Massively Parallel Selection of NanoCluster Beacons

Yu-An Kuo, Cheulhee Jung, Yu-An Chen, Hung-Che Kuo, Oliver S. Zhao, Truong D. Nguyen, James R. Rybarski, Soonwoo Hong, Yuan-I Chen, Dennis C. Wylie, John A. Hawkins, Jada N. Walker, Samuel W. Shields, Jennifer S. Brodbelt, Jeffrey T. Petty, Ilya J. Finkelstein, and Hsin-Chih Yeh

Advanced Materials, Volume 34, Issue 41, e2204957 (2022)

DOI: 10.1002/adma.202204957

---

Table of Contents

• Abstract
• 1. Introduction
• 2. Results
• 3. Discussion
• 4. Conclusion
• 5. Experimental Section
• Acknowledgements

---

Abstract

NanoCluster Beacons (NCBs) are multicolor silver nanocluster probes whose fluorescence can be activated or tuned by a proximal DNA strand called the activator. While a single-nucleotide difference in a pair of activators can lead to drastically different activation outcomes, termed the polar opposite twins (POTs), it is difficult to discover new POT-NCBs using the conventional low-throughput characterization approaches. Here we report a high-throughput selection method that takes advantage of repurposed next-generation-sequencing (NGS) chips to screen the activation fluorescence of ~40,000 activator sequences. We find the nucleobases at positions 7-12 of the 18-nucleotide-long activator are critical to creating bright NCBs and positions 4-6 and 2-4 are hotspots to generate yellow-orange and red POTs, respectively. Based on these findings, we propose a “zipper bag model” that could explain how these hotspots lead to the creation of distinct silver cluster chromophores and alter the chromophore chemical yields. Combining high-throughput screening with machine learning algorithms, we establish a pipeline to design bright and multicolor NCBs in silico. Keywords: NanoCluster Beacons, silver nanoclusters, fluorescent nanomaterials, high-throughput screening, next-generation sequencing

Graphical Abstract

We screen ~40,000 distinct NanoCluster Beacons (NCBs) on repurposed next-generation sequencing chips and identify new NCBs that outperform the known best NCBs. We observe dramatic fluorescence changes by substituting a single nucleotide in an NCB, which could be explained by our proposed zipper-bag model. Combining the chip-based high-throughput screening with machine learning algorithms, we successfully design bright and multicolor NCBs in silico. —

1. Introduction

Activatable and multicolor fluorescent probes are indispensable tools in analytical chemistry and quantitative biology as they enable sensitive detection of analytes and diagnostic imaging of biomarkers in complex environments.[1] Whereas activatable probes have greatly simplified the assays by eliminating the need to remove unbound probes, the development of new activatable probes is largely constrained by the scarce activation mechanisms (e.g., FRET), the limited activation colors (e.g., existing FRET pairs) and the poor enhancement ratios (e.g., 10- to 60-fold for a typical molecular beacon).[2] NanoCluster Beacons (NCBs)[3] are a unique class of activatable probes as they provide a palette of activation colors from the same dark origin[4] (not via FRET) and achieve fluorescence enhancement ratios as high as 1,500[5] to 2,400-fold.[6] The core of an NCB is a few-atom silver nanocluster[7] (e.g., Ag8, Ag10 or Ag16) whose fluorescence can be tuned by its surrounding nucleobases.[7b, 7c, 8] To create an NCB, a dark AgNC is first synthesized in a C-rich DNA host (termed the NC probe), and a G-rich overhang (termed the activator) is brought into close proximity of the AgNC (via target-probe hybridization, Figure S1) to activate its fluorescence (Fig. 1A-B).[3-5, 8a, 8d] Being a low-cost probe that can be easily prepared in a single-pot reaction at room temperature,[7a] NCBs have been applied to the detection of nucleic acids,[8a, 8d, 9] proteins,[10] small molecules,[11] enzyme activities[8c] and cancer cells.[12]

Whereas new applications of NCBs are emerging across a broad range of disciplines, it is unclear what sequence features of the activators ultimately control the enhancement ratio and activation color of an NCB. To answer this fundamental question and to unleash the power of NCB in biosensing, we design and study “polar opposite twin” NCBs (hereafter, denoted as POT-NCBs). POTs are similar in appearance (i.e., sequence), but with very different personalities. In NCBs, POTs refer to a pair of NCBs that differ only by a single nucleotide in their activators, but have drastically distinct activation intensities or colors. Whereas POTs hold the key to understanding the NCB activation processes, there is no effective way to rapidly scan the vast activator sequence space and identify the most extreme POT-NCBs. Here we repurpose the next-generation sequencing (NGS) chips for high-throughput screening of fluorescent nanomaterials. In a single experiment, more than 104 activator mutations can be evaluated based on their capabilities in fluorescence activation of a common NC probe (C55 in Fig. 1A). Although the fluorescence properties of tens to hundreds of silver nanocluster species templated in short DNA strands can be studied in DNA microarrays[13] and robotic plates,[8b] less than 3,000 DNA hosts have been investigated as of today using these methods. While NGS chips are repurposed for studying protein-nucleic acid interactions,[14] they have never been used for study, selection, and optimization of fluorescent nanomaterials. By screening more than 40,000 activator sequences on three Illumina MiSeq chips, we not only discover new NCBs that are brighter than the known best NCBs (G15 and G12) but also identify the positions of nucleobases that are key to stabilizing bright AgNC chromophores (termed the critical zone). In the search for the most extreme POTs, the chip platform helps pinpoint the single-nucleotide substitution hotspots for generating yellow-orange and red POT-NCBs, reaching 31-fold and 9-fold differences in the enhancement ratios, respectively (563 vs. 18 for a pair of yellow-orange POTs and 285 vs. 32 for a pair of red POTs). Based on the findings of the critical zone for hosting bright chromophores (positions 7-12) and the hotspots for generating POTs (positions 4-6 for yellow-orange POTs and positions 2-4 for red POTs), we propose a “zipper bag model” that could explain how POT hotspots lead to the creation of distinct AgNC chromophores and alter the chromophore chemical yields. In addition, with proper selection of the sequence features, we build machine learning models that can design yellow-orange and red NCBs in silico. NCBs designed using these tools are 8.5 and 2 times more likely to be bright yellow-orange and red, respectively, as compared to the ones with random activator sequences. Our high-throughput screening and machine learning design pipeline is not only accelerating the discovery of new NCBs for diverse applications, but also providing insights into the chemical yield and the emitter brightness controlled by the sequence features. —

