A digital single-molecule nanopillar SERS platform for predicting and monitoring immune toxicities in immunotherapy

Abstract

The introduction of immune checkpoint inhibitors has demonstrated significant improvements in survival for subsets of cancer patients. However, they carry significant and sometimes life-threatening toxicities. Prompt prediction and monitoring of immune toxicities have the potential to maximise the benefits of immune checkpoint therapy. Herein, we develop a digital nanopillar SERS platform that achieves real-time single cytokine counting and enables dynamic tracking of immune toxicities in cancer patients receiving immune checkpoint inhibitor treatment - broader applications are anticipated in other disease indications. By analysing four prospective cytokine biomarkers that initiate inflammatory responses, the digital nanopillar SERS assay achieves both highly specific and highly sensitive cytokine detection down to attomolar level. Significantly, we report the capability of the assay to longitudinally monitor 10 melanoma patients during immune inhibitor blockade treatment. Here, we show that elevated cytokine concentrations predict for higher risk of developing severe immune toxicities in our pilot cohort of patients.

Fig. 1. Digital single-molecule nanopillar surface-enhanced Raman scattering (SERS) platform for parallel counting of four types of cytokines.

SEM images of a pillar array side view, b nanoboxes, and c a single nanobox on the top of a pillar; d SERS spectra of nanoboxes conjugated with 5,5-dithiobis (2-nitrobenzoic acid) (DTNB), 4-mercaptobenzoic acid (MBA), 2,3,5,6-tetrafluoro-4-mercaptobenzoic acid (TFMBA), or 2‐mercapto‐4‐methyl‐5‐thiazoleacetic acid (MMTAA) Raman reporters; e workflow for multiplex counting of cytokines, including fibroblast growth factor 2 (FGF-2), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), and fractalkine (CX3CL1). Data from one independent experiment.

Fig. 2. Demonstration of the single-particle SERS activity of DTNB-labelled nanoboxes.

a SEM image and b corresponding SERS mapping image of DTNB-labelled nanoboxes on a silicon substrate; c representative SERS spectra of numbered locations indicated in a and b. The red dotted line shows the characteristic peak at 1330 cm⁻¹ from DTNB. Data from one independent experiment. Source data are provided in the Source Data file.

Fig. 3. Study of confocal height on Raman signal intensity.

SERS mapping of FGF-2 SERS nanotags on the silicon substrate with changing confocal height of a 0 nm, b 500 nm, c 1000 nm, and d 1500 nm; selected Raman spectra obtained from e red circles and f blue circles of SERS images. Red dotted lines in e and f indicate peak signal at 1330 cm⁻¹ from DTNB. Data from one independent experiment. Source data are provided in the Source Data file.

Fig. 4. Specificity of digital nanopillar SERS platform for FGF-2 cytokine detection.

Representative confocal SERS images in the presence of a target FGF-2 (1031 aM), and negative controls with non-target controls b G-CSF (1031 aM), c GM-CSF (1031 aM), d CX3CL1 (1031 aM), and e PBS. The median (interquartile range) of active pillars per scanning image for FGF-2, G-CSF, GM-CSF, CX3CL1, and PBS was 72 (63.5–76.75), 1.5 (1.5–2), 2 (1–4), 0.5 (0–1.25), and 1 (1–1.75), respectively. Data from one independent experiment.

Fig. 5. Specificity of the digital nanopillar SERS platform for FGF-2 cytokine detection.

Representative SEM images of pillar array incubated with FGF-2 SERS nanotags in the presence of a, b FGF-2 (1031 aM), c G-CSF (1031 aM), d GM-CSF (1031 aM), e CX3CL1 (1031 aM), and f PBS. The red circles highlight the existence of SERS nanotags. Panel b is the magnified SEM image of the red-highlighted section in a. It is noted that nanofabrication debris on the sidewall of the pillars can also be seen. Data from one independent experiment.

Fig. 6. Sensitivity for the simultaneous detection of four cytokines.

Representative confocal SERS images of fibroblast growth factor 2 (FGF-2), granulocyte colony-stimulating factor (GM-CSF), granulocyte colony-stimulating factor (G-CSF), and fractalkine (CX3CL1) with the concentration of a 2.6 aM, b 26 aM, c 260 aM, d 1031 aM. Colour scale bars indicate Raman intensities from 5,5-dithiobis (2-nitrobenzoic acid) (DTNB), 4-mercaptobenzoic acid (MBA), 2,3,5,6-tetrafluoro-4-mercaptobenzoic acid (TFMBA), or 2‐mercapto‐4‐methyl‐5‐thiazoleacetic acid (MMTAA). The median (interquartile range) of active pillars per scanning image of FGF-2, G-CSF, GM-CSF, CX3CL1 for 2.6 aM: 3 (1.5–3), 1 (1–2), 2 (1–3), 2 (1–3); 26 aM: 8 (5.5–10), 10 (9–13), 7 (6–10), 8 (6–10); 260 aM: 40 (36–48), 40 (35–52), 39 (35–50), 37 (36–49); and 1031 aM: 79 (61.5–97), 78 (72–87.5), 88 (68.5–97), 79 (64–95), respectively. Data represents one experiment from three independent tests.

Fig. 7. Digital nanopillar SERS assay for monitoring melanoma patients during immune checkpoint therapy.

For Patient 1 who developed severe irAEs, SERS images for cytokine detection on a day 7, b day 21, c day 42, d cytokine concentration graph for fibroblast growth factor 2 (FGF-2), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), and fractalkine (CX3CL1). The two shorter horizontal lines denote the interquartile ranges (25th and 75th percentile) and the longer horizontal lines in between denote the median (50th percentile), and e LDA analysis, respectively. For Patient 6 who developed mild irAEs, SERS images for cytokine detection on f day 0, g day 21, h day 42, i four cytokine concentration graph, the two shorter horizontal lines denote the interquartile ranges (25th and 75th percentile) and the longer horizontal lines in between denote the median (50th percentile), and j LDA analysis, respectively. IPI ipilimumab, PEMBRO pembrolizumab; G3 grade 3, G2 grade 2; SD stable disease, PR partial response. For Patient 1, the median (interquartile range) of active pillars per scanning image of FGF-2, G-CSF, GM-CSF, CX3CL1 on day 7: 14 (11–22.5), 23 (21, 29), 12 (7.5–18), 17 (9–25.5); day 21: 30 (19–37.5), 33 (19–41), 26 (17.5–36.5), 29 (21–43); and day 42: 33 (16.5–58.5), 76 (64–128.5), 25 (14–39.5), 48 (26.5–73.5), respectively. For Patient 6, the median (interquartile range) of active pillars per scanning image of FGF-2, G-CSF, GM-CSF, CX3CL1 on day 0: 18 (16–23), 49 (31.5–56), 23 (17.5–28), 20 (14.5–27); day 21: 29 (24–33.5), 53 (46.5–70), 35 (25–46), 22 (19–29.5); and day 42: 13 (8–16.5), 44 (23.5–55.5), 10 (6.5–12.5), 30 (24–34.5), respectively. The data represented three technical replicates obtained from three chips. Nine images were acquired from each chip for cytokine counting. Statistical analysis was based on Kruskal–Wallis test followed by Dunn’s test to correct multiple comparisons (two-sided). Source data are provided in the Source Data file.