PreCoF: counterfactual explanations for fairness.

This paper studies how counterfactual explanations can be used to assess the fairness of a model. Using machine learning for high-stakes decisions is a threat to fairness as these models can amplify bias present in the dataset, and there is no consensus on a universal metric to detect this. The appropriate metric and method to tackle the bias in a dataset will be case-dependent, and it requires insight into the nature of the bias first. We aim to provide this insight by integrating explainable AI (XAI) research with the fairness domain. More specifically, apart from being able to use (Predictive) Counterfactual Explanations to detect explicit bias when the model is directly using the sensitive attribute, we show that it can also be used to detect implicit bias when the model does not use the sensitive attribute directly but does use other correlated attributes leading to a substantial disadvantage for a protected group. We call this metric PreCoF, or Predictive Counterfactual Fairness. Our experimental results show that our metric succeeds in detecting occurrences of implicit bias in the model by assessing which attributes are more present in the explanations of the protected group compared to the unprotected group. These results could help policymakers decide on whether this discrimination is justified or not.

Cardiogeneticsbank@UZA: A Collection of DNA, Tissues, and Cell Lines as a Translational Tool.

Cardiogeneticsbank@UZA is an academic hospital integrated biobank that collects aortic tissue, blood, cell lines (fibroblasts, vascular smooth muscle cells, peripheral blood mononuclear cells, and induced pluripotent stem cells), and DNA from patients with cardiogenetic disorders, for both diagnostic and research purposes. We adhere to a quality management system and have established standard protocols for the sampling and processing of all cardiogenetic patient related materials. Cardiogeneticsbank@UZA is embedded in the Biobanking and Biomolecular Resources Research Infrastructure Belgium (BBMRI.be) and samples from this biobank are available for commercial and academic researchers, through an established access procedure. Currently, the extremely valuable cardiogenetics collection consists of more than 8,700 DNA samples, 380 tissue samples, and 500 cell lines of 7,578 patients, and is linked with extensive clinical data. Some interesting potential research applications are discussed.

Biobanking for Viral Hepatitis Research.

Introduction: Viral hepatitis is a worldwide, important health issue. The optimal management of viral hepatitis infections faces numerous challenges. In this paper, we describe how biobanking of biological samples derived from viral hepatitis patients collected both in-hospital and during community outreach screenings provides a unique collection of samples. Materials and Methods: All samples and materials were provided with a study code within the SLIMS system Study protocols and an informed consent form were approved by the Antwerp University Hospital/University of Antwerp Ethical Committee. Systematic biobanking was initiated in October 2014. Collected sample types include: (1) serum and plasma of all newly diagnosed HBV, HCV, HDV, and HEV positive patients; (2) left-over serum and plasma samples from all PCR analyses for HBV and HCV performed in the context of routine clinical care; (3) left-over liver tissue not needed for routine histological diagnosis after liver biopsy; and (4) additional virus-specific, appropriate sample types using a scientific rationale-based approach. A community outreach screening program was performed in three major Belgian cities. Serum, EDTA, Tempus Blood RNA and BD Vacutainer CPT were collected. CPT tubes were centrifuged on-site and mononuclear cells collected within 24 h. Results: Concerning community screening: 298 individuals supplied all 4 sample types. Samples were stored at -150°C and were logged in the biobank SLIMS database. Samples were used for HBV-related immunological and biomarker studies. DNA isolated from plasma samples derived from chronic HBV patients was used to investigate Single Nucleotide Polymorphism rs 1790008. Serum samples collected from chronic hepatitis C patients were used to assess the efficacy of HCV treatment. Peripheral Blood Mononuclear Cells (PBMC) isolated from chronic HBV patients and healthy controls were used for different immunological study purposes. Virus isolated from biobanked stool of a chronic hepatitis E patient was used to establish a mouse model for Hepatitis E infections, allowing further HEV virology studies. Conclusion: The establishment of a biobank with samples collected both in-hospital and during community-outreach screening resulted in a unique, continuously expanding collection of biological samples which provides an excellent platform for prompt answers to clinically and translational relevant research questions.