2. Results

2.1. High-throughput selection of red and yellow-orange NCBs on NGS chips

In our first library design (library_1), the 18-nt-long canonical activator G15[3-4, 8a, 8d] (GGGTGG GGTGGG GTGGGG) was divided into three 6-nt-long segments, and each segment was separately randomized to create 3×46-2 = 12,286 activator mutations (two were G15 duplicates in the 3×46 combinations, Table S1 and Methods). This approach significantly reduced the library size from 418 (the entire ligand composition space) down to ~104 in a single experiment. After mixing with the fiducial markers (PhiX), the library sequences were immobilized, bridge amplified and sequenced on a Illumina MiSeq chip. As the sequencing-needed barcodes and adapters (i.e., SP2/barcode/P7 adapter, Table S1) could suppress or alter the NCB fluorescence (Figure S3), they were removed using a restriction enzyme, leaving behind 20-nt-long activators (the library) that were only 2 nt (a pair of CG dinucleotides) longer than the activators used in the traditional low-throughput test-tube selections[3-4, 8a, 8d] (Fig. 1C and Table S1). After enzymatic cleavage, the quality of the library was checked by staining the library with an Atto647N-labeled probe, before using the library for NCB selections (Figure S4). We first aimed to discover new activators that light up the common C55 NC probe (CCCCCTTAATCCCCC, which hosts a dark AgNC) more intensely than G15 in yellow-orange emission (within 570-640 nm). We also searched for activators that give distinct emission colors (e.g., red emission within 663-738 nm). The yellow-orange and the red emissions here were simply defined by the filter cubes (i.e., TRITC and Cy5) commonly used for fluorescence imaging. Our color channel definitions were different from those used by other AgNC researches,[15] where the color definitions were due to some structural factors. Once the complementary AT sequences on the C55 probe and the activators hybridized, AgNC emission developed.[4-5] In the traditional test-tube experiments, the NC probe-activator mixtures were heated to 90-95°C for a minute and gradually cooled down to room temperature to allow hybridzation.[3-4] On the MiSeq chip, a constant temperature of 40°C was maintained to enable hybridization (Figure S6; Methods). This relatively low hybridization temperature extended the life of the chip, allowing us to go through at least 20 rounds of activation experiments on a single chip (hybridized with C55 probes, washed and imaged, and then removed C55 probes using an alkaline solution) without showing any degradation, thus providing highly reproducible selection results (Spearman’s ρ = 0.93 ± 0.005 for the red NCBs and 0.88 ± 0.024 for the yellow-orange NCBs, Figure S12-S13). We emphasize that when using 40°C and 40 minutes for NCB hybridization in test tubes, the resulting fluorescence activation was indistinguishable from that carried out at 90 °C for one minute (Figure S16-S17 and Table S3-S4). After injecting the common C55 NC probes into the MiSeq chip, the chip was scanned using a wide-field fluorescence microscope equipped with a metal halide illuminator, a sCMOS camera and an xyz translation stage. The activated NCBs were sequentially imaged using the filter cubes designed for conventional red emitters (e.g., Cy5, Ex/Em: 620/60, 700/75 nm) and yellow-orange emitters (e.g., TRITC, Ex/Em: 535/50, 605/70 nm). These two filter cubes were selected due to their popularity in fluorescence imaging. A custom bioinformatics and imaging processing pipeline[14d] was employed to identify activator sequences behind each activated NCB spot (Fig. 1C). By ranking the activators based on their median intensity (baseline corrected; Methods) on the chip surface (each activator had 457 ± 308 polonies on the MiSeq chip; Methods), we could clearly distinguish strong activators from weak activators in the two emission channels (Fig. 1C right). Compared to other high-throughput screening methods that rely on fluorescence from single molecules for characaterization,[16] photobleaching is not a severe issue in our approach, as each activator polony contains hundreds of activated NCBs. Polony is a contraction of “polymerase colony,” which is a small colony of DNA amplified from a library sequence. By employing an auto-scan algorithm and shutter control (Methods), the excitation dose to each polony is precisely regulated, avoiding any uneven photobleaching and ensuring consistent imaging results. By acquiring a fluorescence image every 200 ms, an intensity time trace of each polony is obtained, which can be fitted with a single-exponential decay. After one second of wide-field illumination (~10 W/cm2), polony intensity decreases by ~20% at most (Figure S10). Based on the chip screening results, we randomly select 20 activators from the 827 activators that are brighter than G12 (ranked 828th) and 20 activators from the 11,458 activators that are darker than G12 for further investigation in test tubes by fluorometry. Using the G12 activator (ATCCGGGGTGGGGTGGGG) as the standard for red NCB comparison, the MiSeq chip screening results were 85% accurate in both true positive and true negative selections (brighter or darker than G12 in chip screening and confirmed in fluorometry, Fig. 1D, Figure S14-S15, and Table S2). In particular, rAct1 (TCCATTGGTGGGGTGGGG) was found 2.94-fold brighter than G12 in the red channel (enhancement ratios were 1,292 vs. 439, Figure 1E, Figure S14 and Table S2). We then evaluated the “spectral purity” of selected NCBs by computing the intensity ratio between the two colors. For the selected bright red NCBs, 70% of them were spectrally pure in red emission as their integrated intensity under the red window (663-738 nm) was at least 3.5-fold larger than that under the yellow-orange window (570-640 nm). When comparing the chip screening results with the test-tube results, a coefficient of determination of 0.50 was obtained (Fig. 1D). We noticed the intensity differences found on MiSeq chips were substantially smaller than those found in test tubes (e.g., 1.20-fold red-emission difference was found between rAct1 and G12 NCBs on a MiSeq chip, but it became 2.94-fold in test tubes). The underestimation was attributed to the relatively higher fluorescence background on MiSeq chips and the variations in polony numbers among the library sequences (for instance, ~240 activators had less than 30 polonies in library_1). Out of the twelve yellow-orange candidates, we found yAct1 (TTGGTGGGTGGGGTGGGG) 2.03-fold brighter than G15 (the standard for yellow-orange NCB comparison) in activating C55 in the yellow-orange channel (enhancement ratios were 1125 vs. 553, Figure S17 and Table S4). We emphasize that the small-scale investigations carried out in test tubes, such as single-nucleotide substitutions from G15 at each nucleobase position (totally 3×18 = 54 variants), would lead to an activator that is only 44% brighter than G15 (Figure S18).

