimal 35S promoter-Tobacco Mosaic Virus Ω leader was also designed (Table S2). A mEmerald transgene cassette including this 4xRAD51 promoter and the 3CPMV-nos 3 0 UTR was used to assemble the p4xRAD51 transformation vector (Figure 1c). Both native and synthetic promoter constructs included both a visual and selectable marker for selection of transgenic events ( Figure S3). Once transformed into potato, these phytosensor constructs can be evaluated for radiation inducibility as well as fluorescence at a 3-m standoff.

Testing of gamma radiation sense-and-report
Transgenic potato events were created using each of the senseand-report cassettes described above (StPCNA pro , StUVH1 pro , AtRAD51 pro and 4xRAD51 pro ). Three independent transgenic events (events 1-3) were selected for each construct, and events were genotyped using Southern blot to select representatives with both single and multiple copies of the transgene (Table S3, Figure S4). In addition to determining the number of inserts, the baseline expression of the reporter gene, mEmerald, was measured relative to the housekeeping gene StEF1α (Table S3, Figure S4). Not surprisingly, 4xRAD51 pro event 1 had the highest baseline expression of >2.6-fold relative to the housekeeping gene, since this event contained four copies of the 4xRAD51 pro :: mEmerald transgene inserted into the plant genome. For initial testing, the downselected events were challenged as four-weekold plantlets in tissue culture at the highest dose, 80 Gy, to determine if the phytosensor architecture functionally expressed mEmerald after insult by gamma radiation. Based on qRT-PCR analysis, all events for 4xRAD51 pro and StPCNA pro showed  (Figure 2a). The largest fold induction was observed for the 4xRAD51 pro events 1-3 with significant (P < 0.05) induction of 11.80-fold, 5.35-fold and 9.89-fold, respectively. To a lesser degree, StPCNA pro events 1-3, were significantly (P < 0.05) upregulated from 1.74-2.09-fold. Only two of the three AtRAD51 pro events (2 & 3) showed significant fold induction, while the StUVH1 pro events showed both increased and decreased expression depending on the event ( Figure 2a). As the goal was to develop a fully functional phytosensor capable of reporting at standoff, it was necessary to determine if the mEmerald signal produced in response to the radiation treatment was detectable at >3 m. All 4xRAD51 pro events were able to be detected by the fluorescence-inducing laser projector (FILP) (Rigoulot et al., 2021) system, demonstrating the potential of these events as functional radiation phytosensors (Figure 2b). AtRAD51 pro , StUVH1 pro and StPCNA pro events did not show a clear increase in fluorescence after treatment, and thus were excluded from further testing (Figure 2b). Based on the initial testing at the maximum dose of 80 Gy, the 4xRAD51 pro events met the criteria for a radiation phytosensor. However, an 80 Gy dose is a relatively high dose for monitoring scenarios, therefore these 4xRAD51 pro phytosensor events were selected for further testing at a variety of doses and as whole plants in pots.
For in-depth characterization of the 4xRAD51 pro phytosensing events, four-week-old plants in soilless media were treated with 0, 5, 10 and 40 Gy of gamma radiation. Both spectrofluorometer and FILP image data confirmed high fluorescent reporter accumulation after radiation treatments in all transgenic events (Figure 3a,b). Based on measurement of the mEmerald reporter, peak induction of events 1 (6.3-fold) and 2 (4.9-fold) were reached at 48 h, while the peak of event 3 (3.6-fold) was Expression time course of three putative radiation-inducible genes in four-week-old wild-type potato treated (5-80 Gy) and non-treated with gamma radiation. Samples were taken before treatment (0 h), then 1.5, 9 and 24-h posttreatment ( 60 Co, black arrow). Graphs represent RT-qPCR relative expression data (2 -ΔΔCT ) of the indicated radiation inducible genes (StPCNA, StUVH1 and StUVR7) vs the endogenous reference gene, StEF1α. Data are expressed as mean AE standard error (SE) of 4 biological and three technical replicates. Data were analysed using ANOVA (P < 0.05) and comparisons to 0 Gy were evaluated using Post-Hoc Dunnett's T3 test (P < 0.05). Statistical significance is indicated by '***' while non-significance with 'n.s.' (c) Design of radiation phytosensor constructs. The endogenous promoter-5 0 UTR sequence of StPCNA and StUVHI, and AtRAD51 plus each gene's first six codons along with the synthetic 4xRAD51-min35S-TMVΩ promoter-5 0 TMV Ω UTR were used to design phytosensors. All phytosensor constructs contain the mEmerald reporter gene (green fluorescent protein variant) and the Cowpea Mosaic Virus (CPMV)-nos 3 0 UTR and terminator. observed at 72 h. Not surprisingly the 40 Gy dose had the highest reporter signal in all events, with all events able to report 40 Gy at standoff within 24 h of treatment (P < 0.05). The lowest detectable dose at standoff was 10 Gy for all events within 48 h of treatment. 4xRAD51 pro event 3 was the only event that demonstrated the potential to detect signals at 5 Gy; however, this trend was variable between the FILP and fluorometry analysis ( Figure 3c). Among the three events, 4xRAD51 pro event 1 had the strongest reporter signal in both the FILP and fluorometry assays (Figure 3b,c) and therefore was downselected for further testing.
In order to determine the sensitivity to radiation and the time course for functional reporting from a single exposure, data from the 4xRAD51 pro event 1 phytosensor was collected at 8-72 h at a range of exposures ( Figure S5A-C). The spectrofluorometer data showed significant fluorescent induction at 7.5 Gy at 8 h posttreatment (1.6-fold increase compared to 0 Gy plants) and a maximum fluorescence of 5.6-fold above untreated plants when treated with 80 Gy after 72 h. Based on the doses tested, the limit for detection of the 4xRAD51 pro event 1 phytosensor was 7.5 Gy ( Figure S5A). FILP image data collected at 72 h demonstrated a similar trend as spectrofluorometer data, although the FILP signal of plants treated with 7.5 Gy was not significantly different from untreated plants ( Figure S5B). Interestingly, the 4xRAD51 pro event 1 phytosensor produces a strong fluorescent response even when damage is significant enough to cause loss of gravitropic growth ( Figure S5C).

