Immune phenotype-genotype associations in primary clear cell renal cell carcinoma and matched metastatic tissue.

Adjuvant immunotherapy has been recently recommended for patients with metastatic ccRCC, but there are no tissue biomarkers to predict treatment response in ccRCC. Potential predictive biomarkers are mainly assessed in primary tumor tissue, whereas metastases remain understudied. To explore potential differences between genomic alterations and immune phenotypes in primary tumors and their matched metastases, we analyzed primary tumors (PTs) of 47 ccRCC patients and their matched distant metastases (METs) by comprehensive targeted parallel sequencing, whole-genome copy number variation (CNV) analysis, determination of microsatellite instability (MSI) and tumor mutational burden (TMB). We quantified the spatial distribution of tumor-infiltrating CD8+ T cells, and co-expression of the T-cell-exhaustion marker TOX by digital immunoprofiling and quantified tertiary lymphoid structures (TLS). Most METs were pathologically "cold". Inflamed, pathologically "hot" PTs were associated with a decreased disease-free survival (DFS), worst for patients with high levels of CD8+TOX+ T cells. Interestingly, inflamed METs showed a relative increase of exhausted CD8+TOX+ T cells and increased accumulative size of TLS compared to PTs. Integrative analysis of molecular and immune phenotypes revealed BAP1 and CDKN2A/B deficiency to be associated with an inflamed immune phenotype. Our results highlight the distinct spatial distribution and differentiation of CD8+ T cells at metastatic sites, and the association of an inflamed microenvironment with specific genomic alterations.

CohortFinder: an open-source tool for data-driven partitioning of digital pathology and imaging cohorts to yield robust machine-learning models.

Batch effects (BEs) refer to systematic technical differences in data collection unrelated to biological variations whose noise is shown to negatively impact machine learning (ML) model generalizability. Here we release CohortFinder (http://cohortfinder.com), an open-source tool aimed at mitigating BEs via data-driven cohort partitioning. We demonstrate CohortFinder improves ML model performance in downstream digital pathology and medical image processing tasks. CohortFinder is freely available for download at cohortfinder.com.

Prognostic impact and causality of age on oncological outcomes in women with endometrial cancer: a multimethod analysis of the randomised PORTEC-1, PORTEC-2, and PORTEC-3 trials.

BACKGROUND: Numerous studies have shown that older women with endometrial cancer have a higher risk of recurrence and cancer-related death. However, it remains unclear whether older age is a causal prognostic factor, or whether other risk factors become increasingly common with age. We aimed to address this question with a unique multimethod study design using state-of-the-art statistical and causal inference techniques on datasets of three large, randomised trials. METHODS: In this multimethod analysis, data from 1801 women participating in the randomised PORTEC-1, PORTEC-2, and PORTEC-3 trials were used for statistical analyses and causal inference. The cohort included 714 patients with intermediate-risk endometrial cancer, 427 patients with high-intermediate risk endometrial cancer, and 660 patients with high-risk endometrial cancer. Associations of age with clinicopathological and molecular features were analysed using non-parametric tests. Multivariable competing risk analyses were performed to determine the independent prognostic value of age. To analyse age as a causal prognostic variable, a deep learning causal inference model called AutoCI was used. FINDINGS: Median follow-up as estimated using the reversed Kaplan-Meier method was 12·3 years (95% CI 11·9-12·6) for PORTEC-1, 10·5 years (10·2-10·7) for PORTEC-2, and 6·1 years (5·9-6·3) for PORTEC-3. Both overall recurrence and endometrial cancer-specific death significantly increased with age. Moreover, older women had a higher frequency of deep myometrial invasion, serous tumour histology, and p53-abnormal tumours. Age was an independent risk factor for both overall recurrence (hazard ratio [HR] 1·02 per year, 95% CI 1·01-1·04; p=0·0012) and endometrial cancer-specific death (HR 1·03 per year, 1·01-1·05; p=0·0012) and was identified as a significant causal variable. INTERPRETATION: This study showed that advanced age was associated with more aggressive tumour features in women with endometrial cancer, and was independently and causally related to worse oncological outcomes. Therefore, our findings suggest that older women with endometrial cancer should not be excluded from diagnostic assessments, molecular testing, and adjuvant therapy based on their age alone. FUNDING: None.

Prediction of recurrence risk in endometrial cancer with multimodal deep learning.