2.2. Identification of critical nucleobases in stabilizing bright AgNC chromophores

The screening results from library_1 (Table 1) clearly indicated that segment_2 prefers to be conserved (GGTGGG) in order to maintain NCB brightness (Fig. 2B). In contrast, randomizing segment_1 still produced many bright red NCBs, especially when segment_1 became C-rich. The effect of segment_3 was diverse, suggesting an indirect activation role. The 6-segment interrogation further revealed that segment_22 (positions 10-12) is more important than segment_21 (positions 7-9) in creating bright red NCBs (Fig. 2C and Figure S20). Independent investigations using library_2 and library_3 (~28,000 frame-shifted activators, Table S1) also confirmed that the nucleobases in positions 10-12 are critical in C55 activation (Figure S20). These results indicated that bright AgNC chromophores are most likely “clamped” by the two strands at positions 7-12 (Fig. 2A), possibly forming silver-mediated pairs between the two strands.[17]

| | | |---|---| | Canonical | 5′ – GGGTGGGGTGGGGTGGGG – 3′ | | Segment_1, library_1 | **NNNNNN**GGTGGGGTGGGG | | Segment_11 | **NNN**TGGGGTGGGGTGGGG | | Segment_12 | GGG**NNN**GGTGGGGTGGGG | | Segment_2, library_1 | GGGTGG**NNNNNN**GTGGGG | | Segment_21 | GGGTGG**NNN**GGGGTGGGG | | Segment_22 | GGGTGGGGT**NNN**GTGGGG | | Segment_3, library_1 | GGGTGGGGTGGG**NNNNNN** | | Segment_31 | GGGTGGGGTGGG**NNN**GGG | | Segment_32 | GGGTGGGGTGGGGTG**NNN** |

Drawing from the Table 1 and Fig. 2 results, one possible design rule for creating bright red NCBs on C55 is to develop an activator with a C-rich segment_1, a GC-rich segment_21, a G-rich segment_22, and a TC-rich segment_3. Nevertheless, the activator CCCCCCGCGGGGTTTCCC (termed G5) actually had a low red enhancement ratio (39, as compared to 439 for G12; Figure S22 and Table S9). This result clearly indicated that segments do not work alone - cooperativities among the segments determine the activation color and intensity of an NCB. Whereas previous investigations showed that more guanines in the activator generally lead to brighter red emission,[3] our large-scale investigations revealed a different design rule - brighter red NCBs can be achieved with fewer numbers of guanines (e.g., 10G_5 in Figure S23 and Table S5). As the results from our high-throughput screening could not be easily transformed into simple design rules, we trained machine learning algorithms on the dataset and used them to design bright NCBs in silico.

2.3. Discovery of polar opposite twins using NGS chips

Taking advantage of the chip screening platform, we searched for POT-NCBs that have the most extreme color or intensity differences (Fig. 3A-C). Although NCBs were previously used for single-nucleotide polymorphism detection,[4, 8a] only tens of activators were tested, providing little information on the rules to design POTs. In contrast, library_1 alone contained more than 110,000 pairs of twin NCBs (Figure S24), where the top 2,000 pairs were readily candidates for POTs. Upon examining these 2,000 pairs of twin NCBs (Fig. 3D), it was clear that the nucleobases in positions 4-6 are critical for creating yellow-orange POTs (e.g., bright (x-axis)→dark (y-axis) conversion by G/T→C/A substitution at position 5 and G→C substitution at positions 4 and 6), while the positions 2-4 are critical for creating red POTs (bright→dark conversion by C→ATG substitution). Fifteen top POT pairs were further investigated in test tubes. The most extreme yellow-orange and red POTs had 31-fold (yPOT5-yPOT6 with G→C substitution at position 5) and 9-fold (rPOT5-rPOT6 with C→T substitution at position 4) differences in their enhancement ratios, respectively (Fig. 3E). For the ease of comparison, we termed the difference in the enhancement ratios in a pair of POT the “POT difference ratio” (Figure S25-S26 and Table S7-S8), where the pair with the largest POT difference ratio is the most extreme POT.

