UNSEG: unsupervised segmentation of cells and their nuclei in complex tissue samples

Multiplexed imaging technologies have made it possible to interrogate complex tumor microenvironments at sub-cellular resolution within their native spatial context. However, proper quantification of this complexity requires the ability to easily and accurately segment cells into their sub-cellular compartments. Within the supervised learning paradigm, deep learning based segmentation methods demonstrating human level performance have emerged. However, limited work has been done in developing such generalist methods within the label-free unsupervised context. Here we present an unsupervised segmentation (UNSEG) method that achieves deep learning level performance without requiring any training data. UNSEG leverages a Bayesian-like framework and the specificity of nucleus and cell membrane markers to construct an a posteriori probability estimate of each pixel belonging to the nucleus, cell membrane, or background. It uses this estimate to segment each cell into its nuclear and cell-membrane compartments. We show that UNSEG is more internally consistent and better at generalizing to the complexity of tissue morphology than current deep learning methods. This allows UNSEG to unambiguously identify the cytoplasmic compartment of a cell, which we employ to demonstrate its use in an exemplar biological scenario. Within the UNSEG framework, we also introduce a new perturbed watershed algorithm capable of stably and automatically segmenting a cluster of cell nuclei into individual cell nuclei that increases the accuracy of classical watershed. Perturbed watershed can also be used as a standalone algorithm that researchers can incorporate within their supervised or unsupervised learning approaches to extend classical watershed, particularly in the multiplexed imaging context. Finally, as part of developing UNSEG, we have generated a high-quality annotated gastrointestinal tissue (GIT) dataset, which we anticipate will be useful for the broader research community. We demonstrate the efficacy of UNSEG on the GIT dataset, publicly available datasets, and on a range of practical scenarios. In these contexts, we also discuss the possibility of bias inherent in quantification of segmentation accuracy based on F 1 score. Segmentation, despite its long antecedents, remains a challenging problem, particularly in the context of tissue samples. UNSEG, an easy-to-use algorithm, provides an unsupervised approach to overcome this bottleneck, and as we discuss, can help improve deep learning based segmentation methods by providing a bridge between unsupervised and supervised learning paradigms.

Immune microenvironment profiling of normal appearing colorectal mucosa biopsied over repeat patient visits reproducibly separates lynch syndrome patients based on their history of colon cancer

Introduction

Lynch syndrome (LS) is the most common hereditary cause of colorectal cancer (CRC), increasing lifetime risk of CRC by up to 70%. Despite this higher lifetime risk, disease penetrance in LS patients is highly variable and most LS patients undergoing CRC surveillance will not develop CRC. Therefore, biomarkers that can correctly and consistently predict CRC risk in LS patients are needed to both optimize LS patient surveillance and help identify better prevention strategies that reduce risk of CRC development in the subset of high-risk LS patients.

Methods

Normal-appearing colorectal tissue biopsies were obtained during repeat surveillance colonoscopies of LS patients with and without a history of CRC, healthy controls (HC), and patients with a history of sporadic CRC. Biopsies were cultured in an ex-vivo explant system and their supernatants were assayed via multiplexed ELISA to profile the local immune signaling microenvironment. High quality cytokines were identified using the rxCOV fidelity metric. These cytokines were used to perform elastic-net penalized logistic regression-based biomarker selection by computing a new measure - overall selection probability - that quantifies the ability of each marker to discriminate between patient cohorts being compared.

Results

Our study demonstrated that cytokine based local immune microenvironment profiling was reproducible over repeat visits and sensitive to patient LS-status and CRC history. Furthermore, we identified sets of cytokines whose differential expression was predictive of LS-status in patients when compared to sporadic CRC patients and in identifying those LS patients with or without a history of CRC. Enrichment analysis based on these biomarkers revealed an LS and CRC status dependent constitutive inflammatory state of the normal appearing colonic mucosa.

Discussion

This prospective pilot study demonstrated that immune profiling of normal appearing colonic mucosa discriminates LS patients with a prior history of CRC from those without it, as well as patients with a history of sporadic CRC from HC. Importantly, it suggests the existence of immune signatures specific to LS-status and CRC history. We anticipate that our findings have the potential to assess CRC risk in individuals with LS and help in preemptively mitigating it by optimizing surveillance and identifying candidate prevention targets. Further studies are required to validate our findings in an independent cohort of LS patients over multiple visits.