Predicting distant recurrence of endometrial cancer (EC) is crucial for personalized adjuvant treatment. The current gold standard of combined pathological and molecular profiling is costly, hampering implementation. Here we developed HECTOR (histopathology-based endometrial cancer tailored outcome risk), a multimodal deep learning prognostic model using hematoxylin and eosin-stained, whole-slide images and tumor stage as input, on 2,072 patients from eight EC cohorts including the PORTEC-1/-2/-3 randomized trials. HECTOR demonstrated C-indices in internal (n = 353) and two external (n = 160 and n = 151) test sets of 0.789, 0.828 and 0.815, respectively, outperforming the current gold standard, and identified patients with markedly different outcomes (10-year distant recurrence-free probabilities of 97.0%, 77.7% and 58.1% for HECTOR low-, intermediate- and high-risk groups, respectively, by Kaplan-Meier analysis). HECTOR also predicted adjuvant chemotherapy benefit better than current methods. Morphological and genomic feature extraction identified correlates of HECTOR risk groups, some with therapeutic potential. HECTOR improves on the current gold standard and may help delivery of personalized treatment in EC.

Image-based consensus molecular subtyping in rectal cancer biopsies and response to neoadjuvant chemoradiotherapy.

The development of deep learning (DL) models to predict the consensus molecular subtypes (CMS) from histopathology images (imCMS) is a promising and cost-effective strategy to support patient stratification. Here, we investigate whether imCMS calls generated from whole slide histopathology images (WSIs) of rectal cancer (RC) pre-treatment biopsies are associated with pathological complete response (pCR) to neoadjuvant long course chemoradiotherapy (LCRT) with single agent fluoropyrimidine. DL models were trained to classify WSIs of colorectal cancers stained with hematoxylin and eosin into one of the four CMS classes using a multi-centric dataset of resection and biopsy specimens (n = 1057 WSIs) with paired transcriptional data. Classifiers were tested on a held out RC biopsy cohort (ARISTOTLE) and correlated with pCR to LCRT in an independent dataset merging two RC cohorts (ARISTOTLE, n = 114 and SALZBURG, n = 55 patients). DL models predicted CMS with high classification performance in multiple comparative analyses. In the independent cohorts (ARISTOTLE, SALZBURG), cases with WSIs classified as imCMS1 had a significantly higher likelihood of achieving pCR (OR = 2.69, 95% CI 1.01-7.17, p = 0.048). Conversely, imCMS4 was associated with lack of pCR (OR = 0.25, 95% CI 0.07-0.88, p = 0.031). Classification maps demonstrated pathologist-interpretable associations with high stromal content in imCMS4 cases, associated with poor outcome. No significant association was found in imCMS2 or imCMS3. imCMS classification of pre-treatment biopsies is a fast and inexpensive solution to identify patient groups that could benefit from neoadjuvant LCRT. The significant associations between imCMS1/imCMS4 with pCR suggest the existence of predictive morphological features that could enhance standard pathological assessment.

Prognostic and Predictive Value of Immunoscore in Stage III Colorectal Cancer: Pooled Analysis of Cases From the SCOT and IDEA-HORG Studies.

PURPOSE: Immunoscore (IS) is prognostic in stage III colorectal cancer (CRC) and may predict benefit of duration (6 v 3 months) of adjuvant infusional fluorouracil, leucovorin, and oxaliplatin (FOLFOX) chemotherapy. We sought to determine IS prognostic and predictive value in stage-III CRC treated with adjuvant FOLFOX or oral capecitabine and infusional oxaliplatin (CAPOX) in the SCOT and IDEA-HORG trials. METHODS: Three thousand sixty-one cases had tumor samples, of which 2,643 (1,792 CAPOX) were eligible for IS testing. Predefined cutoffs (IS-Low and IS-High) were used to classify cases into two groups for analysis of disease-free survival (3-year DFS) and multivariable-adjusted hazard ratios (mvHRs) by Cox regression. RESULTS: IS was determined in 2,608 (99.5%) eligible cases, with 877 (33.7%) samples classified as IS-Low. IS-Low tumors were more commonly high-risk (T4 and/or N2; 52.9% IS-Low v 42.2% IS-High; P < .001) and in younger patients (P = .024). Patients with IS-Low tumors had significantly shorter DFS in the CAPOX, FOLFOX, and combined cohorts (mvHR, 1.52 [95% CI, 1.28 to 1.82]; mvHR, 1.58 [95% CI, 1.22 to 2.04]; and mvHR, 1.55 [95% CI, 1.34 to 1.79], respectively; P < .001 all comparisons), regardless of sex, BMI, clinical risk group, tumor location, treatment duration, or chemotherapy regimen. IS prognostic value was greater in younger (≤65 years) than older (>65 years) patients in the CAPOX cohort (mvHR, 1.92 [95% CI, 1.50 to 2.46] v 1.28 [95% CI, 1.01 to 1.63], PINTERACTION = .026), and in DNA mismatch repair proficient than deficient mismatch repair disease (mvHR, 1.68 [95% CI, 1.41 to 2.00] v 0.67 [95% CI, 0.30 to 1.49], PINTERACTION = .03), although these exploratory analyses were uncorrected for multiple testing. Adding IS to a model containing all clinical variables significantly improved prediction of DFS (likelihood ratio test, P < .001) regardless of MMR status. CONCLUSION: IS is prognostic in stage III CRC treated with FOLFOX or CAPOX, including within clinically relevant tumor subgroups. Possible variation in IS prognostic value by age and MMR status, and prediction of benefit from extended adjuvant therapy merit validation.