The POT difference ratio reflected the sample brightness difference at the ensemble level, which is equal to the product of “chromophore chemical yield ratio” and “single-emitter brightness ratio”. Using fluorescence correlation spectroscopy (FCS),[18] we found the chromophore chemical yield of rPOT5 NCB 5.46-fold higher than that of rPOT6 NCB (24.41% vs. 4.47%), and the single-emitter brightness of rPOT5 1.64-fold higher than that of rPOT6 (5.67 kHz vs. 3.45 kHz, Figure S27). The product of the 5.46× chromophore chemical yield ratio and the 1.64× single-emitter brightness ratio was indeed the 9× POT difference ratio measured by fluorometry. Similarly, the chromophore chemical yield of yPOT5 NCB was 16.33-fold higher than that of yPOT6 NCB (25.80% vs. 1.58%) and the single-emitter brightness of yPOT5 NCB was 2.19-fold higher than that of yPOT6 NCB (7.04 kHz vs. 3.22 kHz, Figure S28). The product of the 16.33× chromophore chemical yield ratio and the 2.19× single-emitter brightness ratio was close to the 31× POT difference ratio measured at the ensemble level. Our investigation of POTs led to two important findings. First, the red AgNCs in rPOT5 and rPOT6 were actually different species as their excitation peak wavelengths (610 vs. 605 nm, Figure S25A), absorption spectra (a clear peak at 610 nm in rPOT5 NCBs spectrum but no clear peak in rPOT6 NCBs spectrum, Figure S29C), and single-emitter brightness were all different. Second, the chemical yield of AgNC chromophores could be significantly altered by substituting single nucleobases at positions (2-6 in Fig. 3D) outside the critical zone (7-12 in Fig. 2A). Based on these findings, we proposed a zipper bag model that could explain the mechanism behind POT formation (Fig. 3A). —

3. Discussion

3.1. Investigation of the zipper bag model

In our zipper bag model, the bag is the critical zone (positions 7-12) that holds the AgNC chromophore while the zipper is the POT hotspot that seals the bag. A subtle change in the zipper can alter the sealing condition of the zipper bag, which perturbs the short-range ligand environment around the AgNC inside the bag and possibly changes its binding footprint with the bag (Fig. 3A). We have previously shown that by slightly shifting the position of the activator with respect to the NC probe, a new ligand environment can be created around the AgNC that alters its emission spectrum,[8a, 8d] and we believe such a nucleobase-AgNC interaction is within a short range (≤ 1 nm)[8a]. When studying AgNC structures using 193 nm activated-electron photodetachment mass spectrometry (a-EPD MS), we have previously found two Ag10 clusters can be completely distinct chromophores due to very different binding footprints in their DNA hosts.[19] The earlier structural studies by extended X-ray absorption fine structure (EXAFS) spectra complemented the a-EPD MS footprint results.[7d, 20] Here by using FCS, we further demonstrated that a change in zipper may result in not only a distinct AgNC chromophore in the bag but also a different chemical yield of the chromophore, thus providing a basis for POT formation. Why are the zipper locations different for the yPOT5-yPOT6 pair (at position 5) and the rPOT5-rPOT6 pair (at position 4)? One possibility is red AgNC chromophores have higher silver stoichiometries and larger footprints in the bag, pushing the zipper locations (positions 2-4) further away from the bag (positions 7-12, Figure S25B). Using electrospray ionization mass spectrometry (ESI-MS), Gwinn’s group has previously shown a general trend for yellow chromophores having a smaller core (Ag10-Ag11) while red chromophores having a larger core (Ag14-Ag16).[7c] According to their rod-shaped model,[7b] red chromophores are expected to have larger footprints in their DNA hosts. However, in Gwinn ‘ s experiments, their AgNC chromophores are most likely stabilized inside dimers of 10-mers, whose ligand environments are different from our activator/NC probe systems. To investigate the silver stoichiometries of the two major chromophores in our POT experiments, we purified yPOT5, yPOT6, rPOT5 and rPOT6 NCB samples using 20% native polyacrylamide gel electrophoresis (native PAGE) (Figure S30; Methods) and analyzed the purified samples by ESI-MS (Figure S31). Since our NCB constructs were much larger (45-nt long for the NC probe strand and 48-nt long for the activator strand) than the constructs used in previous studies involving ESI-MS analysis[7c, 21] (10- to 26-nt-long), extensive cationic metal adduction during the ESI process prevented us from deciphering the exact silver stoichiometries in purified NCB samples. To circumvent this issue, an ion-pairing reagent, octylamine, was added to the ESI samples to suppress salt adduct formation.[22] Although the number of adducts was reduced, the addition of octylamine destabilized the duplex NCB structures. This generated highly polydisperse MS spectra that only provided us with rough estimates of silver stoichiometries in NCB duplexes (Figure S31). Without knowing the exact silver stoichiometries and AgNC binding sites in a pair of POT, the proposed zipper bag model is just one of the many possibilities behind POT formation (see Discussion section in Supplementary Information). Our future work is to continue investigating binding footprints and silver stoichiometries in purified NCB samples.