Pressure test of radiation phytosensor in mesocosms
To evaluate the performance of the top radiation phytosensor in a simulated natural environment, 4xRAD51 pro event 1 plants were grown in mesocosms with and without weedy competitors ( Figure S6). 4xRAD51 pro event 1 phytosensors and wild-type controls were analysed in time-course by both spectrofluorometer and FILP starting at 8 h until complete decay to background fluorescent level (0 Gy). Spectrofluorometer data showed significant reporting of 40 Gy at 8 h post-treatment but not 10 Gy. Significant reporting of 10 Gy via spectrofluorometer and FILP occurred at 24 h post-treatment and persisted at 48 h posttreatment in both mesocosm conditions (Figure 4a-c). Plants treated with 10 Gy returned to background fluorescence levels after 7 days when observed by spectrofluorometer and 10 days via FILP, while 40 Gy-treated plants took 1 month to return to preradiation levels (Figure 4a,b). In the FILP images, as shown in previous figures, the stems and leaf veins are visibly brighter than Figure 2 Testing of down-selected gamma radiation phytosensor lines with a high dose of radiation. (a) Induction of expression of the reporter gene (mEmerald) in three phytosensor events of each construct (pUVH1, pPCNA, pRAD51 and p4xRAD51) grown in vitro and analysed at 1.5 h after dosing with 80 Gy of gamma radiation. Graphs represent RT-qPCR relative expression data (2 -ΔΔCT ) of mEmerald coding sequence vs the endogenous reference gene, StEF1α. Data are expressed as mean AE standard error (SE) of 4 biological (exceptions: pUVH1 Events 2 and 3, 0 Gy treatment have 3 biological replicates) and three technical replicates. Data were analysed using ANOVA (P < 0.05) and statistical significance is indicated with '***' while non-significance with 'n.s.'. Wild-type individuals were used as a negative control; 3 biological replicates and 3 technical replicates per treatment. No expression of the mEmerald coding sequence was detected. (b) Fluorescence-inducing laser projector (FILP) images of the same phytosensor plants lines (pUVH1, pPCNA, pRAD51 and p4xRAD51) treated (+; 80 Gy) and non-treated (-; 0 Gy) with gamma radiation and analysed at 24 h post-treatment. Images taken using the following settings: mEmerald fluorescence (2 watts, ex. 465 nm, em. 525/50 nm) (green), chlorophyll fluorescence (1-2 watts, ex. 465 nm, em. 645/75 nm). Scale bar = 10 cm. To test whether phytosensors were able to persist in the environment, plant phenotypic analysis was performed over the plants' lifecycle in mesocosms. Both wild-type and 4xRAD51 pro plants were significantly affected by competition or exposure to radiation, but there was no significant interaction between competition and radiation dose for either genotype ( Figure S7, Table S4). Wild-type and phytosensor plants grown in competitive mesocosms demonstrated significant reductions in fresh total aboveground biomass, fresh tuber mass and number of tubers compared to those grown alone (Table S4). Both genotypes exhibited a significant decrease in number of tubers when exposed to both 10 and 40 Gy radiation in mesocosms without competition ( Figure S7, Table S4). Fresh total aboveground biomass was affected by exposure to radiation in a dosespecific manner in both wild-type and 4xRAD51 pro plants and while there was no significant change in biomass at 10 Gy, there was a significant increase in the aboveground biomass at 40 Gy ( Figure S7, Table S4). Both genotypes exhibited a significant reduction in tuber number and mass after exposure to 40 Gy radiation when grown alone ( Figure S7, Table S4). Leaf number was not affected by competition or radiation in either genotype ( Figure S7, Table S4). No significant interaction between Figure 3 Performance of the p4xRAD51 phytosensor. (a) Four-week-old p4xRAD51 radiation phytosensors grown in potting soil were treated at 0, 5, 10 or 40 Gy of gamma radiation. Graphs showing mEmerald fluorescence signal (cps, count per second) of phytosensor lines measured via spectrofluorometer (ex. 465 nm, em. 509-511 nm) at 8-, 24-, 48-and 72-h post-treatment. Background subtracted from mean using mean spectrofluorometer measurements of wild-type which underwent the same treatment as the transgenic plants. Data are expressed as mean AE standard error (SE) of 3 biological and six technical replicates. Data were analysed using ANOVA (P < 0.05) and comparisons to 0 Gy were evaluated using Post-Hoc Dunnett's test (P < 0.05). Significant difference from 0 Gy via the Dunnett's test is indicated by '***' and groups of comparisons where there was no significant effect of radiation treatment via ANOVA are marked with 'n.s.' (b) mEmerald pixel intensity of plant phytosensors collected using the fluorescence-inducing laser projector (FILP) apparatus. Data are expressed as mean AE standard error (SE) of 3 biological and 1 technical replicate. Data were analysed using ANOVA (P < 0.05) and comparisons to 0 Gy were evaluated using Post-Hoc Dunnett's test (P < 0.05). Statistical significance via the Dunnett's test is indicated by '***' while non-significance is left unmarked. genotype and community or dose was observed, indicating that there is no difference in how the phenotype of the genetically engineered phytosensor responds to dose and/or community compared to wild-type controls (Table S4).