Domain generalization across tumor types, laboratories, and species - Insights from the 2022 edition of the Mitosis Domain Generalization Challenge.

Recognition of mitotic figures in histologic tumor specimens is highly relevant to patient outcome assessment. This task is challenging for algorithms and human experts alike, with deterioration of algorithmic performance under shifts in image representations. Considerable covariate shifts occur when assessment is performed on different tumor types, images are acquired using different digitization devices, or specimens are produced in different laboratories. This observation motivated the inception of the 2022 challenge on MItosis Domain Generalization (MIDOG 2022). The challenge provided annotated histologic tumor images from six different domains and evaluated the algorithmic approaches for mitotic figure detection provided by nine challenge participants on ten independent domains. Ground truth for mitotic figure detection was established in two ways: a three-expert majority vote and an independent, immunohistochemistry-assisted set of labels. This work represents an overview of the challenge tasks, the algorithmic strategies employed by the participants, and potential factors contributing to their success. With an F1 score of 0.764 for the top-performing team, we summarize that domain generalization across various tumor domains is possible with today's deep learning-based recognition pipelines. However, we also found that domain characteristics not present in the training set (feline as new species, spindle cell shape as new morphology and a new scanner) led to small but significant decreases in performance. When assessed against the immunohistochemistry-assisted reference standard, all methods resulted in reduced recall scores, with only minor changes in the order of participants in the ranking.

Pathway level subtyping identifies a slow-cycling biological phenotype associated with poor clinical outcomes in colorectal cancer.

Molecular stratification using gene-level transcriptional data has identified subtypes with distinctive genotypic and phenotypic traits, as exemplified by the consensus molecular subtypes (CMS) in colorectal cancer (CRC). Here, rather than gene-level data, we make use of gene ontology and biological activation state information for initial molecular class discovery. In doing so, we defined three pathway-derived subtypes (PDS) in CRC: PDS1 tumors, which are canonical/LGR5+ stem-rich, highly proliferative and display good prognosis; PDS2 tumors, which are regenerative/ANXA1+ stem-rich, with elevated stromal and immune tumor microenvironmental lineages; and PDS3 tumors, which represent a previously overlooked slow-cycling subset of tumors within CMS2 with reduced stem populations and increased differentiated lineages, particularly enterocytes and enteroendocrine cells, yet display the worst prognosis in locally advanced disease. These PDS3 phenotypic traits are evident across numerous bulk and single-cell datasets, and demark a series of subtle biological states that are currently under-represented in pre-clinical models and are not identified using existing subtyping classifiers.

SST-editing: in silico spatial transcriptomic editing at single-cell resolution.

MOTIVATION: Generative Adversarial Nets (GAN) achieve impressive performance for text-guided editing of natural images. However, a comparable utility of GAN remains understudied for spatial transcriptomics (ST) technologies with matched gene expression and biomedical image data. RESULTS: We propose In Silico Spatial Transcriptomic editing that enables gene expression-guided editing of immunofluorescence images. Using cell-level spatial transcriptomics data extracted from normal and tumor tissue slides, we train the approach under the framework of GAN (Inversion). To simulate cellular state transitions, we then feed edited gene expression levels to trained models. Compared to normal cellular images (ground truth), we successfully model the transition from tumor to normal tissue samples, as measured with quantifiable and interpretable cellular features. AVAILABILITY AND IMPLEMENTATION: https://github.com/CTPLab/SST-editing.