3.2. In silico design of bright NCBs using chip screening results and machine learning models

Since the results from our high-throughput screening (Table 1) cannot be easily transformed into simple design rules, we take advantage of machine learning algorithms to classify NCBs and uncover sequence features that give bright NCBs. Machine learning approaches have previously helped identify sequence features in DNA hosts that preferentially stabilize bright AgNCs, establishing the first statistical model for rational design of AgNCs with desired colors.[8b, 15a] However, such a model was built upon the emission properties of ~2,000 AgNCs sandwiched between two identical strands (8-mers to 16-mers).[15b] Our AgNCs were different, as each of them was stabilized within an 18-nt long activator and a 15-nt long C55. Although both robotic-well-plate studies[15b] (2,000 short strands) and NGS screening (40,000 NCBs in this report) covered only a small fraction of the overall ligand composition spaces (10-mers and 18-mers, respectively), statistical models could be built based on these small fractions of data. We extracted sequence features (which include motifs and motif locations) only from the bright activators in library_1. As the ranking of activators in the yellow-orange channel was quite different from that in the red channel (Figure S32A), we performed separate trainings on yellow-orange and red NCBs. Following the approach proposed by Copp and Gwinn,[8b, 15a, 15b] the top 30% NCBs (3,600) were labeled as the “bright” class and the bottom 30% were labeled as the “dark” class. 339 and 567 features from the bright and dark classes in the yellow-orange channel (denoted as bright yellow-orange and dark yellow-orange features) and 402 and 1,164 features from the bright and dark classes in the red channel (denoted as bright red and dark red features) were separately identified by MERCI.[23] To decrease the chance of overfitting, we further narrowed down to a set of the most discriminative features with 61 bright yellow-orange, 121 dark yellow-orange, 103 bright red and 112 dark red features using Weka[24] (Table S14-S15; Methods). A feature vector was employed to describe the location of each activator in the high-dimensional space for classification model development (Fig. 4A).

A number of models were established for classifying the chip screening results, based on algorithms such as logistic regression (LR), linear discriminant analysis (LDA), decision tree (DT), AdaBoost (ADA), and support vector machines (SVM) (Table S13). All models were built on “bright yellow-orange vs. dark yellow-orange” or “bright red vs. dark red” classification. To evaluate the model performance, we defined the accuracy of the model (Acc) as (TB + TD)/(TB + FB + TD + FD), where TB is the number of true predictions that the model makes for bright activators, TD and FD are the numbers of true and false dark predictions, and FB is the number of false bright predictions. In other words, Acc represented the fraction of test sequences that the model correctly identifies as “bright” or “dark” activators. We found the model built on LR has the best performance, achieving an average accuracy of 0.89 and 0.87 in the yellow-orange and red emission classification, respectively. In our model development process, the categorized dataset was divided into a training set (80% of the selected sequences) and a test set (20% of the selected sequences). The selection process iterated 5 times while rotating the training set and the test set, resulting in a 5-fold cross-validation (CV) that guarantees the model consistency (Figure S32B-C). As expected, the propensity of an activator to be “bright” is not only determined by having “bright” motifs within the activator but also by positioning these motifs at proper locations. Separately, based on the most discriminative features identified by Weka,[24] 1,000 bright yellow-orange and 1,000 bright red activator candidates were designed in silico (Figure S33; Methods). In the high-dimensional space, we employed the minimal “edit distance”[15b, 25] to identify the “closest” activator sequence in the library_1 dataset for each of the candidates. Consequently, when the closest library sequence was not among the top 200 bright activators screened, the candidate was discarded. Besides, when the candidate sequence had less than 3 or more than 5 single-base mutations, they were also discarded to restrict our search space. After going through these candidate refining steps, we were down to 100 bright yellow-orange and 54 bright red candidates. Among these candidates, the LR model classified 85 and 41 of them as bright yellow-orange and red activators, respectively. From these most promising candidates, 20 were randomly selected, synthesized, hybridized with the C55 probes in test tubes, and measured by a fluorometers. Compared to the randomly generated activator sequences (Figure S34 and Table S9), our design and classification pipeline produced new activators that were 8.5 times (85% vs. 10%) and 1.9 times (75% vs. 40%) more likely to be bright yellow-orange and red, respectively (Figure S35-S36 and Table S10-S11). Besides, the average enhancement ratio of in silico designed yellow-orange and red activators were 22 times and twice higher than that of random sequences, respectively (Table S9). Moreover, while all random sequences gave more or less red emission, our pipeline successfully produced NCBs with yellow-orange emission (Fig. 4B). In particular, among all bright candidates tested, we identified a new yellow-orange activator (yPred1: GTGTTGGGTGGTCGGGGG, with only 12 guanines) and a new red activator (rPred9: ATCCCTCGGGGAGGGGGC, with only 9 guanines) that were 2.13-fold and 1.30-fold brighter than the gold standards G15 and G12 in activating the C55 NC probe in test tubes, respectively (Fig. 4C, Table S10-S11). —

