Synthesized and purified molecular beacons and complimentary DNA oligomers were purchased from Integrated DNA Technologies (IDT, Coralville, IA, United States). All the beacons in the present study use as fluorophore 6-carboxyfluorescein (6-FAM) at the 5′end, and as quencher 4-((4-(dimethylamino)phenyl)azo)benzoic Acid (Dabcyl) at the 3′end.

To explore the role of the guanine (G) cytosine (C) content in the stem of MB in the kinetics of MB-target formation, we designed MBs with 3 different percentages of GC content (40,60, and 100%). The loop section of the MBs was the same in all of them. The MB used in these experiments was Q − 3 ′ − gagagAAACCAATAGATCGACATActctc- 5 ′ -F   ( 40 %   GC   content ) Q − 3 ′ − aggagAAACCAATAGATCGACATActcct- 5 ′ -F   ( 60 %   GC   content ) Q − 3 ′ − gggggAAACCAATAGATCGACATAccccc- 5 ′ -F   ( 100 %   GC   content )

In all caps is shown the sequence of the loop in the MB and in lowercase the sequence corresponding to the stem.

To explore the overlapping of the target in the loop and stem from the MB, the following target sequences were designed:

The sequences from the targets that are complementary to the loop from the MB are shown in all caps. Lowercase shows the sequences complementary to the stem region of the MB. The numbers to the right indicate the number of nucleotides overlapping with the stem from the MB, both unilateral (first 6 target sequences) and bilateral (five next target sequences).

We follow a bioinformatics strategy to identify the best sequences in the different influenza genes to ensure high specificity when designing the MBs.

Influenza virus sequences were obtained from the Influenza Virus Resource: http://www.ncbi.nlm.nih.gov/genomes/FLU/FLU.html. All available sequences of the hemagglutinin HA (H1, H3) and neuraminidase NA (N1, N2) genes were downloaded from the aforementioned database and aligned using the command line version of the ClustalW program. The most conserved regions of each HA and NA subtype were selected using BioEdit software, from the previously generated alignments. Each of the conserved sequences was analyzed using the mfold server http://dinamelt.bioinfo.rpi.edu/quikfold.php; this analysis served as a filter to obtain those conserved sequences that contain a minimum secondary structure that does not interfere with the hairpin formation of the MBs. Some of these conserved sequences did not work to specifically detect each virus subtype since large regions are conserved in the gene. Due to this, it was necessary to carry out an “in silico” coverage analysis (Durai and Schulz, 2019), to ensure the specificity of the sequences for each of the influenza virus subtypes. The best sequences obtained after the analysis are shown in the table below:

The “in silico” coverage analysis help define the sensitivity and specificity of each sequence (Durai and Schulz, 2019). The results obtained are illustrated in the table below:

These sequences were used as the loop portion of the MBs, and a stem sequence was added to produce the final MBs sequences shown below (these MB sequences are protected by patent 344,352):

Underlined are the influenza genes’ complementary sequences corresponding to the loop regions of the MBs.

There are different methods to generate silk hydrogels under controlled conditions. In the present study, we induced the hydrogel formation with methanol, as previously described (Numata et al., 2014; Numata and Kaplan, 2010). Briefly, cocoons from the silkworm Bombyx mori were grounded and boiled for 30 min in a 20 mM Na₂CO₃ solution and then washed with distilled water to remove sericin proteins and wax. Extracted proteins were dried and dissolved in a warm 9 M LiB solution and kept for 2 h. The silk solution was dialyzed using distilled water for 2 days using a dialysis membrane with a molecular weight cut-off of 3,500 (Thermo Fisher Scientific, Waltham, MA, United States). The ilk solution was then concentrated. To produce the hydrogels, an aliquot of the concentrated silk solution was introduced into a silicon cylinder and the cylinder was immersed in methanol for 12 h at room temperature (25–27°C). The resulting silk hydrogel was mixed with 1 µM of the different molecular beacons used in this study.

The silk hydrogel containing the different molecular beacons described in the previous sections was utilized to manually print the microarrays using a Gilson P2 micropipette and the TIRF microarrayer shown below in panel A (TIRF Labs, Cary NC, United States). Typically, each microarray spot was approximately 1 mm in diameter and about 1 mm in height when dispensing 1 µl on each spot (phase contrast image from the microscope in panel B).

The microarray slides were mounted on a conventional epifluorescence inverted microscope (Nikon Instruments, Japan) using a ×10 objective. Images were acquired using an inexpensive 5 MP CCD camera from Amscope (MT5000-CCD). All microarray data were analyzed using the ImageJ line profile tool and Igor Pro v7 (WaveMetrics, Portland, OR, United States). A target was applied to the microarray slide using a Gilson micropipette.