Discussion
Initial irradiation of the S. tuberosum chassis indicated that potato is viable for radiation phytosensing. Our results suggest that gamma radiation treatments below 40 Gy stimulate vegetative growth at the cost of tuber mass, a desirable trait for an ionizing radiation phytosensor. This is in contrast to work done in tomato, showing nearly half the vegetative biomass accumulation in plants treated with 25 Gy from a Calcium heavy ion source (Arena et al., 2019). The difference in radiotolerance between potato and tomato could be due to a myriad of factors, including the difference between ionizing radiation from a Cobalt-60 source versus a Calcium heavy ion source, as well as the increase in ploidy from potato to tomato diluting the effect of mutations (Sparrow et al., 1961). The sunken meristem phenotype observed in this study aligns well with the phenotypes observed in pine species around Chernobyl (Kozubov and Taskaev, 2002) and Fukushima (Watanabe et al., 2015;Yoschenko et al., 2016). Based on biomass data and the sunken meristem phenotype, it is likely that ≥20 Gy disorganizes the balance of gene expression which maintains the meristematic tissues (Carles and Fletcher, 2003), with 80 Gy and above destroying meristematic growth. These field studies do not report the stimulatory 'hormesis' effect of ionizing radiation on vegetative growth seen here and widely in the literature (Kuzin et al., 1986;Maity et al., 2005;Sax, 1955;Volkova et al., 2022), though these pine species were receiving a constant low dose while this study used quick, relatively intense doses. The intention of this sensor was to produce maximum aboveground vegetation to be visible to detectors, therefore phenotypes such as increased stem biomass after ≤40 Gy irradiation aid in sensor performance rather than harm the sensor. The potato chassis is therefore an adequate short-term sensor at 80 Gy and below, while long-term establishment of the sensor can experience a maximum of 40 Gy. Both radiation levels are well above what humans can survive.
Performance of the synthetic promoter construct was superior to the native potato promoter constructs tested in this study. Induction of native gene expression was greatest at 1.5-h posttreatment and continues for StPCNA through 24 h, which is consistent with similar experiments in Arabidopsis (Culligan et al., 2006;Yoshiyama et al., 2009). The StPCNA promoter was expected to perform well based on the genes' essential role as a protein clamp holding DNA polymerases (Bravo et al., 1987;Strzalka et al., 2009) onto lesion sites to coordinate repair of nearly all types of DNA lesions (Strzalka and Ziemienowicz, 2011). Native StPCNA appeared to be highly inducible (26.9-fold at 40 Gy) but this inducibility did not transfer to the StPCNA pro phytosensor construct ( ∼ 2-fold at 80 Gy across 3 transgenic events). When comparing basal expression ( Figure S4A) to FILP images of the 4xRAD51 pro , StPCNA pro and AtRAD51 pro events (Figure 2b), 0.5-1-fold expression compared to the StEF1α reference gene results in fluorescence that can be plainly observed via FILP. If the StPCNA pro was behaving as it did in its native context, these sensors should show no visible fluorescence when uninduced and very light fluorescence when induced. In all three events, there was much more uninduced mEmerald transcript present compared to StPCNA transcript but much lower relative induction from the higher basal level. It is possible that the 3 0 UTR from cowpea mosaic virus included in the phytosensor construct is raising the basal level of transcript by preventing degradation of the mRNA (Meshcheriakova et al., 2014). Additionally, plant PCNA promoters are known to be regulated by a number of cell cycle factors such as E2F transcription factors (Egelkrout et al., 2002;Uemukai et al., 2005). These sites are typically within -600 bp of the transcription start site (Kosugi et al., 1995), though lower induction after gamma radiation treatment could be due to unidentified cis elements that were not included in the StPCNA promoter region used for the phytosensor construct.
Surprisingly, native StUVH1 was only induced at 40 and 80 Gy, StUVR7 was uninduced at all treatments and the StUVH1 pro construct was downregulated in two of the three transgenic events. Studies in Arabidopsis indicate that the UVH1/UVR7 endonuclease is essential for nucleotide excision repair of pyrimidine dimers and that these proteins play a role in genome stability and gamma radiation tolerance (Hefner et al., 2003;Vannier et al., 2009). Poor induction of StUVH1 paired with the fact that its endonuclease partner, StUVR7, was not induced at all by gamma radiation indicates that either: (1) these genes in potato do not have the same role in homologous recombination and nucleotide excision repair as they do in Arabidopsis (Dubest et al., 2004); (2) these genes' role in DNA repair can be carried out without induction of transcript when irradiated with 40 Gy or less; or (3) the genomic context of potato and polyploidy results in other unidentified homologues of the loci tested being upregulated. The possibility of UVH1 and UVR7 having reduced roles in homologous recombination should be investigated, as most plant DNA repair research is conducted in Arabidopsis and rice (Gimenez and Manzano-Agugliaro, 2017). If the process of homologous recombination in Solanum species differs from these model species, this could have major impacts on our understanding of plant DNA repair. Additionally, this gap in DNA repair knowledge could hinder applications such as CRISPR-Cas mediated insertion of foreign DNA into commercially important Solanum species, a process which relies on homologous recombination or non-homologous end joining (Collonnier et al., 2017;Lee et al., 2019).
The AtRAD51 promoter did not behave as expected in the context of its phytosensor construct. Previous characterizations of the AtRAD51 promoter report over 40-fold induction in Arabidopsis seedlings when treated with the radiomimetic Zeocin (Ogita et al., 2018). Qualitatively similar results were seen when the promoter was used in the context of a transgene-driving GUS expression in that same paper. Additionally, the induction of AtRAD51 in response to gamma radiation in its native context is consistent enough to be considered a dosimeter for gamma radiation (Ryu et al., 2018). Like the performance of the StPCNA pro , the difference between the previous reports on native AtRAD51 and the AtRAD51 pro phytosensor construct could be due to the 3 0 UTR added to stabilize the mEmerald transcript. This data could also suggest that other motifs in the AtRAD51 promoter may assist in activation of the gene and that these motifs are not bound by trans-activators in potato. Plant promoters exhibit an incredible amount of variation even among close relatives; furthermore, the AtRAD51 and putative StRAD51 promoter regions exhibit only 44.9% homology and there are no CTT(N) 7 AAG motifs within 2000 bp of the StRAD51 start codon in the published potato genome. This lack of homology suggests that the AtRAD51 promoter would not be regulated similarly to the native Arabidopsis context, even though the overall DDR machinery appears to remain consistent between Arabidopsis and potato.
