If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
University of Arizona Cancer Center, Tucson, Arizona, USADivision of Dermatology, University of Arizona College of Medicine, Tucson, Arizona, USASkin Cancer Institute, University of Arizona, Tucson, Arizona, USA
Department of Dermatology and Cutaneous Biology, Sydney Kimmel Cancer Center, Thomas Jefferson University, Philadelphia, Pennsylvania, USADepartment of Otolaryngology Head and Neck Surgery, Sydney Kimmel Cancer Center, Thomas Jefferson University, Philadelphia, Pennsylvania, USA
Understanding early tumorigenic events has obvious diagnostic and prognostic implications, particularly in the skin because this is the site associated with more cancer diagnoses in the United States than any other organ (
). Actinic keratoses (AKs) are dysplastic keratinocyte lesions characterized by irregularly shaped scaly papules frequently manifesting on sun-damaged skin and are considered precursor lesions that can develop into cutaneous squamous cell carcinoma (cSCC), the second most common keratinocyte cancer (
). Delineating the molecular events that characterize transition from normal skin to AK and on to established skin cancer has been the focus of a growing number of studies employing various technologies to identify markers of both initiation and progression of AK and cSCC (
Traditionally, cancer biology has focused on changes within the tumor compartment for diagnosis and staging. Recent genomic profiling of AK has been similar, as multiple studies have focused on DNA analysis to identify clonal chromosomal abnormalities and somatic mutations (
). This approach is in line with a mutation-centric view of cancer, supported by the clear requirement for a tumor to harbor multiple somatic DNA changes without consideration of the dermal compartment, populated predominantly by fibroblasts and, in the case of AK and cSCC, infiltrating immune cells. Because sun-damaged skin can harbor multiple so-called tumor driver-gene mutations caused by prolonged exposure to UVR (
), it becomes challenging to differentiate tumor from nontumor and tumor from precancerous lesions based on mutation burden alone.
As part of a pilot study to evaluate current technologies available to assess a large cohort of archivally matched patient samples, we employed the NanoString Digital Spatial Profiler to compare transcriptional change in sun-protected skin with different regions of matched AK from the six individuals. The NanoString Digital Spatial Profiler allows for the selection of tissue compartments of interest for RNA analysis using photocleavable RNA probes isolated based on antibody staining. In this study, we used a pan-cytokeratin antibody to differentiate the epidermal from dermal compartment and assayed over 18,000 RNA probes from multiple regions of interest comparing sun-protected skin (n = 5–6 from 6 individuals for a total of 34 regions of interest), the center of an AK lesion (n = 3 for a total of 18 regions of interest), and the edge of the AK lesion (n = 2–3 for a total of 17 regions of interest) representing sun-damaged skin, from six unrelated individuals (Figure 1a).
Figure 1Spatial profiling identifies the greatest transcriptional change in the cytokeratin negative compartment when comparing SP skin with AK. (a) The Digital Spatial Profiler identified ROI and highlighted examples of areas of patient tissue sampled for transcriptional profiling. (b) Differentially expressed genes identified using a pairwise FDR-adjusted P < 0.05 and an overall t test statistic of 0.05 highlight a greater number of DEGs (y-axis) in the CK− comparing SP normal skin with the Cen. Cen, actinic keratosis center; CK−, cytokeratin negative compartment; CK+, cytokeratin positive compartment; DEG, differentially expressed gene; Edge, actinic keratosis edge; FDR, false discovery rate; ROI, region of interest; SP, sun-protected; v, versus.