Immune Microenvironment Profiling of Normal Appearing Colorectal Mucosa Biopsied Over Repeat Patient Visits Reproduciably Separates Lynch Syndrome Patients Based on Their History of Colon Cancer

Introduction

Lynch syndrome (LS) is the most common hereditary cause of colorectal cancer (CRC), increasing lifetime risk of CRC by up to 70%. Despite this higher lifetime risk, disease penetrance in LS patients is highly variable and most LS patients undergoing CRC surveillance will not develop CRC. Therefore, biomarkers that can correctly and consistently predict CRC risk in LS patients are needed to both optimize LS patient surveillance and help identify better prevention strategies that reduce risk of CRC development in the subset of high-risk LS patients.

Methods

Normal-appearing colorectal tissue biopsies were obtained during repeat surveillance colonoscopies of LS patients with and without a history of CRC, healthy controls (HC), and patients with a history of sporadic CRC. Biopsies were cultured in an ex-vivo explant system and their supernatants were assayed via multiplexed ELISA to profile the local immune signaling microenvironment. High quality cytokine signatures were identified using rx COV fidelity metric. These signatures were used to perform biomarker selection by computing their selection probability based on penalized logistic regression.

Results

Our study demonstrated that cytokine based local immune microenvironment profiling was reproducible over repeat visits and sensitive to patient LS-status and CRC history. Furthermore, we identified sets of biomarkers whose differential expression was predictive of LS-status in patients when compared to sporadic CRC patients and in identifying those LS patients with or without a history of CRC. Enrichment analysis based on these biomarkers revealed an LS and CRC status dependent constitutive inflammatory state of the normal appearing colonic mucosa.

Discussion

This prospective pilot study demonstrated that immune profiling of normal appearing colonic mucosa discriminates LS patients with a prior history of CRC from those without it, as well as patients with a history of sporadic CRC from HC. Importantly, it suggests existence of immune signatures specific to LS-status and CRC history. We anticipate that our findings have the potential to assess CRC risk in individuals with LS and help in preemptively mitigating it by optimizing surveillance and identifying candidate prevention targets. Further studies are required to validate our findings in an independent cohort of LS patients over multiple visits.

Outcome of trans-catheter edge-to-edge mitral valve repair in patients with severe left ventricular dilatation

Introduction

The COAPT trial established the benefit of trans-catheter edge-to-edge mitral valve repair (TEER) for patients with symptomatic secondary mitral regurgitation (MR). However, patients with severe LV dilatation were excluded from this study. Furthermore, disproportionate LV dilatation has been proposed as a possible explanation for the lack of benefit in the Mitra-FR trial.

Purpose

The purpose of this study was to evaluate the safety and efficacy of TEER in patients with severe LV dilatation.

Methods

We categorized patients with secondary MR and severe LV dilatation based on LVEDD/BSA ratio using a cut-off of >3.6cm/m2 in men and >3.7cm/m2 in women.

Results

We included 121 patients, 25 patients with severe LV dilatation (S) and 96 without (NS). Mean LVEDD was 68.9±6.6mm in the S group vs 58.5±6.5mm in the NS group. Compared to the NS patients, patients with severe LV dilatation were younger (age 69±10 vs 74±10 years), had lower LVEF (37+14% vs 29+7%), and lower GFR (44±21 vs 55±26). There was no significant difference in MR grade (3.3 vs 3.3), however the S group had numerically greater EROA (EROA 0.45 vs 0.35). No significant differences were noted in NYHA class (2.8 vs 2.76) or SPAP (59mmHg for both). Peri-procedural adverse events were rare in both groups. One peri-procedural death was noted, 2 patients underwent urgent dialysis, 1 had cardiogenic shock, and 2 had prolonged ventilation after the procedure, all of these occurred in the NS group. At 1-year follow-up, there was marked improvement in MR grade (2.4±1.4 vs 2.2±1.2 for S and NS respectively). Both groups had a small reduction in LVEDD of about 3mm. Patients in the S group had improved GFR at 1-year follow-up (44±21 to 51±32). No differences were noted in 1, 2, or 5 year mortality between groups (Figures).

Conclusion

Trans-catheter edge-to-edge mitral valve repair in patients with severe LV dilatation is safe and seems to provide similar short-and long-term outcomes compared to patients with less severe dilatation. Further studies are needed to establish the efficacy of TEER for these patients.

Funding Acknowledgement

Type of funding sources: None.