Despite the poor performance of the AtRAD51 promoter in the context of a potato radiation phytosensor, four repeats of the AtRAD51 SOG1 binding site provided an adequate promoter for a radiation phytosensor. At 10 Gy, this sensor produced a maximum 2.17-fold above basal during the 24-72 h peak reporting window when detected by FILP. This is consistent with other published phytosensors, with reporting beginning at 24 h post-treatment and fluorescence fold-changes in the single digits appear to be the limit regardless of stimuli (Abd El-Halim et al., 2020;Fethe et al., 2014;Persad-Russell et al., 2022). The performance of the sensor is not hindered by field conditions and the sensor construct was not a metabolic burden on the potato chassis. Based on the performance of the 4xRAD51 pro phytosensor, plants deployed in field conditions can report ≥10 Gy of ionizing radiation to a drone-mounted FILP apparatus with a relatively low-powered laser and camera system flying three meters away as soon as 24 h after its release into the environment. This sensing threshold and timescale would be highly useful during a meltdown similar to Chernobyl or Fukushima where radioactive solids were released into the environment and continued to irradiate their surroundings for weeks. In these scenarios, widely dispersed plants could pinpoint presence of an ionizing radiation source based on nearby sensor's peak fluorescence intensity and the duration in which the sensor remains turned on. Based on the sensor's performance in mesocosms and relatively high sensing threshold (LD 99 for humans is 8 Gy), it is unlikely that the 4xRAD51 pro sensor would report DNA damage from stimuli other than ionizing radiation.
Additionally, the 4xRAD51 construct appeared to have 'leaky' expression when uninduced, leading to a relatively high basal fluorescence compared to wild-type potato. Uninduced fluorescence for 4xRAD51 event 1 was 100 000-200 000 cps while wildtype S. tuberosum cv. 'Desiree' was 20 000-60 000 cps, meaning basal fluorescence of the sensor was typically 5-fold higher than the background fluorescence of a potato leaf. This is consistent with the increased rate of the cell cycle occurring in young tissue which will inherently experience more DNA lesions due to doublestranded breaks, stalled replication forks and repairing incorrectly inserted bases (Hu et al., 2016;Schärer, 2003). These expression patterns seen in the potato phytosensor suggest that previous characterization of SOG1-induced DDR in Arabidopsis (Ogita et al., 2018) and rice (Nishizawa-Yokoi et al., 2022) remains consistent in Solanaceae. Though basal stem fluorescence does allow sensor plants to stick out from a surrounding plant canopy, new iterations of the sensor will need to improve upon the established SOG1 transcriptional switch to either lower basal fluorescence associated with growing tissue and/or increase SOG1 presence in leaf tissue types to increase fluorescence in the sensor plant's canopy. This sensor represents an indirect biosensor (Liu et al., 2022) which relies on many molecular interactions which leads to SOG1 binding the 4xRAD51 promoter. Indirect biosensors are inherently prone to more error than the direct biosensors reviewed by Liu et al. (2022) where the reporter is directly activated by the stimulus, which is clear given the high level of basal fluorescence in all 4xRAD51 pro events. If a direct biosensor could be devised for ionizing radiation, this could offer improvements on the current sensor design.
Analytically, leaf spectroscopy was more sensitive in detecting differences between untreated and treated plants compared to standoff detection using FILP. A large factor in each of these measurements was the proportion of vascular tissue in the reading, whether it be leaf veins in the area read by the spectrofluorometer or all vascular tissue visible when imaged with the FILP. Fluorescence was much higher in vascular tissue compared to 'leaf' tissue types (mesophyll, palisade, epidermis, etc.) in all 4xRAD51 events across all experiments. This expression pattern is consistent with reported higher expression of SOG1 in actively growing and vascular tissue (Klepikova et al., 2016;Pagano et al., 2022) and that NAC family transcription factors are broadly associated with meristem and vascular tissue types (Ohtani et al., 2011;Takada et al., 2001;Yamaguchi et al., 2008). The spectrofluorometer used in this work is designed to read flat leaf tissue and attempts to read the stem directly allowed ambient light into the detector. Efficient extraction of FILP pixel data relies on the thresholding function of ImageJ (Schindelin et al., 2012) being done on chlorophyll a fluorescence images of the phytosensor plants, meaning that all green tissue of the plant is averaged into the mean pixel value for a sensor. To optimize sensor performance, the observation equipment used should be tailored to detect stem and vascular fluorescence while disregarding tissue types such as mesophyll which do not fluoresce as brightly in the 4xRAD51 pro sensors.
This sensor currently uses a green fluorescent protein as the reporter which can only be observed with specialized equipment such as a FILP. This reporter was chosen due to low native fluorescence in potato's leaves in the GFP emission spectrum, resulting in a very specific reporter in the context of plant tissue. The requirement of specialized mechanical equipment does not address the problem associated with current radiation sensors, which is that the mechanical parts can either be damaged or unavailable in the event of an emergency. Future iterations of the sensor using a specific, plainly visible reporter would be ideal so that the general public can be warned of danger without the need for equipment. To this end, reporter phenotypes such as high purple anthocyanin accumulation (Zuluaga et al., 2008), bright pink tissue (He et al., 2020), or bleaching (Mochizuki et al., 2001) can be achieved by using different genetic circuits to change cytosolic and plastid metabolism. The key to these reporting strategies is making sure various stressors and life stage transitions do not produce a similar phenotype, thus making the sensor non-specific. The first iteration of the gamma radiation phytosensor is deployable when using the green fluorescent reporter, though future iterations should incorporate a reporter that functions without extra equipment.