Multiplex analysis of intratumoural immune infiltrate and prognosis in patients with stage II-III colorectal cancer from the SCOT and QUASAR 2 trials: a retrospective analysis.

BACKGROUND: Tumour-infiltrating CD8+ cytotoxic T cells confer favourable prognosis in colorectal cancer. The added prognostic value of other infiltrating immune cells is unclear and so we sought to investigate their prognostic value in two large clinical trial cohorts. METHODS: We used multiplex immunofluorescent staining of tissue microarrays to assess the densities of CD8+, CD20+, FoxP3+, and CD68+ cells in the intraepithelial and intrastromal compartments from tumour samples of patients with stage II-III colorectal cancer from the SCOT trial (ISRCTN59757862), which examined 3 months versus 6 months of adjuvant oxaliplatin-based chemotherapy, and from the QUASAR 2 trial (ISRCTN45133151), which compared adjuvant capecitabine with or without bevacizumab. Both trials included patients aged 18 years or older with an Eastern Cooperative Oncology Group performance status of 0-1. Immune marker predictors were analysed by multiple regression, and the prognostic and predictive values of markers for colorectal cancer recurrence-free interval by Cox regression were assessed using the SCOT cohort for discovery and QUASAR 2 cohort for validation. FINDINGS: After exclusion of cases without tissue microarrays and with technical failures, and following quality control, we included 2340 cases from the SCOT trial and 1069 from the QUASAR 2 trial in our analysis. Univariable analysis of associations with recurrence-free interval in cases from the SCOT trial showed a strong prognostic value of intraepithelial CD8 (CD8IE) as a continuous variable (hazard ratio [HR] for 75th vs 25th percentile [75vs25] 0·73 [95% CI 0·68-0·79], p=2·5 × 10-16), and of intrastromal FoxP3 (FoxP3IS; 0·71 [0·64-0·78], p=1·5 × 10-13) but not as strongly in the epithelium (FoxP3IE; 0·89 [0·84-0·96], p=1·5 × 10-4). Associations of other markers with recurrence-free interval were moderate. CD8IE and FoxP3IS retained independent prognostic value in bivariable and multivariable analysis, and, compared with either marker alone, a composite marker including both markers (CD8IE-FoxP3IS) was superior when assessed as a continuous variable (adjusted [a]HR75 vs 25 0·70 [95% CI 0·63-0·78], p=5·1 × 10-11) and when categorised into low, intermediate, and high density groups using previously published cutpoints (aHR for intermediate vs high 1·68 [95% CI 1·29-2·20], p=1·3 × 10-4; low vs high 2·58 [1·91-3·49], p=7·9 × 10-10), with performance similar to the gold-standard Immunoscore. The prognostic value of CD8IE-FoxP3IS was confirmed in cases from the QUASAR 2 trial, both as a continuous variable (aHR75 vs 25 0·84 [95% CI 0·73-0·96], p=0·012) and as a categorical variable for low versus high density (aHR 1·80 [95% CI 1·17-2·75], p=0·0071) but not for intermediate versus high (1·30 [0·89-1·88], p=0·17). INTERPRETATION: Combined evaluation of CD8IE and FoxP3IS could help to refine risk stratification in colorectal cancer. Investigation of FoxP3IS cells as an immunotherapy target in colorectal cancer might be merited. FUNDING: Medical Research Council, National Institute for Health Research, Cancer Research UK, Swedish Cancer Society, Roche, and Promedica Foundation.

Poor Diagnostic Reproducibility in the Identification of Nonconventional Dysplasia in Colitis Impacts the Application of Histologic Stratification Tools.

Due to their increased cancer risk, patients with longstanding inflammatory bowel disease are offered endoscopic surveillance with concomitant histopathologic assessments, aimed at identifying dysplasia as a precursor lesion of colitis-associated colorectal cancer. However, this strategy is beset with difficulties and limitations. Recently, a novel classification criterion for colitis-associated low-grade dysplasia has been proposed, and an association between nonconventional dysplasia and progression was reported, suggesting the possibility of histology-based stratification of patients with colitis-associated lesions. Here, a cohort of colitis-associated lesions was assessed by a panel of 6 experienced pathologists to test the applicability of the published classification criteria and try and validate the association between nonconventional dysplasia and progression. While confirming the presence of different morphologic patterns of colitis-associated dysplasia, the study demonstrated difficulties concerning diagnostic reproducibility between pathologists and was unable to validate the association of nonconventional dysplasia with cancer progression. Our study highlights the overall difficulty of using histologic assessment of precursor lesions for cancer risk prediction in inflammatory bowel disease patients and suggests the need for a different diagnostic strategy that can objectively identify high-risk phenotypes.