We use an illumination device that excites the sample from a lateral pathway (TIRFLabs, Cary, NC). The device can be mounted on a conventional epifluorescence inverted microscope (Asanov et al., 2012). The use of the shallow angle illumination device results in having independent excitation and emission optical pathways, promoting reduced stray light and cross talk between the excitation and emission channels (Asanov et al., 2012). Shallow angle illumination results in the excitation of the sample only a few microns above the surface of the coverslip, preventing the excitation of the bulk solution. With each microarray spot having a height of 1 mm, with shallow angle illumination, we ensure that the excitation light is focused primarily on the microarray and not the solution above. A detailed description of the optical properties and physics behind shallow angle illumination can be found elsewhere (Asanov et al., 2012; Asanov et al., 2010).

Silk hydrogel was prepared using a synthetic molecular beacon for H1 (MB_H1) in which the 5′ quencher was not included in order to have a MB that emits fluorescence continuously in the absence of the target. Silk hydrogel microarrays were printed with this MB and confocal microscopy experiments were conducted to reconstruct the tridimensional shape of the droplets. An Olympus FV1000 confocal microscope was utilized for the morphological characterization of the hydrogel droplets. Images of the entire microarray were obtained with a low magnification (×20) objective and a resolution of 2,048 × 2,048 pixels in the x-y plane. The z-plane optical slices were obtained every 200 nm. The entire image stack was loaded into Imaris v8.1 (Oxford Instruments, United Kingdom) for the tridimensional reconstruction and plane projections illustrated in Supplementary Figure S1.

We use our shallow angle illumination device mounted on a conventional epifluorescence inverted microscope (material and methods) (Asanov et al., 2012). Unlike conventional microarray readers, which are basically confocal microscopes, our device utilizes a camera to acquire the fluorescence of the entire microarray in one step (conventional microarray readers read one pixel at a time). This procedure provides a time course of fluorescence increments, that reflects the interaction between the bio analyte (target) and the molecular beacon (MB).

Figure 1 illustrates the typical reaction of a MB and its target. Under resting conditions (without the presence of the target), the MB should be in the rest (non-fluorescent) state, forming a hairpin which brings together the fluorophore and its quencher, preventing light emission in this way. Upon the addition of the target, the bimolecular association MB-target is formed, resulting in the spatial separation of the fluorophore and its quencher, and therefore light emission. Ideally, the bimolecular MB-target state should be more favorable to prevent dissociation of the duo, resulting in reduced fluorescence signal. To ensure that the bimolecular state is favorable in this reaction, the stem from the MB must be carefully designed. Typically, a high GC content in the stem will prevent spontaneous separation of the fluorophore and its quencher but will also make more difficult the formation of the MB-target binomial (since the thermodynamics would favor the hairpin state). Since traditionally MBs have been used in PCR reaction, where temperature is controlled by a thermal cycler, the MBs used in PCR reactions have typically high GC content in their stems (70%–100%). This high GC content would not work in assays conducted at room temperature, and therefore is not suitable for point-of-care diagnostics using simple equipment. In fact, the thermal transition of MBs is such that at room temperature, there is basically no florescence emitted by the probe (Tyagi and Kramer, 1996).

To overcome this design, we explored the GC content usable for MB-target associations at room temperature (25–27°C). We also explored several stem lengths to identify the best combination of stem length and GC content to favor MB-target association at room temperature.

The MBs contained in our silk hydrogel were printed on the surface of silica microarray glass (1 mm thick) and the slide was mounted on the inverted microscope and fluorescence was monitored using a CCD camera (Material and Methods) and our shallow angle illumination device. Silk printed hydrogels form a tridimensional droplet of approximately 1 mm in diameter and 700–800 microns in height (Supplementary Figure S1). A typical fluorescence experiment is illustrated in Supplementary Video S1. The mean fluorescence intensity was plotted over time to illustrate the time course of association of the MB-target and its kinetics. The video shows how the area of each microarray spot was measured using a region of interest (ROI) in ImageJ (material and methods). For each microarray experiment a replica of at least six spots was measured to obtain the mean ± standard deviation. At least 5 independent microarrays were evaluated for each of the conditions explored. Supplementary Video S2 shows the time course of fluorescence increments upon target application.

Because leakage of the MBs from the hydrogel would result in reduction in the fluorescence intensity of the microarray, we characterized the time course of MBs leakage from the hydrogel. Supplementary Figure S2 (and Supplementary Video S3) shows that the leakage of the MBs is a very slow process that does not interfere with the fast association times of MB-target interactions.

To explore the effect of the GC content in the stem domain of MBs, we designed 3 MBs with different stem GC content. In all three cases, the stem length was 5 nucleotides long, and we varied the GC content from 40, 60, and 100%.