Conclusion
In this work, we developed a usable gamma radiation phytosensor able to sense and report ≥10 Gy of gamma radiation in the environment at a standoff distance of 3 m. The Grey unit is a measure of absorbed dose (joules/kg) but this is in contrast to radiation conditions during radiation emergencies. During these emergencies, radioactive material has contaminated the environment and is constantly releasing ionizing radiation at a certain intensity. For example, just after the Chernobyl disaster ambient levels of radiation were estimated to range from 300 Sieverts/h near the reactor core to 0.1 sieverts/h at the nearby concrete mixing unit (Medvedev, 1989). Considering the sievert-to-Grey conversion for gamma radiation, the radiation phytosensor here developed would reach its 10 Gy sensing threshold after 2 min and 100 h, respectively. Treatment times during this experimentation ranged from 2 to 16 h with transcript abundance peaking 1.5 h after treatment and fluorescence peaking 48-72 h after treatment. In this gamma radiation range, 40 Gy was reported as early as 8 h while 10 Gy required 24 h to produce a significant response, meaning this sensor would have been useful for those responding to the Chernobyl disaster and catastrophes like it. More investigation is needed to understand the sensor's performance under longer, low-intensity irradiation. At what radiation intensity will the plant's DNA repair machinery overtake the time the sensor requires to accumulate enough reporter protein? This and other specification questions need to be answered before the sensor can be deployed.
Though improvements and further testing should be carried out, 4xRAD51 pro event 1 is the first ionizing radiation phytosensor that can report at a 3-meter standoff, and is the first selfpropagating, self-repairing dosimeter. This sensor, and phytosensors generally, are important tools as humanity seeks out a more sustainable existence. Through this academic work and future phytosensor research, electricity-independent environmental sensing will become a viable tool for humans to understand our impact on the environment.

