Synthetic metabolic and genetic networks for medical diagnostics – SynBioDiag

The project is composed of four main tasks. In a first task (Task 1), clinical samples are collected in serum, urine, and tissues of prostate cancer (PCa) patients at the University Hospital of Montpellier. The sample biomarkers are then quantified by liquid chromatography and targeted high-resolution mass spectrometry (LC-MS). We then design biosensors for biomarkers with differential concentrations based on literature and LC-MS analysis on patients and control samples. Hybrid analog/digital circuits are then constructed in E. coli (Task 2) to detect multiplexed signals (i.e., multiple biomarkers) through metabolic pathways and to amplify the signals using genetic switches. The circuits expressed in E. coli cell extracts are lyophilized and patterned on paper (Task 3). Finally (Task 4), the biosensors are rehydrated and tested on the clinical samples acquired in the first task. The biosensor results (fluorescence levels) are compared to LC-MS data and PSA screens (prostate cancer antigen screens). Biosensors are optimized by modulating metabolic pathways and genetic switches.

Biomarker selection and biosensor design. By means of a bibliographic study coupled with bioinformatics analysis, we established a list of 15 metabolite biomarkers of prostate cancer (PCa). This list of biomarkers was sent to the LC-MS analysis laboratory in order to characterize their concentration in clinical samples collected by the CHU of Montpelier. Biosensors (hubs) allowing the indirect detection of several biomarkers via various transducers have been particularly prioritized. Indeed, these hubs are necessary for the development of more complex circuits (adders, perceptron...).

Analogous detection systems. A series of selected biosensors have been constructed and characterized in a cell-free environment. Following a methodology developed during a proof of concept (construction of complex systems realized for benzoic acid derivatives), we have developed and optimized a cell-free biosensor detecting hydrogen peroxide, a central effector (hub) to which we have been able to connect several transducers allowing the detection of various previously identified biomarker molecules. Biosensors detecting Sarcosine and Lactate have been developed and optimized.

Hybrid analog/digital biosensors. We have placed recombinase switches under the control of transcription factor hubs such as H202 capable of being integrated with metabolic transducer signals (lactate dehydrogenase, converting lactate to H202). The use of a metabolic transducer allowed the activation of the recombinase switch by the molecule of interest. These results validate the feasibility of our approach. The switches thus generated will now be connected to different metabolic transducers allowing the conversion of biomarkers of interest (e.g. sarcosine).

Established biomarkers of prostate cancer also include non-coding RNA, in particular PCA3. While we did not anticipate this type of detection in our initial project, a toehold switch biosensor allowing the detection of PCA3 RNA has also been developed to eventually be coupled with the other biosensors previously developed. This new development required the use of bioinformatics approaches to develop sequence-specific RNA detectors regulating translation in vivo or in vitro and allowing the detection of PCa transcriptional markers.

We have also started to develop a series of hybrid devices integrating concentration information from several types of molecules (RNAs, metabolites, and proteins) for the detection of model molecules (benzoate derivatives) and biomarkers of PCa. In particular, we have developed a method to build weighted adders on cell-free systems. These weighted adders have allowed us to develop a perceptron (the basic unit of artificial neural network systems) to classify samples based on their metabolic compositions. While this work, published in Nature Communication, has been exemplified on a model system, it is currently being implemented for the multiplex detection of PCa biomarkers (sarcosine, lactate, PCA3, lactate dehydrogenase/LDH).

Pandi A, Koch M, Voyvodic PL, Soudier P, Bonnet J, Kushwaha M*, Faulon JL*. Metabolic Perceptrons for Neural Computing in Biological Systems. Nature Communications, 10: 3880, 2019. | doi: 10.1038/s41467-019-11889-0

Batista A, Soudier P, Kushwaha M, Faulon JL*. Optimizing protein synthesis in cell-free. IET Engineering Biology, 5(1): 10-19, 2021. | doi: //doi.org/10.1049/enb2.12004
Voyvodic PL and Bonnet J. Cell-free biosensors for biomedical applications. Current Opinions in Biomedical Engineering 13: 9-15, 2020.

Pandi A., Borkowski O., Faulon JL. (2020) Synthetic Biology at the Hand of Cell-Free Systems. In: Singh V. (eds) Advances in Synthetic Biology. Springer, Singapore | doi.org/10.1007/978-981-15-0081-7_2

Du nouveau dans l’apprentissage automatique via des systèmes biologiques, INRAe/INSERM press release, 16/09/2019, www.inrae.fr/actualites/du-nouveau-lapprentissage-automatique-systemes-biologiques

Organisms have evolved to sense large panels of metabolites and human have used these sensing abilities throughout history. Recently, synthetic biology has expanded and rationalized biosensing by engineering whole-cell biosensors (WCBs) with sophisticated signal processing capabilities. The main advantages of cell-based technologies over abiotic detection based on purified antibodies, nucleic acid hybridization, or metabolomics analysis are lower cost, improved stability, and the possibility to be ultimately used at the point-of-care or as a personal healthcare home device.

Our project aims to engineer a scalable, programmable WCB platform for the multiplexed detection of biomarkers in clinical samples. As a model system, we will focus on the detection of chronic diseases biomarkers, which clinical samples are abundant and multiple gold standard tests exist to which benchmark our platform. To this end, serum, urine and tissue samples will be collected from patients of the Montpellier University Hospital.

Our project is timely and highly relevant as the partners have recently addressed two main bottlenecks in developing WCB for the clinic namely: (i) scaling-up the number of analytes that can actually be detected and (ii) solving reliability issues of WCB operation in complex biofluids media.

Typical WCB are triggered by no more than half a dozen input signals. To palliate this shortcoming, the MICALIS partner has recently expanded the range of biologically detectable molecules up to a thousand by systematically designing and engineering metabolic pathways that transform non-detectable chemicals into molecules for which sensors already exist. The method has been successfully benchmarked to engineer biosensors that detected pollutants, drugs and biomarkers such as benzoic acid and hippuric acid.

The other main challenge limiting WCB deployment has been their unreliable operation and low signal to noise ratio in complex and heterogeneous samples like physiological fluids. The CBS partner solved this issue by developing a signal-processing platform using amplifying genetic switches and demonstrated the detection of abnormal glycosuria into the urine of diabetic patients. This work validates the operational capacity of WCB technology in clinical samples.

Here we will combine the advantages of the above recent developments by first sensing biomarkers, through synthetic metabolic pathways acting as analog devices and then triggering a digital response using genetic switches robustly operating in clinical samples. Such a hybrid analog/digital system has not yet been developed for biosensors.

Our WCBs will be implemented with bacteria (non-pathogenic strains of E. coli) and on paper-based cell-free systems. Bacterial sensors can encompass complex sensing circuits, do not need expensive resources even for long times scale operation and are cheap to produce. However, biomarkers that are toxic or that do not cross cell membrane are difficult to detect. Toxicity and transport issues are irrelevant with paper-based cell-free biosensors. With paper-based cell-free systems, biosensors can be stored when freeze dried for a long time, the sensing circuits can be easily optimized by simply changing the amount of DNA encoding the different functions, and since cell-free systems are abiotic they present less release concerns and regulatory hurdles.

While the platform will be benchmarked for detecting prostate neoplasia biomarkers, it will be set up through a robust engineering workflow supporting fast repurposing to many other medical diagnostics especially those sharing biomarkers with prostate disorder (such as schizophrenia, uremia, chronic renal failure).