Swiss digital pathology recommendations: results from a Delphi process conducted by the Swiss Digital Pathology Consortium of the Swiss Society of Pathology.

Integration of digital pathology (DP) into clinical diagnostic workflows is increasingly receiving attention as new hardware and software become available. To facilitate the adoption of DP, the Swiss Digital Pathology Consortium (SDiPath) organized a Delphi process to produce a series of recommendations for DP integration within Swiss clinical environments. This process saw the creation of 4 working groups, focusing on the various components of a DP system (1) scanners, quality assurance and validation of scans, (2) integration of Whole Slide Image (WSI)-scanners and DP systems into the Pathology Laboratory Information System, (3) digital workflow-compliance with general quality guidelines, and (4) image analysis (IA)/artificial intelligence (AI), with topic experts for each recruited for discussion and statement generation. The work product of the Delphi process is 83 consensus statements presented here, forming the basis for "SDiPath Recommendations for Digital Pathology". They represent an up-to-date resource for national and international hospitals, researchers, device manufacturers, algorithm developers, and all supporting fields, with the intent of providing expectations and best practices to help ensure safe and efficient DP usage.

National digital pathology projects in Switzerland: A 2023 update.

The Swiss Digital Pathology Consortium (SDiPath) was founded in 2018 as a working group of the Swiss Society for Pathology with the aim of networking, training, and promoting digital pathology (DP) at a national level. Since then, two national surveys have been carried out on the level of knowledge, dissemination, use, and needs in DP, which have resulted in clear fields of action. In addition to organizing symposia and workshops, national guidelines were drawn up and an initiative for a national DP platform actively codesigned. With the growing use of digital image processing and artificial intelligence tools, continuous monitoring, evaluation, and exchange of experiences will be pursued, along with best practices.

Digital image analysis and artificial intelligence in pathology diagnostics-the Swiss view.

Digital pathology (DP) is increasingly entering routine clinical pathology diagnostics. As digitization of the routine caseload advances, implementation of digital image analysis algorithms and artificial intelligence tools becomes not only attainable, but also desirable in daily sign out. The Swiss Digital Pathology Consortium (SDiPath) has initiated a Delphi process to generate best-practice recommendations for various phases of the process of digitization in pathology for the local Swiss environment, encompassing the following four topics: i) scanners, quality assurance, and validation of scans; ii) integration of scanners and systems into the pathology laboratory information system; iii) the digital workflow; and iv) digital image analysis (DIA)/artificial intelligence (AI). The current article focuses on the DIA-/AI-related recommendations generated and agreed upon by the working group and further verified by the Delphi process among the members of SDiPath. Importantly, they include the view and the currently perceived needs of practicing pathologists from multiple academic and cantonal hospitals as well as private practices.

Accounting for intensity variation in image analysis of large-scale multiplexed clinical trial datasets.

Multiplex immunofluorescence (mIF) imaging can provide comprehensive quantitative and spatial information for multiple immune markers for tumour immunoprofiling. However, application at scale to clinical trial samples sourced from multiple institutions is challenging due to pre-analytical heterogeneity. This study reports an analytical approach to the largest multi-parameter immunoprofiling study of clinical trial samples to date. We analysed 12,592 tissue microarray (TMA) spots from 3,545 colorectal cancers sourced from more than 240 institutions in two clinical trials (QUASAR 2 and SCOT) stained for CD4, CD8, CD20, CD68, FoxP3, pan-cytokeratin, and DAPI by mIF. TMA slides were multi-spectrally imaged and analysed by cell-based and pixel-based marker analysis. We developed an adaptive thresholding method to account for inter- and intra-slide intensity variation in TMA analysis. Applying this method effectively ameliorated inter- and intra-slide intensity variation improving the image analysis results compared with methods using a single global threshold. Correlation of CD8 data derived by our mIF analysis approach with single-plex chromogenic immunohistochemistry CD8 data derived from subsequent sections indicates the validity of our method (Spearman's rank correlation coefficients ρ between 0.63 and 0.66, p ≪ 0.01) as compared with the current gold standard analysis approach. Evaluation of correlation between cell-based and pixel-based analysis results confirms equivalency (ρ > 0.8, p ≪ 0.01, except for CD20 in the epithelial region) of both analytical approaches. These data suggest that our adaptive thresholding approach can enable analysis of mIF-stained clinical trial TMA datasets by digital pathology at scale for precision immunoprofiling.