4. Conclusion

We have performed high-throughput screening on more than 40,000 activators using repurposed NGS chips. Not only did we discover new NCBs (yAct1 and rAct1) that are 2-3 times brighter than the known best NCBs (G15 and G12, Fig. 1), but we also identified a critical zone in the activator (positions 7-12) that stabilizes bright AgNC chromophores (Fig. 2). In the search for the most extreme POT-NCBs, the chip platform helped identify a red pair (rPOT5-rPOT6) and a yellow-orange pair (yPOT5-yPOT6) with POT difference ratios as high as 9 and 31, respectively (Fig. 3). By probing the NCBs at the near single-molecule level, we confirmed the observed brightness difference at the ensemble level is attributed to the differences in the single-emitter brightness and the chromophore chemical yield (Figure S27-S28). Based on the findings of the critical zone (positions 7-12) and the POT hotspots (positions 2-4 for the red POTs and positions 4-6 for the yellow-orange POTs), we proposed a zipper bag model that could explain how POT hotspots lead to the creation of distinct AgNC chromophores and alter the chromophore chemical yields (Fig. 3). We emphasize that our zipper bag model is only one of many possibilities. Further investigation on silver stoichiometries and AgNC footprints in purified NCBs using ESI-MS and a-EPD MS[19] is necessary to prove the zipper bag model. As the results from high-throughput screening could not be easily converted into simple rules for designing bright NCBs, we employed machine learning algorithms to classify the screening results and built a pipeline to design bright NCBs in silico. Forty new NCBs were generated by the pipeline, clearly showing two designated color bands (Fig. 4). We also found brighter NCBs could be achieved with fewer numbers of guanine bases in the activators. Although the overall sequence space investigated in this report (~40,000 variants) represents a small fraction of the entire 18-mer space (418 variants), our NGS platform has the scalability to broaden the search space by another one to three orders of magnitudes (by using a NextSeq or NovaSeq chip), thus leading to a more expansive machine learning model in the future. Our chip screening platform can facilitate the development of new chemical sensors based on DNA-templated AgNCs[26] or the study of other metal nanoclusters templated in DNA.[27] The chemical repertoire of natural nucleic acids can be expanded by connecting functional moieties to the alkyne-modified C5 site on dU[28], thus allowing us to study the interactions between a wide variety of functional moieties and AgNCs on NGS chips. The discoveries of critical zones and interaction hotspots lead to the zipper bag model, which provides us with a picture how surrounding nucleobases can interact with the AgNCs encapsulated in NCBs and control their chemical yields and emission properties. Moreover, the knowledge of POTs can be directly applied to design new sensors for single-nucleotide polymorphism detection. To our knowledge, this article is the first report that NGS chips are repurposed for high-throughput screening of fluorescent nanomaterials. Our high-throughput screening and machine-learning-based design pipeline is not only accelerating the discovery of new NCBs for diverse applications, but also providing insights into the chemical yield and the emitter brightness controlled by the sequence features. We anticipate new NCB-based sensors and new fluorescence barcodes will soon be developed based on the design strategy that we lay out in this report. —

5. Experimental Section

Statistical analysis:

Fluorescence images collected for high-throughput screening were first aligned using CHAMP program[14d] and alignment results were further analyzed using customized Python script. Fluorimetry results were analyzed by customized Python script. POT results were expressed as mean ± standard deviation (SD) with 3 repeats. The statistical significance relating to NGS chip imaging results was performed by Mann-Whitney U-test. Differences were ranked significant when P < 0.05, **P < 0.01, **P < 0.001.

---

Acknowledgements

This work was supported by the Welch Foundation to J.S.B. (F-1155), H.-C.Y. (F-1833), and I.J.F (F-1808), the Texas 4000 to H.-C.Y., National Institutes of Health to H.-C.Y. and I.J.F. (GM129617), and National Science Foundation to H.-C.Y. (CBET2041345) and J.S.B. (CBET2029266), and a UT-Austin Catalyst Grant to I.J.F.

Data availability

CHAMP program is available at https://github.com/finkelsteinlab/champ. All other relevant raw/analyzed data are available from the corresponding authors upon reasonable request.

