Description of an activity-based enzyme biosensor for lung cancer detection

Background Lung cancer is associated with the greatest cancer mortality as it typically presents with incurable distributed disease. Biomarkers relevant to risk assessment for the detection of lung cancer continue to be a challenge because they are often not detectable during the asymptomatic curable stage of the disease. A solution to population-scale testing for lung cancer will require a combination of performance, scalability, cost-effectiveness, and simplicity. Methods One solution is to measure the activity of serum available enzymes that contribute to the transformation process rather than counting biomarkers. Protease enzymes modify the environment during tumor growth and present an attractive target for detection. An activity based sensor platform sensitive to active protease enzymes is presented. A panel of 18 sensors was used to measure 750 sera samples from participants at increased risk for lung cancer with or without the disease. Results A machine learning approach is applied to generate algorithms that detect 90% of cancer patients overall with a specificity of 82% including 90% sensitivity in Stage I when disease intervention is most effective and detection more challenging. Conclusion This approach is promising as a scalable, clinically useful platform to help detect patients who have lung cancer using a simple blood sample. The performance and cost profile is being pursued in studies as a platform for population wide screening.

Supplementary Table 1: Enumeration of number of layers in explosion based turbostratic graphene: Graphene biosensors were examined using a 200kV Hitachi H8100 thermionic field emission transmission electron microscope at an electron acceleration voltage of 200 kV.TEM images were captured using a normative and standardized electron dose on eucentric specimen stage and a constant defocus value from the carbon-coated surfaces.Images were collected at 300,000x magnification.Numbers of layers were counted on images that showed clear edges.Analyzer.Oxygen was analyzed by pyrolyzing the sample in a helium environment in contact with a nickel-plated carbon catalyst at 1060 °C.The resulting nitrogen, hydrogen, and carbon monoxide gases are then separated and the carbon monoxide is analyzed in a thermal conductivity analyzer.
Oxygen content was then calculated from the amount of carbon monoxide produced.Results are normalized to 100% w/w and averaged for replicates.
The working solutions were prepared and sonicated in a waterbath for 1 min per ml of solution.Samples were vortexed once immediately before sampling 2 µL three times.The optimal absorbance was established by measuring the absorbance across 250 nm to 650 nm on a Nanodrop.The samples were then allowed to sit for 1 hour without any agitation.After 60 min, 2 µL samples were taken with a pipette tip submerged to 50% depth in the sensor solution.The absorbance spectrum was repeated on the time60 sample.The performance is displayed as a percentage change in absorbance efficiency at OD265.
Alternatively, working solutions were further stored at room temperature for 9 months.After 9 months, the solutions were vortexed and 2 µl samples were measured at OD265.The change in absorbance from time0 is presented as percent change.The analysis shows the stability of sensor reflected in the low change in absorbance over time.This observation is not dependent on the biosensor measured.Supplementary Table 8: Precision.Intra assay precision was assessed by evaluating the frequency that an assay on a single sample repeated on three different days gave the same result.The entire dataset included 150 precision samples that were assessed on three independent assays.The Test set precision was evaluated on 50 Test Set out of sample intra-assay repeats.
Supplementary Figure 1: Raman analysis of a sample of graphene-polymer based nanoparticles.
Raman spectrum of 4 separate lots of detonated graphene powder was measured using the Reinshaw inVia Reflex Raman and a 532nm excitation.Raw dry graphene powders were measured unprocessed and at ambient air and pressure.The Raman spectra of detonated graphene collected with 532 nm excitation are measured.The spectra contained the G band at 1575 cm -1 , a strong D band at 1340 cm -1 and a symmetrical 2D band at 2680 cm -1 , but weaker than the G band.The relative position and sharpness of the G band across the various lots represents consistent multilayer thickness graphene.The presence of a strong D band indicates reduced in-plane sp 2 domains due to the ring breathing mode from sp 2 carbon rings adjacent to graphene edges or defects; the D band is expected to be significant in our detonated graphene powder material.Since 2D band was symmetrical, the sample was not graphite but graphene, which was multilayer graphene since the 2D band was weaker than the G band.The 2D/G intensity ratio also confirms consistent multilayer graphene present in our material.The presence of a sharp and symmetrical 2D band is the result of a two phonon lattice vibrational process always present in graphene and not graphite.The Raman spectra of the sample were typical of multilayer graphene.Overall size can be seen and the layered structure of the graphene sheets are revealed.5 mg of each biosensor was mixed with 1mL of 100% ethanol and sonicated for 5 minutes.
5 µL of diluted biosensor solution was placed on a 300-mesh carbon-coated copper grids and incubated for 1 minute.After incubation, a piece of filter paper was used to wick away the remaining solution and grid was placed onto a clean piece of filter paper to let air dry.
The grids with nanobiosensors were examined using a 200kV Hitachi H8100 thermionic field emission transmission electron microscope at an electron acceleration voltage of 200 kV.TEM images were captured using a normative and standardized electron dose on eucentric specimen stage and a constant defocus value from the carbon-coated surfaces.
Images were randomly acquired at 10 different locations within the grid.

2 Supplementary Figure 4 :
GEO data sets comparing primary tumor and healthy human tissue.A search of the GEO Profiles database revealed differences between lung cancer and non lung cancer (GEO accession GDE40275; Kastner et al., 2012 1 .GSE30118; Ihsan et al., 2011 2 .GSE6044; Rohrbeck et al., 2008 3 ) accessed in October 2020.Graphs showing differentially expressed matrix metalloproteinase and cathepsin enzyme targets including Arginase, as post translational modifiers that distinguish lung cancer and non-lung cancer samples by expression.Also, the same panel might have efficacy distinguishing NSCLS and SCLC based on a similar list.(* p ≤ 0.001, **p ≤ 0.0001)The peptide designed for CTSL was insoluble in several designs and so was not included in the final panel.This list of targets was expanded to include peptides included in previous studies with utility in breast, pancreatic and lung cancers described in Table1.

Table 6 :
Coefficient of variation at 1:10 sera input.14 sera were assayed with triplicate measurements and the variance between the triplicates were evaluated for each biosensor to determine the intra-plate variance.Data presented are for the 50 min measurement.
Supplementary Table5: Standard Curve.A serial dilution of TCPP-peptide conjugate was prepared in assay buffer.Triplicate wells were run with each assay plate and a polynomial standard curve was used to calculate the concentration of TCPP-peptide product in individual biosensor reactions.Across 85 plates, the average R 2 for all standard curves was 0.9999.One example curve is presented showing average relative fluorescence units (RFU) at each dose +/-s.d.Supplementary