Analysis of these Digital Spatial Profiler data identified the expected changes in the epidermal compartment comparing sun-protected skin to AK, showing an upregulation of gene expression associated with P38 MAPK, TP53, extracellular signal–regulated kinase, and DNA damage–associated pathways, in line with previous studies (
) (Figure 2a and Supplementary Tables S1 and S2). A comparison of the dermal compartment revealed upregulation of pathways associated with matrix metalloproteinases, cell adhesion, and extracellular matrix remodeling, as well as multiple immune-related pathways including the innate immune system, major histocompatibility complex class II antigen presentation, and toll-like receptor signaling (Supplementary Tables S3 and S4). Differential expression analysis of batch-corrected, normalized read counts using a pairwise false discovery rate–adjusted P < 0.05 for those genes with an overall t test statistic of 0.05 revealed the greatest change in transcriptional activity in the dermal compartment when comparing sun-protected skin with the center of the AK lesion. The fewest transcriptional changes in our matched samples were evident when comparing the AK edge with the AK center, again with the greatest change being in the dermal compartment (Figure 1b and Supplementary Table S5). Interestingly, comparing sun-protected skin with AK edge revealed more change in the epidermal compartment. This suggests a progressive model where early transcriptional change from sun-protected skin to AK edge (assumed to be sun-damaged normal skin) is evident in the epidermal compartment, which leads to activation of multiple signaling cascades within the dermal compartment and significant changes to extracellular matrix remodeling, presumably with a major contribution from infiltration of immune cells (Figure 2b).
Figure 2Pathway changes in the progression from SP skin to AK. (a) Biplots showing CLR-transformed reads by comparing SP and AK samples. Individual samples are shown with three color and symbol combinations, and the mean values are represented by larger symbols on the PCA plane with the CLR standardized gene expression plotted. For CK-negative samples, the SP and AK edges cluster together, and for CK-positive samples, the AK edge and center cluster together relative to the SP samples. The bar charts help interpret the directionality of the projections, which can be inferred from the PCA plot but is simplified by the bar graphs. (b) Graphical representation of transcriptional change moving through SP skin to AK. Major pathways (identified using methodology described in
) changing compared with each of the three compartments are indicated within the arrows representing the overall extent of transcriptional change; green arrows represent transcriptional change within the CK positive compartment, whereas orange arrows represent the dermal (CK negative) compartment. AK, actinic keratosis; CK, cytokeratin; CLR, centered log-ratio; ECM, extracellular matrix; ERK, extracellular signal–regulated kinase; PCA, principal component analysis; SP, sun-protected normal skin.
These data are intriguing because earlier studies of AK have focused on the cell of origin, the keratinocyte, and although clear differences in transcription have been identified comparing AK with sun-protected or UV-exposed normal skin, fewer changes are observed comparing AK with cSCC (
), and a compelling hypothesis supported by the data presented in this study is that key molecular changes that influence the progression of AK are in fact found in the dermal compartment. Further work with a wider sample set and the inclusion of matched cSCC samples will be needed to test this hypothesis, and the Digital Spatial Profiler platform offers the ability to do so in archival, formalin-fixed samples. A recent study, combining single-cell and spatial transcriptomics to compare 10 squamous cell carcinoma samples with patient-matched normal skin, also included the separation of the dermal compartment and highlighted intercellular communication between the stroma and the leading edge of invasive squamous cell carcinoma (
). This observation was also noted in an earlier single-cell study of head and neck squamous cell carcinoma, where the leading edge of the tumor was highlighted as the most prognostic region (
). These studies support the idea that the dermal compartment may well hold key prognosticating information in the context of AK and cSCC.
Collectively, these data suggest that although the epidermal compartment harbors the cell of origin for squamous cell carcinoma and AK, the greatest change in transcriptional homeostasis in the progression from sun-protected skin to AK is observed in the dermal compartment. This may well be a fruitful avenue of investigation for delineating molecular markers of initiation and progression.
All participants provided written, informed consent, and this protocol was approved by the institutional review board at the University of Arizona (Protocol Title: Skin Cancer Prevention Program Biorepository; Protocol Number: 1200000229).
This research was funded by the National Cancer Institute, grant numbers P01CA229112 and T32CA078447. This project used the Genomics and Flow Cytometry Shared Resources at the Sidney Kimmel Cancer Center, supported by the National Cancer Institute, grant 5P30CA056036-17.