Cystatin C is glucocorticoid responsive, directs recruitment of Trem2+ macrophages, and predicts failure of cancer immunotherapy.

Cystatin C (CyC), a secreted cysteine protease inhibitor, has unclear biological functions. Many patients exhibit elevated plasma CyC levels, particularly during glucocorticoid (GC) treatment. This study links GCs with CyC's systemic regulation by utilizing genome-wide association and structural equation modeling to determine CyC production genetics in the UK Biobank. Both CyC production and a polygenic score (PGS) capturing predisposition to CyC production were associated with increased all-cause and cancer-specific mortality. We found that the GC receptor directly targets CyC, leading to GC-responsive CyC secretion in macrophages and cancer cells. CyC-knockout tumors displayed significantly reduced growth and diminished recruitment of TREM2+ macrophages, which have been connected to cancer immunotherapy failure. Furthermore, the CyC-production PGS predicted checkpoint immunotherapy failure in 685 patients with metastatic cancer from combined clinical trial cohorts. In conclusion, CyC may act as a GC effector pathway via TREM2+ macrophage recruitment and may be a potential target for combination cancer immunotherapy.

Transformer-based biomarker prediction from colorectal cancer histology: A large-scale multicentric study.

Deep learning (DL) can accelerate the prediction of prognostic biomarkers from routine pathology slides in colorectal cancer (CRC). However, current approaches rely on convolutional neural networks (CNNs) and have mostly been validated on small patient cohorts. Here, we develop a new transformer-based pipeline for end-to-end biomarker prediction from pathology slides by combining a pre-trained transformer encoder with a transformer network for patch aggregation. Our transformer-based approach substantially improves the performance, generalizability, data efficiency, and interpretability as compared with current state-of-the-art algorithms. After training and evaluating on a large multicenter cohort of over 13,000 patients from 16 colorectal cancer cohorts, we achieve a sensitivity of 0.99 with a negative predictive value of over 0.99 for prediction of microsatellite instability (MSI) on surgical resection specimens. We demonstrate that resection specimen-only training reaches clinical-grade performance on endoscopic biopsy tissue, solving a long-standing diagnostic problem.

Bone marrow haematopoiesis in patients with COVID-19.

AIMS: Severe acute respiratory syndrome coronavirus type 2 (SARS-CoV-2) infection broadly affects organ homeostasis, including the haematopoietic system. Autopsy studies are a crucial tool for investigation of organ-specific pathologies. Here we perform an in-depth analysis of the impact of severe coronavirus disease 2019 (COVID-19) on bone marrow haematopoiesis in correlation with clinical and laboratory parameters. METHODS AND RESULTS: Twenty-eight autopsy cases and five controls from two academic centres were included in the study. We performed a comprehensive analysis of bone marrow pathology and microenvironment features with clinical and laboratory parameters and assessed SARS-CoV-2 infection of the bone marrow by quantitative polymerase chain reaction (qPCR) analysis. In COVID-19 patients, bone marrow specimens showed a left-shifted myelopoiesis (19 of 28, 64%), increased myeloid-erythroid ratio (eight of 28, 28%), increased megakaryopoiesis (six of 28, 21%) and lymphocytosis (four of 28, 14%). Strikingly, a high proportion of COVID-19 specimens showed erythrophagocytosis (15 of 28, 54%) and the presence of siderophages (11 of 15, 73%) compared to control cases (none of five, 0%). Clinically, erythrophagocytosis correlated with lower haemoglobin levels and was more frequently observed in patients from the second wave. Analysis of the immune environment showed a strong increase in CD68+ macrophages (16 of 28, 57%) and a borderline lymphocytosis (five of 28, 18%). The stromal microenvironment showed oedema (two of 28, 7%) and severe capillary congestion (one of 28, 4%) in isolated cases. No stromal fibrosis or microvascular thrombosis was found. While all cases had confirmed positive testing of SARS-CoV-2 in the respiratory system, SARS-CoV-2 was not detected in the bone marrow by high-sensitivity PCR, suggesting that SARS-CoV-2 does not commonly replicate in the haematopoietic microenvironment. CONCLUSIONS: SARS-CoV-2 infection indirectly impacts the haematological compartment and the bone marrow immune environment. Erythrophagocytosis is frequent and associated with lower haemoglobin levels in patients with severe COVID-19.