Currently, there are two types of defect detection systems used to monitor the health of freight railcar bearings in service: wayside hot-box detection systems and trackside acoustic detection systems. These systems have proven to be inefficient in accurately determining bearing health, especially in the early stages of defect development. To that end, a prototype onboard bearing condition monitoring system has been developed and validated through extensive laboratory testing and a designated field test in 2015 at the Transportation Technology Center, Inc. in Pueblo, CO. The devised system can accurately and reliably characterize the health of bearings based on developed vibration thresholds and can identify defective tapered-roller bearing components with defect areas smaller than 12.9 cm2 while in service.
The use of engineered microbes as smart or responsive chemical production systems that produce different outputs in response to different environmental stimuli holds great potential for diverse applications, such as on-demand pharmaceutical production, targeted long-term drug delivery, and minimal-equipment diagnostics. Microbial cells respond to their surroundings, produce complex chemicals, self-replicate, and thrive in complex and harsh environmental conditions, making them excellent candidates to meet the need for responsive, robust, and scalable chemical production. However, the creation of such responsive microbial factories is hindered by the disconnect between biosensor development and microbial metabolite production: most complex cell sensors produce simple reporter proteins in response to target signals1,2, and most metabolically engineered cells constitutively produce a single target chemical. Incorporating metabolite outputs into microbial sensors brings new difficulties and complications that require new biosensing and metabolic engineering strategies.
Engineering cells to selectively produce only one of multiple possible metabolite outputs (rather than protein outputs) poses many challenges. Expression of multiple enzymatic pathways can be metabolically burdensome3, which leads to low cell viability and construct stability. Perhaps even more importantly, small amounts of uninduced, baseline enzyme expression can produce large amounts of metabolites at undesired times or conditions, confounding cell output selectivity and preventing a controlled response4,5. Metabolic engineering can improve yields and titers of desired metabolites, but the requirement of high output selectivity is not a consideration in typical metabolic engineering strategies5. In fact, this selectivity is actually at odds with typical emphases on improving titer, which often lead to leakiness even under repression.
Beyond selective production of target metabolites, sensor cells must also respond to relevant ranges of analyte concentrations. A diversity of effective approaches to shift sensor response points could help facilitate the development of tunable, more applicable biosensors. Most sensors that use bioprospected components respond to concentrations dictated by the natural sensor-analyte affinity, which often fall outside of industrially or clinically relevant response ranges. Currently, the most effective approaches to shift sensor response curves require extensive protein engineering6 or the use of multiple receptors with varied binding affinities for the target analyte7,8, which can necessitate further bioprospecting or protein engineering.
Here, we describe the development of a robust whole-cell zinc biosensor that accurately quantifies clinically relevant concentrations of zinc in human serum. We design and engineer a library of dual-input promoters that decouple pigment metabolism from cell growth and thus enable visible output within four hours of sample addition in laboratory conditions. We then use transcriptional, translational, post-translational, and metabolic control methods to tune which zinc concentrations activate different pigment pathways, ultimately shifting the threshold zinc concentrations by nearly an order of magnitude compared to our previous proof of principle and into the clinically relevant range. The resulting cells can be used to assay unprocessed serum in a robust, field-friendly fashion: lyophilized cells can be rehydrated with small volumes of human serum, and they produce one of three different visible pigments based on the concentration of serum zinc. The user can assess output qualitatively through visual test assessment or quantitatively with an easy-to-use smartphone app. This work is a major step towards a field-deployable micronutrient biosensor that would impact the treatment of millions of people, and demonstrates generalizable strategies for the development of robust and tunable sensor cells.