Author Contributions
Conceptualization: BL, CCL, ET, JP, LW, HHSC, APS; Data Curation: BL, CCL, ET, JP, LW, HHSC, APS; Formal Analysis: BL, CCL, ET, JP, LW, HHSC, APS; Funding Acquisition: BL, CCL, ET, JP, LW, HHSC, APS; Investigation: BL, CCL, ET, JP, LW, HHSC, APS; Methodology: BL, CCL, ET, JP, LW, HHSC, APS; Project Administration: BL, CCL, ET, JP, LW, HHSC, APS; Resources: BL, CCL, ET, JP, LW, HHSC, APS; Software: BL, CCL, ET, JP, LW, HHSC, APS; Supervision: BL, CCL, ET, JP, LW, HHSC, APS; Validation: BL, CCL, ET, JP, LW, HHSC, APS; Visualization: BL, CCL, ET, JP, LW, HHSC, APS; Writing - Original Draft Preparation: BL, CCL, ET, JP, LW, HHSC, APS; Writing - Review and Editing: BL, CCL, ET, JP, LW, HHSC, APS
Supplementary Materials and Methods
Patient details
The details about patients are as follows:
•
SAZT20099: male, 75 years old, white not Hispanic individual
•
SAZT20088: male, 67 years old, white not Hispanic individual
•
SAZT200086: male, 78 years old, white not Hispanic individual
•
SAZT200080: female, 61 years old, white not Hispanic individual
•
SAZT200015: female, 75 years old, white not Hispanic individual
•
SAZT200007: male, 75 years old, white not Hispanic individual
Formalin-fixed, paraffin-embedded tissues
All human tissue was sourced from the Skin Cancer Prevention Program Biorepository at the University of Arizona Cancer Center. The protocol has been approved by the institutional review board at the University of Arizona (Protocol Title: Skin Cancer Prevention Program Biorepository; Protocol Number: 1200000229). A set of matched sun-protected and actinic keratosis from six individuals were used for this study. Tissue blocks were sectioned with a 5-μm thickness and mounted in the center of the slide (Superfrost Plus Micro slides, VWR, Radnor, PA) within an area of 35.3 mm long by 14.1 mm wide to allow detection by the NanoString (Seattle, WA) GeoMx Digital Spatial Profiler (DSP) instrument.
NanoString GeoMx digital space profiling whole transcriptome atlas for next-generation sequencing
Sample preparation was performed following the Leica Biosystems Bond RX FFPE RNA Slide Preparation Protocol in the GeoMx NGS Slide Preparation User Manual (NanoString, MAN-10115-03 for software version 2.1). This assay relies on photocleavable RNA probes coupled to oligonucleotide tags that allow the selection of tissue compartments of interest for RNA analysis. Slides were baked at 60 °C for 60 minutes. Samples were loaded into Leica Bond Rx for deparaffinization and antigen retrieval with the following settings: Bake and Dewax preparation program, 150-μl buffer dispense volume, HIER 20 minutes with ER2 at 100 °C, and 1 μ/ml proteinase K for 15 minutes. Slides were briefly stored in 1× PBS. Excess PBS was removed from slides before the addition of 200-μl Whole Transcriptome Assay probe sets diluted 1:10 in Buffer R (NanoString), covered with Hybrislips (Grace Biolabs, 714022, Bend, OR), and incubated at 37 °C overnight in a hybridization chamber lined with Kimwipes moistened with 2× saline-sodium citrate (SSC). Unbound probes were removed by washing twice at 37 °C for 25 minutes in 50% formamide (Thermo Fisher Scientific, AM9324, Waltham, MA)/2× SSC, followed by two washes at room temperature in 2× SSC for 2 minutes. Slides were blocked with 200-μl Buffer W (NanoString) for 30 minutes followed by 1-hour incubation at room temperature in a humidity chamber with morphology marker solution comprising 0.66-μl pan-cytokeratin (Novus Biologicals, NBP2-33200AF488) and 10-μl Syto83 (Thermo Fisher Scientific, S11364) in Buffer W (NanoString) to 200 μl per slide. Slides were subsequently washed twice with 2× SSC before loading onto the NanoString GeoMx DSP instrument.