Plant material and growth conditions
Wild-type Solanum tuberosum cv. 'Desiree' was procured from the Wisconsin Seed Potato Certification Program at the University of Wisconsin-Madison and kept in sterile culture on modified Murashige and Skoog media for transformation (Chronis et al., 2014). After transformation and genotyping qPCR, events were maintained in tissue culture.
For experiments where plants were in pots, plants were removed from tissue culture and placed into a 4-inch plastic pot filled with a soilless media and allowed to adjust to ambient humidity for 1 week under a closed lid inside of a growth chamber. After this the lid was removed, plants were grown within the growth chamber an additional 3 weeks until the beginning of the experiment. Growth chambers were set to a 16h light/8-h dark daylight regime, with daytime temperature set to 20°C and nighttime temperature set to 18°C. Plants were fertilized with Peter's 20-20-20 fertilizer after hardening and after 3 weeks in soil.

Gamma radiation treatment
Gamma radiation treatment was done by adjusting plants' distance and dosage time from a Cobalt-60 source to reach a total dose (reported in this paper in Grey). Treatment times and distances can be found for each of the experiments in Table S1.

Plant transformation
Transformation of Solanum tuberosum cv. 'Desiree' was performed using an established method (Chronis et al., 2014). At least ten shoots from separate calli were recovered and propagated separately per construct, with ten being kept after successful genotyping via PCR. Three of those events were then selected for further testing based on vigorous growth phenotype and a range of basal fluorescence.