As illustrated in Figure 2, using a stem with 100% GC resulted in very slow association kinetics between the MB and its target (shown in fuchsia). Using the same target, we next explored the association between the MB with 60% GC content and its target (shown in red). In this case, the association kinetics were faster, with the fastest association obtained when using 40% GC in the stem (shown in blue).

In the next series of experiments, we were interested in exploring if the length of the stem affected the association kinetics between MB and its target. For this purpose, we designed 3 MBs with stems 5, 7 and 9 nucleotides long (Figure 3A). Using a stem of 9 nucleotides (even when the stem contained 40% GC as in the previous experiments), the association kinetics were very slow (shown in black). Reducing the stem length to 7 nucleotides sped up the association kinetics between the MB and its target (shown in red), but the fastest association was obtained with MBs formed by a stem 5 nucleotides long (shown in green).

Reducing the length of the stem to 4 nucleotides resulted in increased background fluorescence in the absence of target, indicating that a portion of the MB was in the open (fluorescent) conformation (Figure 3B). A greater background fluorescent level was observed when using MBs with 3 nucleotide long stems (Figure 3B).

These results indicate that the fastest association kinetics at room temperature are obtained with MBs formed by a stem 5 nucleotides long and with a GC content of 40%. This MB configuration did not compromise the MB conformation in resting (non-florescent) conditions, which results in low background levels and reduced signal-to-noise ratios in the absence of the target.

After identifying the best configuration for MBs consisting of 5 nucleotide long stems with 40% GC content, we proceeded to determine the effect of overlapping the target only in the loop region of the MB versus overlapping the loop and invading a portion of the stem.

To conduct these experiments, we designed targets that hybridize with the entire loop and with 2, 4, and 5 nucleotides from the stem on one side (unilateral, Figure 4A) or with the same procedure but overlapping on both sides of the stem (bilateral, Figure 4B).

The results obtained indicate that overlapping unilaterally or bilaterally made no difference in the association kinetics between the MB and its target. However, the number of nucleotides invading the stem did make a difference, with increasingly faster kinetics as the number of invading nucleotides augmented. Unilaterally or bilaterally, the fastest kinetics were obtained when the target complemented the entire stem sequence (5 nucleotides long, compare the blue lines in 4A and 4B).

The higher background fluorescence reported here agrees with previously published (Tsourkas et al., 2003) results indicating that MBs with stems consisting of 4 nucleotides still recognize their target but show high background fluorescence in the absence of the target.

To explore the feasibility of using our MB-silk microarray technology in molecular diagnostics, we designed experiments directed at identifying cDNA sequences from different influenza virus subtypes. To design the most selective MBs for influenza we conducted in silico analysis of sequences to identify unique signatures that may serve in the design of MBs selective for each influenza virus subtype (material and methods). After all the in-silico analysis, we identified specific sequences for hemagglutinin (H1, H2 and H3) and neuraminidase (N1, N2) genes from human and two genes from swine origin (H1s and N1s). The sequences of these MBs are shown in the Material and Methods section.

Figure 5A illustrates the fluorescence changes induced after target administration for the different hemagglutinin (H1, H2, H3 and H1s) cDNAs, and Figure 5B shows the results obtained for the neuraminidase (N1, N2 and N1s).

As illustrated in the figure, only the MBs designed to interact with their respective targets fluoresced upon target application. No cross-reactivity was observed between the hemagglutinin or neuraminidase MBs; they responded only to their respective targets. To demonstrate the reproducibility of the response, we printed 12 spots for each MB (Figure 5). The time course of the fluorescence change in the MBs is more clearly illustrated in Supplementary Video S4. The video also shows the selectivity of the MBs for their specific target. When the cDNA for N1 is applied, the MBs do not fluoresce since they are designed for N2. Only when N2 is applied the respective MB emit fluorescence (approximately halfway through the video sequence).

To determine the sensitivity of our microarray method, we conducted experiments aimed at identifying the smallest amount of sample that our MBs can detect (Figure 6).

Maintaining a constant concentration of 1 µM MB in the tridimensional silk microarray, we varied the concentration of the cDNA for the different genes of influenza. As illustrated in Figure 6, the minimum amount of cDNA detected under these conditions was 0.1 nM (100 p.m.). The dotted line shows 3 times the standard deviation from noise as a threshold to separate positive signals from the background. Similar values were obtained for all influenza genes explored, indicating that the sensitivity of the method was not dependent on the target to be identified but rather a property of the MB concentration and the optical properties of our silk microarray. The sensitivity of our system is higher than what has been reported for conventional microarray technology, with our microarray technology showing at least 5 times greater sensitivity (Thissen et al., 2014).