Abstract 1930: Unsupervised sub-cellular segmentation of complex tissue and cell samples using highly multiplexed imaging-derived a priori knowledge

Spatial context of heterocellular interactions within solid tumor microenvironments (TMEs) is important for deciphering the mechanistic underpinnings of malignant TME phenotypes and leveraging that knowledge to improve personalized and precision medicine. Recent development of highly multiplexed imaging approaches, such as co-detection by indexing (CODEX), cyclic immunofluorescence (cycif), and imaging mass cytometry (IMC) has, for the first time, allowed deep interrogation of this heterocellular spatial complexity. Proper quantitation of this complexity, however, requires the ability to easily and accurately segment and localize cells and their sub-cellular compartments within the spatial context of the tumor microenvironment. Seminal approaches based on semi-supervised and supervised learning methods, including deep learning techniques, have been developed to segment cells and their nuclei. However, generalization of these methods to segmenting heterogenous and complex cell and tissue samples with varying resolution, magnification, and dynamic range, remains a persistent bottleneck. In case of deep learning, the requirement of large and accurate annotated datasets further adds to the challenge of seamless integration of systems-based methods in pathology and cancer research. Here, we demonstrate that by leveraging cell-compartment specific a priori knowledge captured by these imaging modalities, we can segment cells in complex tissue and cell samples in an unsupervised manner, without requiring model training. We specifically show that using nucleus and cell-membrane markers, we can accurately segment sub-cellular compartments of a diversity of tissue samples imaged at different resolutions, magnifications and dynamic ranges, and cell samples at varying levels of confluency. We also demonstrate that our method is fast. Given its ease of use, accuracy, robustness, and no requirement of large, annotated datasets, our unsupervised segmentation method fills a much-needed gap toward integration of spatial systems biology and cancer research within the convergence science paradigm.

Citation Format: Bogdan Kochetov, Phoenix D. Bell, Rebecca Raphael, Benjamin J. Raymond, Brian J. Leibowitz, Jingshan Tong, Brenda Diergaarde, Jian Yu, Reetesh K. Pai, Robert E. Schoen, Lin Zhang, Aatur Singhi, Shikhar Uttam. Unsupervised sub-cellular segmentation of complex tissue and cell samples using highly multiplexed imaging-derived a priori knowledge [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1930.

Tumor necrosis factor receptor 2 promotes neuroprotection during chronic autoimmune neuroinflammation

Multiple sclerosis (MS) is an inflammatory autoimmune disorder affecting the central nervous system (CNS) which affects over 400,000 Americans and over 2.5 million people worldwide. Although most patients are initially diagnosed with relapsing-remitting MS, the majority of these patients later develop a chronic-progressive form of MS, for which there is no well-established mouse model. The most common genetic factor associated with MS susceptibility is the Human Leukocyte Antigen (HLA)-DR2b haplotype. Additionally, studies in the experimental autoimmune encephalomyelitis (EAE) mouse model of MS, showed that lack of TNF signaling through its receptor TNFR2 leads to disease exacerbation and severe demyelination. Here, we developed a mouse model which expresses HLA-DR2b and lacks TNFR2, designated DR2bΔR2. Strikingly, DR2bΔR2 mice develop progressive EAE with pathology and clinical features observed in progressive MS patients. Adoptive transfer studies revealed that the clinical phenotype of EAE in DR2bΔR2 mice are largely dependent on TNFR2 expression in the CNS. Subsequently, we showed that DR2bΔR2 mice have a significant increase of lesions in the cerebellum, associated with reduced of oligodendrocyte progenitor cells (OPC) recruitment or function. Moreover, we showed that DR2bΔR2 mice are presented with chronic astrogliosis in demyelinating lesions. Our studies provide key insights into CNS repair and regulatory mechanisms controlled by TNFR2 during neuroinflammation and provide novel therapeutic strategies for treating progressive MS.

Protein kinase C contributes to desensitization of ANG II signaling in adult rat cardiac fibroblasts

We have studied G(q)-linked ANG II signaling [inositol phosphate (IP) accumulation, Ca(2+) mobilization] in primary cultures of rat cardiac fibroblasts (CFs) and have found that ANG II initiates a protein kinase C (PKC)-mediated negative feedback loop that rapidly terminates the ANG II response. Pharmacological inhibition of PKC by staurosporine and GF-109203X doubled IP production over that achieved in response to ANG II alone. Inhibition of PKC also led to larger Ca(2+) transients in response to ANG II, suggesting that Ca(2+) mobilization was proportional to G(q)-phospholipase C-IP(3) activity under the conditions studied. Depletion of cellular PKC by overnight treatment with phorbol 12-myristate 13-acetate (PMA) similarly augmented ANG II-induced IP production. Acute activation of PKC by PMA halved IP formation, with an EC(50) approximately 1 nM; 4alpha-PMA was inactive. Time course data demonstrated that ANG II-mediated IP production fully desensitized within 30 s; PKC inhibition reduced the rate and extent of this desensitization. In cells desensitized to ANG II, a purinergic agonist still mobilized intracellular Ca(2+), indicating that desensitization was homologous. The ANG II-induced Ca(2+) signal was fully resensitized within 30 min. The data demonstrate that a large portion of the IP-Ca(2+) responses of rat CFs to ANG II are short-lived because of rapid, PKC-mediated desensitization.