Genotyping via southern blot
Southern blots were performed to determine transgene copy number for the transgenic radiation phytosensor events by established methods (Mellars and Gomez, 2011). Briefly, 5 μg of genomic DNA from three biological replicates of each phytosensor construct and line were extracted using the CTAB method. After extraction, DNA was cut using AflII, BspHI and KasI enzymes and fragments were separated on a 1% agarose gel for 5.5 h. DIG hybridization probes for a 500 bp sequence of mEmerald coding sequence were generated using primers Fw 10 and Rv 18 and the Roche PCR DIG Probe Synthesis Kit. The probe membrane was then placed onto the gel, crosslinked, hybridized and detected according to the above protocol.

qRT-PCR
RNA was extracted from tissue samples stored in RNAlater solution (Sigma-Aldrich) after dosing and extracted using a TRI Reagent extraction protocol (Molecular Research Center) and RNA Clean and Concentrator kit (Zymo) with a DNase I treatment. Two thousand nanograms of RNA from each sample was used to generate cDNA with the Applied Biosystems™ High-Capacity cDNA Reverse Transcription Kit (Fisher). The cDNA was generated, then 1.08 ng cDNA was used for qPCR a single 5 μL qPCR reaction using the PowerUp™ SYBR™ Green Master Mix (Fisher) and its associated protocol. An Applied Biosystems QuantStudio 6 Flex qPCR instrument and its associated software were used. The primers used for native genes, mEmerald and StEF1α can be found in Table S4. For all primer sets, qPCR settings were 2 min at 50°C, 10 min at 95°C, then 40 cycles where temperature begins at 95°C for 15 s, descends 1.6°C per second to 60°C where it holds for 1 min, then ascends to 95°C at the same rate. Results were analysed using the 2 ÀΔΔCT ð Þ method with mEmerald C T values first being set relative to EF1α expression, then relative to the appropriate average 0 Gy ΔC T value to calculate fold change induced by gamma radiation treatment (Livak and Schmittgen, 2001). Primer efficiencies were equivalent and therefore not included in 2 ÀΔΔCT ð Þ calculations.

Spectrofluorometer measurements and data analysis
Measurements were taken with a Fluorolog®-3 spectrofluorometer (Horiba/Jobin Yvon), Excitation for mEmerald measurements was 465 nm and emission was observed from 500 to 515 nm, with 509-511 representing peak emission (Cubitt et al., 1998). The second, third and fourth leaves from the meristem were measured twice on each plant. The counts per second (CPS) for 509-511 nm of each read were then averaged to create a mean CPS for a biological replicate. The mean CPS values for each biological replicate for a construct × event × treatment combination were then averaged to calculate the mean and standard error. Statistics were calculated using the JMP Pro 15 software (SAS, Cary, NC) ANOVA function (P < 0.05) and mean separation with a Tukey's HSD (P < 0.05).

Fluorescence-inducing laser projector images
The fluorescence-inducing laser projector designed by Rigoulot et al. (2021)  laser power of 1-2 watts. All images were processed and analysed for pixel intensity using the ImageJ program (Schneider et al., 2012). Pixel data was analysed in ImageJ by identifying plant tissue in the chlorophyll a image using the Image > Adjust > Thresholding function. Then, Analyse > Analyse particles was used to generate ROIs for the plants in the image. Once generated, the ROIs were measured on the mEmerald fluorescence images and mean pixel values for each plant were averaged to generate a treatment/event average.

Mesocosm experiments
Each mesocosm was contained within a 34.6 cm × 21 cm × 12.4 cm plastic bin filled with a 2 : 1 mix by volume of potting and field soil, collected at the East Tennessee AgResearch and Education Center (Knoxville, TN), respectively. Previous studies have shown a combination of field and potting soil to yield the highest biomass of S. tuberosum. Furthermore, while this soil composition is not solely derived from the field, it is likely to resemble natural conditions around agricultural fields where regular tilling and fertilizing prevent field soil from compacting. Plants were watered as needed.
To assess the effect of competitors on the top performing transgenic S. tuberosum sense and report genotype (p4xRAD51) in response to radiation, communities were constructed with and without weedy neighbours. We chose the following heterospecifics: Cyperus esculentus (yellow nutsedge) and Portulaca oleracea (common purslane). These species were chosen to represent a variety of growth habits (prostrate, erect), life histories (annuals and perennials), status (native and invasive) and reproductive strategies (outcrossing, self-fertilizing and asexual propagation), as these factors have been shown to influence competition among plants (Fazlioglu et al., 2016). In addition, all species are known to commonly occur across the globe and are considered invasive species outside of their native ranges. These competitors were germinated on potting soil in the greenhouse and grown for 4 weeks before being transplanted into communities with focal individuals. Each mesocosm containing neighbours had two replicates per transgenic event and wildtype S. tuberosum, two replicates C. esculentus, and four replicates P. oleracea totalling 8 plants in each competitive mesocosm. For mesocosms without competition, two replicate individuals of each S. tuberosum genotype were planted at each end of the mesocosm, totalling 4 individuals. Each community mesocosm was replicated three times, totalling 6 biological replicates per community per radiation dosage (72 S. tuberosum total).
Mesocosms were grown in the University of Tennessee: Knoxville North Greenhouse. To supplement the available light, LED growth lights (Fluence SPYDR 2x LED Grow Light, Fluence, Austin, TX) were installed and set to a photoperiod of 16 h light/8 h dark. Mesocosms were watered as needed.
One week after transplanting, to allow for plant establishment and limit stress, mesocosms were transported to the cobalt-60 facility for radiation treatments. During transportation, all mesocosm bins were placed in sealed 14 Qt (13.25 L) plastic bins.