---

References

1. Kobayashi H, Ogawa M, Alford R, Choyke PL, Urano Y. New strategies for fluorescent probe design in medical diagnostic imaging. Chem Rev. 2010;110(5):2620-2640. doi: 10.1021/cr900263j
2. Chinen AB, Guan CM, Ferrer JR, Barnaby SN, Merkel TJ, Mirkin CA. Nanoparticle probes for the detection of cancer biomarkers, cells, and tissues by fluorescence. Chem Rev. 2015;115(19):10530-10574. doi: 10.1021/acs.chemrev.5b00321
3. Rajendran M, Ellington AD. Selecting fluorescent aptamer beacons that light up in the presence of zinc. Nucleic Acids Res. 2003;31(19):5700-5713. doi: 10.1093/nar/gkg764
4. Yeh HC, Sharma J, Han JJ, Martinez JS, Werner JH. A DNA-silver nanocluster probe that fluoresces upon hybridization. Nano Letters. 2010;10(8):3106-3110. doi: 10.1021/nl101773c
5. Yeh H-C, Sharma J, Han JJ, Martinez JS, Werner JH. Fluorescent nanocluster beacons: bright, stable, and activatable probes for labeling and sensing. IEEE Nanotechnology Magazine. 2011;5(1):28-33.
6. Obliosca JM, Babin MC, Liu C, Liu YL, Chen YA, Batson RA, Ganguly M, Petty JT, Yeh HC. A complementary palette of NanoCluster Beacons. ACS Nano. 2014;8(10):10150-10160. doi: 10.1021/nn505338e
7. Ang YS, Woon WWE, Yung LYL. Rational design of hybridization chain reaction monomers for robust signal amplification. Nucleic Acids Research. 2018;46(13):6974-6983. doi: 10.1093/nar/gky521
8. (a) Obliosca JM, Liu C, Yeh HC. Fluorescent silver nanoclusters as DNA probes. Nanoscale. 2013;5(18):8443-8461. doi: 10.1039/c3nr01601c (b) Schultz D, Gardner K, Oemrawsingh SS, Markesevic N, Olsson K, Debord M, Bouwmeester D, Gwinn E. Evidence for rod-shaped DNA-stabilized silver nanocluster emitters. Adv Mater. 2013;25(20):2797-2803. doi: 10.1002/adma.201204624 (c) Copp SM, Schultz D, Swasey S, Pavlovich J, Debord M, Chiu A, Olsson K, Gwinn E. Magic numbers in DNA-stabilized fluorescent silver clusters lead to magic colors. J Phys Chem Lett. 2014;5(6):959-963. doi: 10.1021/jz500146q (d) Petty JT, Sergev OO, Ganguly M, Rankine IJ, Chevrier DM, Zhang P. A segregated, partially oxidized, and compact Ag10 cluster within an encapsulating DNA host. J Am Chem Soc. 2016;138(10):3469-3477. doi: 10.1021/jacs.5b13124 (e) Huard DJE, Demissie A, Kim D, Lewis D, Dickson RM, Petty JT, Lieberman RL. Atomic structure of a fluorescent Ag8 cluster templated by a multistranded DNA scaffold. J Am Chem Soc. 2019;141(29):11465-11470. doi: 10.1021/jacs.8b12203 (f) Cerretani C, Kanazawa H, Vosch T, Kondo J. Crystal structure of a NIR-emitting DNA-stabilized Ag16 nanocluster. Angew Chem Int Ed Engl. 2019. doi: 10.1002/anie.201906766
9. Yeh HC, Sharma J, Shih Ie M, Vu DM, Martinez JS, Werner JH. A fluorescence light-up Ag nanocluster probe that discriminates single-nucleotide variants by emission color. Journal of the American Chemical Society. 2012;134(28):11550-11558. doi: 10.1021/ja3024737
10. Li JJ, Zhong XQ, Zhang HQ, Le XC, Zhu JJ. Binding-induced fluorescence turn-on assay using aptamer-functionalized silver nanocluster DNA probes. Analytical Chemistry. 2012;84(12):5170-5174. doi: 10.1021/ac3006268
11. Zhang M, Guo SM, Li YR, Zuo P, Ye BC. A label-free fluorescent molecular beacon based on DNA-templated silver nanoclusters for detection of adenosine and adenosine deaminase. Chem Commun (Camb). 2012;48(44):5488-5490. doi: 10.1039/c2cc31626a
12. Yin J, He X, Wang K, Xu F, Shangguan J, He D, Shi H. Label-free and turn-on aptamer strategy for cancer cells detection based on a DNA-silver nanocluster fluorescence upon recognition-induced hybridization. Analytical Chemistry. 2013;85(24):12011-12019. doi: 10.1021/ac402989u
13. Richards CI, Choi S, Hsiang J-C, Antoku Y, Vosch T, Bongiorno A, Tzeng Y-L, Dickson RM. Oligonucleotide-stabilized Ag nanocluster fluorophores. Journal of the American Chemical Society. 2008;130(15):5038-5039. doi: 10.1021/ja8005644
14. (a) Nutiu R, Friedman RC, Luo S, Khrebtukova I, Silva D, Li R, Zhang L, Schroth GP, Burge CB. Direct measurement of DNA affinity landscapes on a high-throughput sequencing instrument. Nat Biotechnol. 2011;29(7):659-664. doi: 10.1038/nbt.1882 (b) Tome JM, Ozer A, Pagano JM, Gheba D, Schroth GP, Lis JT. Comprehensive analysis of RNA-protein interactions by high-throughput sequencing-RNA affinity profiling. Nat Methods. 2014;11(6):683-688. doi: 10.1038/nmeth.2970 (c) Buenrostro JD, Araya CL, Chircus LM, Layton CJ, Chang HY, Snyder MP, Greenleaf WJ. Quantitative analysis of RNA-protein interactions on a massively parallel array. Nat Biotechnol. 2014;32(6):562-568. doi: 10.1038/nbt.2880 (d) Jung C, Hawkins JA, Jones SK Jr., Xiao Y, Rybarski JR, Dillard KE, Hussmann J, Saifuddin FA, Savran CA, Ellington AD, Ke A, Press WH, Finkelstein IJ. Massively parallel biophysical analysis of CRISPR-Cas complexes on next generation sequencing chips. Cell. 2017;170(1):35-47. doi: 10.1016/j.cell.2017.05.044 (e) Layton CJ, McMahon PL, Greenleaf WJ. Large-scale, quantitative protein assays on a high-throughput DNA sequencing chip. Mol Cell. 2019;73(5):1075-1082. doi: 10.1016/j.molcel.2019.02.019