Short-term responsiveness of manual thoracic end-play assessment to spinal manipulation: a randomized controlled trial of construct validity

OBJECTIVE

To evaluate the short-term responsiveness of rotatory thoracic end-play assessment to spinal manipulation and, thereby, motion palpation construct validity.

DESIGN

Prospective, single-blind, randomized, controlled trial (randomized blocks design).

SETTING

Laboratory, Center for Technique Research.

PARTICIPANTS

Sixty first-year chiropractic college student volunteers; seventy-three possible candidates were screened.

INTERVENTIONS

The treatment group received manual high-velocity, low-amplitude rotatory manipulation. The control group received no intervention to minimize nonspecific effects of sham treatment.

MAIN OUTCOME MEASURES

End-play response, defined as the change from restricted to normal end play immediately after intervention. Responsiveness, defined as the percentage of the end-play response attributable to spinal manipulation: relative response attributable to the maneuver, RRAM = (treatment group response--control group response)/treatment group response.

RESULTS

Ten percent of the tests were positive for restriction of end play in left or right rotation from T3-T4 to T12-L1; the average rate was 2.1 restrictions per subject (SD = 1.4). End-play response was 60% in the treatment group, in contrast to the 37% response in the control group (z = 1.86, p = .04). More than a third of the response in the treatment group was attributable to spinal manipulation (RRAM = 39%). For one examiner, RRAM = 51%. Mild symptomatology did not affect responsiveness.

CONCLUSIONS

The data suggest a moderate short-term responsiveness of rotatory thoracic end-play restriction to spinal manipulation, hence it has utility as a posttreatment evaluative test. This study was the first to use an external standard (manipulation) to demonstrate that segmental end-play restriction changes, hence end-play restriction itself, are detectable in human subjects with manual palpation by chiropractors. Further research is required to determine the generalizability of the study findings.

Low back pain outcome measurement assessment in chiropractic teaching clinics: responsiveness and applicability of two functional disability questionnaires

OBJECTIVE

The major aims were to evaluate responsiveness and clinical/research applicability of the Revised Oswestry Disability Questionnaire (ODQ) and the Dallas Pain Questionnaire (DPQ). Construct and content validity were assessed. Patient characteristics and outcomes were also documented.

DESIGN

Longitudinal observational study.

SETTING

College outpatient clinics.

PARTICIPANTS

Six hundred sixty-three consecutive new patients accepted for treatment of low back pain (LBP) at the clinics over a 1-yr period, age 18 or older.

INTERVENTIONS

Treatment of low back pain by senior interns under the supervision of staff clinicians.

MAIN OUTCOME MEASURES

ODQ and DPQ administered at baseline, 2 wk, 1 month, and monthly up to 6 months. Responsiveness: mean standardized change score (delta' = mudif/sigma dif), relative efficiency ([RE = delta' ODQ/delta DPQ']2), and improvement rates (IR). Applicability: instrument completion rates. Construct validity: correlation with VAS for pain intensity.

RESULTS

ODQ responsiveness was generally consistent over time (delta' = .70-.83) and negligibly better than the DPQ activities of daily living scale (RE = 1.00-1.35); most patients self-reporting improvement showed positive outcomes (IR = 97%). For large samples (n > 100): delta' = .47-.63 and IR = 81% for the DPQ work/leisure scale; delta' = .17-.40 and IR = 54% for the DPQ anxiety/depression and social dimensions. Completion rates: 65%-78% of all instruments; 81%-100% of individual scales. Construct validity: r = .44-.68 for the ODQ, DPQ activities of daily living, and DPQ work/leisure scales; r = .20-.40 for the anxiety/depression scale.

CONCLUSIONS

The ODQ and the activities of daily living and work/leisure scales from the DPQ appear appropriate for monitoring LBP patients returning for care to chiropractic teaching clinics. The social and anxiety/depression dimensions of the DPQ do not appear to be responsive in this population. The latter scale may be unsuitable on the grounds of misinterpretations.