Microenvironmental immune cell alterations across the spectrum of nodular lymphocyte predominant Hodgkin lymphoma and T-cell/histiocyte-rich large B-cell lymphoma

Background The clinicopathological spectrum of nodular lymphocyte predominant Hodgkin lymphoma (NLPHL), also known as nodular lymphocyte predominant B-cell lymphoma, partially overlaps with T-cell/histiocyte-rich large B-cell lymphoma (THRLCBL). NLPHL histology may vary in architecture and B-cell/T-cell composition of the tumour microenvironment. However, the immune cell phenotypes accompanying different histological patterns remain poorly characterised. Methods We applied a multiplexed immunofluorescence workflow to identify differential expansion/depletion of multiple microenvironmental immune cell phenotypes between cases of NLPHL showing different histological patterns (as described by Fan et al, 2003) and cases of THRLBCL. Results FOXP3-expressing T-regulatory cells were conspicuously depleted across all NLPHL cases. As histology progressed to variant Fan patterns C and E of NLPHL and to THRLBCL, there were progressive expansions of cytotoxic granzyme-B-expressing natural killer and CD8-positive T-cells, PD1-expressing CD8-positive T-cells, and CD163-positive macrophages including a PDL1-expressing subset. These occurred in parallel to depletion of NKG2A-expressing natural killer and CD8-positive T-cells. Discussion These findings provide new insights on the immunoregulatory mechanisms involved in NLPHL and THLRBCL pathogenesis, and are supportive of an increasingly proposed biological continuum between these two lymphomas. Additionally, the findings may help establish new biomarkers of high-risk disease, which could support a novel therapeutic program of immune checkpoint interruption targeting the PD1:PDL1 and/or NKG2A:HLA-E axes in the management of high-risk NLPHL and THRLBCL.

Dynamic pixel thresholding was executed in 'ImageJ' (FIJI distribution; version 1.53q) to generate secondary thresholded channels with relative insensitivity to wider technical variations in signal/background characteristics.
Functions were executed in the following sequence (with further rationales for their inclusion provided): 1. "Despeckle" on primary channels for all markers (not including DAPI).
− Rationale: this de-noising function helped to make generate more informative Pearson correlation coefficient measurements between channel pairs (see following sections).
2. "Duplicate" on primary channels to generate secondary channels.

"8-bit" conversion on secondary channels.
− Rationale: required to ensure consistent Auto Threshold behaviour.

"Auto
Threshold" on secondary channels with settings: method=RenyiEntropy; ignore white; white objects on black background.
− Rationale: to generate secondary thresholded channels with binary pixel values of either zero (in negative background) or 255 (in positive signal).
− Note: this is the 'plugin' and not the 'applet' implementation of FIJI's thresholding functions.NB:: "resetMinAndMax()" was also executed between each step to ensure consistent behaviour of functions, because some can be affected by alterations to the Look Up Table minima and maxima.
Two human investigators (CP/AA) independently quality-checked all outputs from this sequence.Tiles generating a non-satisfactory output secondary thresholded channel (i.e., excessively under or over-thresholded) for any marker were excluded (example satisfactory and non-satisfactory outputs shown in Figure S1).

Digital image analysis workflow (2): Cell segmentation and measurements
Processed images with both including both primary channels and secondary threshold channels were then segmented using 'StarDist' (version 0. Two human investigators (CP/AA) independently manually quality-checked all outputs after segmentation and regions with gross non-cellular under/over-staining artefacts were excluded.Additionally, detections with extremes of nuclear diameter (<4 microns or >11 microns, representing over/under-segmented nuclear detections respectively) were excluded from subsequent phenotyping.In image tiles stained with the lymphocyte panel, CD20-positive LP cells were manually annotated to distinguish them from non-neoplastic B-cells in the subsequent analyses.

Digital image analysis workflow (3): Algorithmic phenotyping
RStudio (version 1.4.1717,running R 4.11) was used to run algorithmic phenotyping and further analysis.Phenotypes of cell detections were primarily determined via thresholded positive coverage (TPC; measured in the whole-cell compartment via secondary thresholded channels).TPC determined the final status for nuclear (FOXP3) and cytoplasmic-predominant (GZMB/GNLY/CD68) markers.However, if detections were provisionally positive for two membranous markers based on TPCs, Pearson correlation coefficients (PCC) between primary channel intensities were implemented as an additional accuracy-improving step in the context of a crowded TME.The rationale being that high PCC can differentiate double-positive phenotypes with colocalising signals from those with low PCC which are likely spurious phenotypes (due to contamination of cell detections by neighbouring cell membranes), and that the difference between TPCs can determine which is the likely contaminant (illustrated in Figure S2).The optimised TPC and PCC base thresholds applied are provided below.
Detections appearing positive for three or more membranous markers were similarly resolved by consensus agreement of pair-wise sub-resolutions (i.e., if a marker is resolved as genuinely positive in all the required pair-wise resolutions it features in, it is considered positive by consensus).

Fixed thresholds applied in algorithmic phenotyping
6. "Close" binary function on secondary thresholded channels with settings: iteration=1; count=2.− Rationale: this partially fills concave or ring-like thresholded objects, thickening the inner aspect of membranous signals and thus augmenting the subsequent detection of truly positive cells.7. "Remove Outliers" binary function on secondary thresholded channels with settings: radius=2; threshold=1.− Rationale: to remove speck-like pixel outliers.8. "Divide" math function on secondary thresholded channels with value = 255.− Rationale: this changes to binary pixel values to either zero or 1, meaning than mean pixel values in subsequent cell detections become equivalent to coverage by thresholded positive signal.
thresholded positive coverage; PCC = Pearson correlation coefficient.* PCC thresholds of 1 indicate that no visually convincing double-positive cells of that kind were identified in representative evaluation sets.

Figure S2 :
Figure S2: Example colocalisation-based resolutions of cell detections with apparent double-positivity for two membranous markers

Figure S3 :
Figure S3: Example evaluation of marker positivity determination

Figure S4 :
Figure S4: Example evaluation of co-localisation-based resolutions

Figure S5 :
Figure S5: Additional lymphocyte panel phenotypes notable for their expansion in the single "Treg-high" THRLBCL case

Figure S6 :
Figure S6: Additional macrophage panel phenotypes notable for their expansion in the single "Treg-high" THRLBCL case