Slides were loaded and scanned on the GeoMx instrument according to the GeoMx DSP Instrument User Manual (NanoString, MAN-10116-03 for software version 2.1) with the following scan exposure times; 200 ms FITC/525 nm, 60 ms Cy3/568 nm, 250 ms Texas Red/615 nm, and 300 ms Cy5/666nm. A total of 12 polygonal regions of interest (ROIs) were selected per slide, with segmentation masks applied to each ROI to select pan-cytokeratin–positive and pan-cytokeratin–negative areas within each ROI. Channel thresholds were adjusted to maximize inclusion of signal areas and minimize inclusion of nonsignal areas with default settings for erode, N-dilate, hole size, and particle size. Nuclei numbers were calculated for each region and are presented in Supplementary Figure S1. Whole transcriptome assay probes were aspirated from each region by way of UV cleavage and deposited into individual wells of a 96-well plate.
Sequencing
Next-generation sequencing libraries were prepared according to the GeoMx NGS Readout Library Prep User Manual (NanoString, MAN-10117-03 for software version 2.1). DSP aspirates were dried overnight at room temperature before being resuspended in 10-μl diethyl pyrocarbonate-treated water for 10 minutes. The 96-well PCR plates were prepared by the addition of 2-μl PCR mix (NanoString), 4 μl of index primer mix (NanoString), and 4 μl of resuspended DSP aspirate. PCR products were generated with the following program: 37 °C for 30 minutes, 50 °C for 10 minutes, and 95 °C for 3 minutes, followed by 18 cycles of 95 °C for 15 seconds, 65 °C for 1 minute, and 68 °C for 30 seconds, with a final extension at 68 °C for 5 minutes and a hold at 4 °C. Indexed libraries were pooled into small and large pools based on ROI, with the small-pool ROI being approximately 4-fold smaller than the large-pool ROI. Pools were incubated with AMPure XP beads (Beckman Coulter, A63880, Brea, CA) at a 1.2× bead-to-sample ratio for 5 minutes. Beads were collected on a magnetic stand for 5 minutes, the supernatant was removed, and the beads were washed twice with 200-μl freshly prepared 80% ethanol in diethyl pyrocarbonate-treated water for 30 seconds each. Excess ethanol was allowed to evaporate, and the beads were resuspended in elution buffer (10 nM Tris-HCl pH 8, 0.05% Tween-20) (15 μl for large pool, 10 μl for small pool). Libraries were rebounded to AMPure (Omega Bio-tek, Norcross, GA) beads by the addition of Ampure Buffer at 1.2× the elution buffer volume, and the above process repeated, and the libraries finally eluted in 12 μl for the small pool and 16 μl for the large pool (with elution buffer).
Total target counts per DSP collection plate for sequencing were calculated from the total sampled areas (μm2) reported in the DSP-generated lab worksheet with the target sequencing depth being 100 counts/μm2. Each library was diluted to 4–10 nM and combined to reach the estimated counts/μm2 per library in the final pool. All sequencing libraries were generated with unique indexes, allowing pooled sequencing. We combined the whole transcriptome assay libraries for sequencing with an Illumina (San Diego, CA) NovaSeq S4 platform at a loading concentration of 250 pM with 5% PhiX. The sequencing parameters used were as follows: read 1, 27 cycles; read 2, 27 cycles; index 1, 8 cycles; index 2, 8 cycles. Q3 normalized data for all 18,000+ genes for each area of interest were generated using the DSP workflow according to the manufacturer’s specifications. Analysis of samples SAZT20099 and SAZT20088 showed clear separation of cytokeratin gene expression within the epidermal and dermal compartments (Supplementary Figure S2).