Our previous efforts to engineer diagnostic zinc sensor cells were plagued by long assay times, construct instability, and poor test interpretability, which needed to be overcome in order to make a potentially deployable diagnostic. To accomplish this goal, we sought to decouple pigment production from cell growth by using a small molecule inducer as a master regulator to control expression of all pigment pathways. During a pre-assay inoculum production stage (in the absence of inducer), cells should be colorless, which should decrease the metabolic burden on cells during this growth phase and thus improve cell viability and genetic circuit construct stability17,20,21,22,23. A large inoculum of these cells would then be added to the sample, and a standard inducer (such as IPTG or arabinose) would activate the color synthesis system, leaving zinc-responsive elements to control pigment production (Fig. 1a). The assay time would then depend only on the pigment biosynthesis rate, and not the time needed for generation of visible biomass from a small (colored) inoculum, as in our proof-of-principle work. This would also improve test interpretation (since colorless cells will indicate an incomplete or nonfunctional test).
We determined the best-responding promoter of each class by calculating the relative change in fluorescence caused by inducer and zinc addition (Supplementary Fig. 3), and then used promoters with the optimal dynamic ranges to control production of the metabolic pathway that produces the purple pigment violacein. A starter culture of uninduced cells that had been grown overnight was used to inoculate tubes containing fresh medium, the appropriate inducer, and different concentrations of zinc. Though all systems show some level of induction, the PLac-based system outperforms the others, as it has visually undetectable violacein production in the overnight culture and the highest level of induced violacein production (Fig. 1e).
Using the optimized PLac-based violacein circuit, we constructed multi-color zinc-responsive sensor cells by incorporating metabolic pathways from the carotenoid pathway, specifically those to produce the red pigment lycopene and the orange pigment β-carotene. We used LacI/IPTG to control expression of the lycopene production pathway and the zinc-responsive activator ZntR to control expression of the enzyme CrtY, which converts lycopene into β-carotene (Fig. 2a). Upon IPTG induction, cells always produce lycopene, but cells should only appear red at intermediate zinc concentrations: at low zinc concentrations, the more intense pigment violacein overwhelms lycopene and cells appear purple, while at high zinc concentrations CrtY converts all lycopene into β-carotene and cells appear orange. The mevalonate pathway genes (which produce the precursor to the carotenoid pathway) were expressed from PBad to increase the rate and intensity of carotenoid production.
As designed, the implemented engineered cells turn one of three colors based on the zinc concentration. After overnight growth (with no inducer), cells are visibly colorless and have no detectable violacein or carotenoid. When these cells are used to inoculate medium containing IPTG and different concentrations of zinc, cells appear visibly purple, red, or orange (Fig. 2b) after four hours of culture.
Using lyophilized cells and body incubation, we next evaluated whether sensor cells could reliably assess zinc levels in serum that we isolated from individual donor blood samples. Because all donors had healthy or borderline zinc levels (Supplementary Table 2), we treated each with a Chelex-100 resin to deplete zinc and then supplemented zinc to create serum with a range of zinc concentrations. In all samples, lyophilized sensor cells produce different colors based on zinc concentration, but the relative color and color intensity varied across experiments, which we attributed to slight differences in body temperature and motion between tests done on different days.
We have demonstrated the development of a field-deployable pigment-based biosensor for micronutrient testing and have developed generalizable strategies to decrease the time to response, increase cell stability, tune the response range, and account for environmental variability. The final sensor cells turn purple, red, or orange to indicate whether a serum sample has dangerously low, borderline, or healthy levels of serum zinc, and results can be interpreted through visual inspection or with a smartphone app. Compared to previous proof-of-principle work, we reduce assay time, tune the sensor response point by an order of magnitude to clinically relevant concentrations, and show that the test can be reliably and accurately used as the basis for a minimal-equipment assay.
The developed sensor cells can serve as the foundation of a zinc diagnostic that meets all requirements for in-field micronutrient assessment. Tests can be stably stored at ambient temperature, require minimal sample processing, and produce easily interpretable results. If incorporated into standard health surveys, this test could provide governments and aid organizations with the information needed to efficiently implement zinc supplementation programs. 2b1af7f3a8