Mesocosm measurements
RNA was extracted from tissue samples of the newest three fully expanded leaves of focal S. tuberosum individuals at 1.5 h posttreatment to measure gene expression of mEmerald. Fluorescence was measured using the Fluorolog and FILP, as described above, at 24, 48, 72, 168, 240, 360 and 672 h.
To quantify the effect of competitors on S. tuberosum, traits were measured 8 weeks after planting, when plants naturally senesced. Fresh and dry total aboveground biomass and tuber mass, and number of tubers, leaves, and primary and secondary branches were measured for each focal S. tuberosum. Fresh biomass and tuber mass and number of tubers, leaves, and primary and secondary branches were collected at harvest. For measurements of dry total aboveground biomass and tuber mass, tissues were dried for 2 weeks at 55°C prior to measurement.

Mesocosm statistical analysis
FILP images and trait measurements from mesocosms were analysed using R v. 4.1.3 (R Core Team 2022). Fluorescence, measured as pixel intensity per unit area, was quantified from FILP-generated images using NIH ImageJ software. To test whether transgenic S. tuberosum fluorescence was affected by competition and radiation dose, pixel intensity per unit area was the dependent variable and radiation dosage ('treat'), community, and their interaction were treated as fixed effects in generalized linear models using the 'glm' function in the stats package. Since wild-type S. tuberosum does not naturally produce GFP, p4XRAD51 event 1 plants at 0 Gy in mesocosms without competition were set as the reference for fluorescence; wild-type was not included in the analysis. Analysis of variance (ANOVA) and covariance (ANCOVA) were performed using the stats package and car package (Fox and Weisberg, 2019), respectively. Since there was no significant interaction between dosage and community, the effect of dosage was analysed using the Dunnett's Test to compare multiple treatments to the control using the 'DunnettTest' function in the DescTools package within each community type (Signorell et al., 2021).
For trait measurements from mesocosms, traits were treated as the dependent variable and genotype, dosage, community and their interactions were treated as fixed effects using generalized linear models as described above; total aboveground biomass and tuber mass measurements were analysed using a Gaussian distribution, while number of tubers, leaves, and primary and secondary branches were analysed using the Poisson distribution. Wild-type plants in mesocosms without competition at 0 Gy were set as the reference. ANOVA, ANCOVA and Dunnett's Test for each trait were performed as described above. In addition, to test for effect of community type on traits, pairwise differences for each trait were calculated between community types for each genotype within each radiation dosage. We used the function 'pairwise' in the emmeans package (Lenth, 2022). We estimated the effect of competition for each estimated marginal means using Tukey's method to adjust significance thresholds for multiple comparisons.
Accession numbers. This work did not utilize any accessions.
Special thanks to Max Cichon, Tamara Isaacs-Smith, and the staff at Auburn University whose assistance was essential to the testing of the sensor.

Author contributions
RGS, SBR, SCL and CNS designed the phytosensor strategy; RGS, HB, EMS and CNB generated transgenic phytosensors in potato; RGS, AO, BM, TK, HB, CNB, EMS, BJ, CB, YY and TMS collected data; RS, AO, BM and YY analysed data; RS, SBR, AO, BM, CNS and SCL wrote the article. All authors approved the final manuscript.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.

Figure S1
Post-irradiation potato phenotype at anthesis. Figure S2 Post-irradiation potato phenotype at harvest. Figure S3 Full design of radiation phytosensor constructs used in the study. Figure S4 Basal expression and transgene copy number for phytosensor lines. Figure S5 Full specification of the p4xRAD51 event 1 phytosensor. Figure S6 Field site mesocosm layout and irradiation. Figure S7 Post-irradiation potato phenotypes after mesocosm irradiation. Table S1 Gamma radiation treatment specifications.