Differentially expressed gene identification and pathways analysis
Beginning with the Log2 transformation of Q3 normalized expression (generated from the DSP), reads were parametrically batch adjusted with a noninformative prior (Combat function in SVA Bioconductor package version 3.38.0, R version 4.2.0, 64-bit Windows build 22000). After random intercept mixed model ANOVA (by identifier) using the R package LMER (version 3.1-3), a significance level of 0.05 for log-ratio c2 test statistic for overall difference in ROIs (sun-protected and AK edge and center) were used for pairwise comparisons with false discovery rate–adjusted P ≤ 0.05.
), and a priori pathways of interest were transformed from centered log-ratio into biplots.
Biplot and centered log-ratio transformed read analysis
The centered log-ratio transformation is a within-sample standardization that centers reads for each gene by the geometric mean over reads for all the genes; this is a transformation that is similar to the count per million and fragments per kilobase of transcript per million fragments mapped transformations, and justification for this transformation in the context of compositional data analysis and biplots can be found in Greenacre (
). Biplots are a combination of principal component analysis and the relationship, both correlation and variance, between the elements making up the PCA plot (
). Biplots can be used to evaluate the quality of dimension reduction using first two principal components (32.7%, 60.3%, 26.4%, and 49.5% of the variability in the TP53, P38, DNA damage, and extracellular matrix pathways, respectively). We provide additional evidence about the relative importance of the pathways shown in Figure 2 by using the Fisher combined probability test (see below for detail), which combines the P-values from individual genes within each pathway into a single test statistic.
Combined Fisher probability analysis
Combing many tests for significance within a population, like pathways, is a possible approach for filtering pathways of interest. This method is based on summation of individual gene-level P-values into a single one-sided c2 test statistic (
). The P-values for our example in Supplementary Figure S3 are the overall-test statistic based on mixed-effects ANOVA that includes an adjustment for correlation that is induced by sampling (both ROIs and areas of interest) within a tissue sample.
Supplementary Figure S1Total nuclei counts from AOIs. Box and whisker graphs show total nuclei counts for each AOI: sun-protected normal skin (Sun Prot), actinic keratosis edge (Edge), and actinic keratosis center (Center), delineated by cytokeratin positivity (+ indicates cytokeratin positive, − indicates cytokeratin negative). Total number of AOIs that passed QC for each grouping was as follows: Sun Prot + = 34, Edge + = 17, Center + = 18, Sun Prot − = 32, Edge − = 17, Center − = 18. AOI, area of interest; QC, quality control.
Supplementary Figure S2Keratin and collagen gene expression differentiates areas of interest. Q3 normalized read counts (y-axis) for the genes encoding keratin 1 (KRT1) and keratin 10 (KRT10), as well as genes encoding collagen 1 (COL1A1) and collagen 3 (COL3A1) components, clearly differentiate cytokeratin positive (Pos_) from cytokeratin negative (Neg_) areas of interest. Data from samples SAZT20099 and SAZT20088 are presented.
Supplementary Figure S3Pathway analysis comparisons. Combined Fisher probability analysis confirms significant changes in p38, toll-like receptor signaling, p53, and DNA damage pathways in cytokeratin positive AOIs (CK+) and significant changes in ECM and DNA damage pathways in cytokeratin negative AOIs (CK−). P-values are given for each compartment as indicated. AOI, area of interest; CK−, cytokeratin negative compartment; CK+, cytokeratin positive compartment; ECM, extracellular matrix.
00018-0&id=gr1.jpg){kind=link}
00018-0&id=gr2.jpg){kind=link}
00018-0&id=fx1.jpg){kind=link}
00018-0&id=fx2.jpg){kind=link}
00018-0&id=fx3.jpg){kind=link}