15. (a) Copp SM, Bogdanov P, Debord M, Singh A, Gwinn E. Base motif recognition and design of DNA templates for fluorescent silver clusters by machine learning. Advanced Materials. 2014;26(33):5839-5845. doi: 10.1002/adma.201401402 (b) Copp SM, Gorovits A, Swasey SM, Gudibandi S, Bogdanov P, Gwinn EG. Fluorescence color by data-driven design of genomic silver clusters. ACS Nano. 2018. doi: 10.1021/acsnano.8b03404 (c) Copp SM, Swasey SM, Gorovits A, Bogdanov P, Gwinn EG. Fluorescence kinetics of silver clusters in DNA. Chemistry of Materials. 2019;32(1):430-437. (d) Gonzàlez-Rosell A, Cerretani C, Mastracco P, Vosch T, Copp SM. Structure and luminescence of DNA-templated silver clusters. Nanoscale Advances. 2021;3(5):1230-1260. doi: 10.1039/d0na01005g
16. Hayward SL, Lund PE, Kang Q, Johnson-Buck A, Tewari M, Walter NG. Ultraspecific and amplification-free quantification of mutant DNA by single-molecule kinetic fingerprinting. J Am Chem Soc. 2018;140(37):11755-11762. doi: 10.1021/jacs.8b06685
17. (a) Swasey SM, Leal LE, Lopez-Acevedo O, Pavlovich J, Gwinn EG. Silver (I) as DNA glue: Ag+-mediated guanine pairing revealed by removing Watson-Crick constraints. Sci Rep. 2015;5:10163. doi: 10.1038/srep10163 (b) Kondo J, Tada Y, Dairaku T, Hattori Y, Saneyoshi H, Ono A, Tanaka Y. A metallo-DNA nanowire with uninterrupted one-dimensional silver array. Nat Chem. 2017;9(10):956-960. doi: 10.1038/nchem.2808
18. (a) Sharma J, Yeh HC, Yoo H, Werner JH, Martinez JS. A complementary palette of fluorescent silver nanoclusters. Chemical Communications. 2010;46(19):3280-3282. doi: 10.1039/b927268b (b) Yeh H-C, Sharma J, Yoo H, Martinez JS, Werner JH. The stable luminescent silver nanodot: Ag31. Proc SPIE. 2010;7576.
19. Blevins MS, Kim D, Crittenden CM, Hong S, Yeh HC, Petty JT, Brodbelt JS. Ultraviolet photodissociation mass spectrometry of the complexes of silver and DNA. ACS Nano. 2019;13(12):14070-14079. doi: 10.1021/acsnano.9b06470
20. Petty JT, Ganguly M, Rankine IJ, Chevrier DM, Zhang P. A DNA-encapsulated and fluorescent Ag10-11 cluster. The Journal of Physical Chemistry C. 2017;121(27):14936-14945.
21. (a) O’Neill PR, Velazquez LR, Dunn DG, Gwinn EG, Fygenson DK. Hairpins with poly-C loops stabilize four types of fluorescent Agn:DNA. The Journal of Physical Chemistry C. 2009;113(12):4229-4233. (b) Schultz D, Gwinn EG. Silver atom and strand numbers in fluorescent and dark Ag:DNAs. Chem Commun (Camb). 2012;48(44):5748-5750. doi: 10.1039/c2cc17675k
22. (a) McGinnis AC, Grubb EC, Bartlett MG. Systematic optimization of ion-pairing agents and hexafluoroisopropanol for enhanced electrospray ionization mass spectrometry of oligonucleotides. Rapid Commun Mass Spectrom. 2013;27(24):2655-2664. doi: 10.1002/rcm.6733 (b) Scalabrin M, Palumbo M, Richter SN. Highly improved electrospray ionization-mass spectrometry detection of G-quadruplex-folded oligonucleotides and their complexes with small molecules. Anal Chem. 2017;89(16):8632-8637. doi: 10.1021/acs.analchem.7b01282 (c) Sutton JM, Bartlett MG. Improving the ESI-MS detection of phosphate tagged oligonucleotides and analogs: The role of ion pairing agents and ESI additives. Rapid Commun Mass Spectrom. 2020;34(7):e8696. doi: 10.1002/rcm.8696 (d) Sutton JM, El Zahar NM, Bartlett MG. Investigation into the ESI-MS behavior of phosphodiester oligonucleotides using experimental design. J Am Soc Mass Spectrom. 2021;32(2):497-506. doi: 10.1021/jasms.0c00380
23. Vens C, Rosso MN, Danchin EG. Identifying discriminative classification-based motifs in biological sequences. Bioinformatics. 2011;27(9):1231-1238. doi: 10.1093/bioinformatics/btr110
24. Hall M, Frank E, Holmes G, Pfahringer B, Reutemann P, Witten IH. The WEKA data mining software: an update. ACM SIGKDD Explorations Newsletter. 2009;11(1):10-18.
25. Ristad ES, Yianilos PN. Learning string-edit distance. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1998;20(5):522-532.
26. Del Bonis-O’Donnell JT, Thakrar A, Hirschberg JW, Vong D, Queenan BN, Fygenson DK, Pennathur S. Dual near-infrared two-photon microscopy for deep-tissue dopamine nanosensor imaging. ACS Chem Neurosci. 2018;9(4):849-857. doi: 10.1021/acschemneuro.7b00444
27. (a) Chen WY, Lan GY, Chang HT. Use of fluorescent DNA-templated gold/silver nanoclusters for the detection of sulfide ions. Anal Chem. 2011;83(24):9450-9455. doi: 10.1021/ac202162u (b) Jia X, Li J, Han L, Ren J, Yang X, Wang E. DNA-hosted copper nanoclusters for fluorescent identification of single nucleotide polymorphisms. ACS Nano. 2012;6(4):3311-3317. doi: 10.1021/nn3002455 (c) Chen J, Ji X, Tinnefeld P, He Z. Multifunctional dumbbell-shaped DNA-templated selective formation of fluorescent copper nanoparticles and graphene oxide nanoassembly for ATP detection. ACS Appl Mater Interfaces. 2016;8(2):1786-1794. doi: 10.1021/acsami.5b09678
28. (a) Drees A, Fischer M. Click chemistry mediated coupling of biomolecules to abasic sites in DNA. International Journal of Molecular Sciences. 2021;22(17):9202. (b) Wu D, Feagin T, Mage P, Rangel A, Wan L, Li A, Coller J, Eisenstein M, Pitteri S, Soh HT. Templated synthesis of fluorescent nanoparticles from DNA-binding dyes. bioRxiv. 2020.

---

For the complete references list, please see the full text on PubMed Central.