Circulating Cell-Free DNA and RNA Analysis as Liquid Biopsy: Optimal Centrifugation Protocol.

The combined analysis of circulating cell-free (tumor) DNA (cfDNA/ctDNA) and circulating cell-free (tumor) RNA (cfRNA/ctRNA) shows great promise in determining the molecular profile of cancer patients. Optimization of the workflow is necessary to achieve consistent and reproducible results. In this study, we compared five centrifugation protocols for the optimal yield of both cfDNA/ctDNA and cfRNA/ctRNA. These protocols varied in centrifugation speed, ambient temperature, time, and number of centrifugation steps. Samples from 33 participants were collected in either BD Vacutainer K2EDTA (EDTA) tubes or cell-free DNA BCT® (Streck) tubes. cfDNA concentration and fragment size, and cfRNA concentration were quantitated in all samples by digital droplet PCR (ddPCR) and quantitative PCR (qPCR). The KRAS-mutated ctDNA and ctRNA fraction was determined via ddPCR. In EDTA tubes, the protocol generating both plasma and platelets was found to produce high quality cfDNA and cfRNA concentrations. Two-step, high-speed centrifugation protocols were associated with high cfDNA but low cfRNA concentrations. High cfRNA concentrations were generated by a one-step, low-speed protocol. However, this coincided with a high amount of genomic DNA (gDNA) contamination. In Streck tubes, two-step, high-speed centrifugation protocols also generated good quality, high cfDNA concentration. However, these tubes are not compatible with cfRNA analysis.

The prevalence of estrogen receptor-1 mutation in advanced breast cancer: The estrogen receptor one study (EROS1).

BACKGROUND: Breast cancer has, due its high incidence, the highest mortality of cancer in women. The most common molecular variety of breast cancer is luminal subtype that expresses estrogen and progesterone receptors. Estrogen receptor alpha (ERα), encoded by the estrogen receptor1 (ESR1) gene, is expressed in approximately 70% of all breast cancers, and hormonal therapy represents a major treatment modality in all stages of ER positive breast cancers. Acquired mutations in the ligand-binding domain (LBD) of ERα, referred as ESR1 mutation, result in resistance to different endocrine therapies leading to disease progression or recurrence. Recent studies reviled that these ESR1 mutations lead to constitutive activity of the estrogen receptor ER, meaning that the receptor is active in absence of its ligand conferring resistance against endocrine therapy and tumor growth. Published studies have not yet been able to determine the exact prevalence rate of ESR1 mutations, but set the outer boundaries between 11-55%. PURPOSE: The goal of the present study is to determine the frequency rate of ESR1 mutations in ER positive recurrent breast cancer by using digital droplet PCR (ddPCR) technique. MATERIALS AND METHODS: This retrospective study was conducted in the Multidisciplinary Breast Clinic of Antwerp University Hospital. The seven most common ESR1mutations (c.1138G>C (p. (E380Q)), c.1610A>G (p.(Y537C)), c.1613A>G (p.(p.D538G)), c.1607T>G (p.(L536R)), c.1387T>C (p.S463R)), c.16410A>C (p.(Y537S)), c.609T>A (p.(Y537N)) were assessed in available baseline plasma samples of 21 patients with ER positive recurrent breast cancer. Inclusion criteria for study participation were: female, age above 18 years, ER positive breast cancer, 5years adjuvant hormonal therapy of primary disease, and disease recurrence or metastasis during or after stop of endocrine therapy. ESR1 mutations were analyzed in cell-free DNA (cfDNA) by using digital droplet PCR (ddPCR). RESULTS: cfDNA was obtained from 21 patients with recurrent breast cancer. ESR1 mutations were found in 4/21 (19%; 95% CI, 5%-42%). The test sensitivity was lower than the targeted value <0.1% in 29% of patients (6/21). No significant statistical difference in baseline clinical characteristics was observed in patients with wild-type and mutant ER (p>0.05). Adjuvant endocrine therapy for primary disease was Tamoxifen (TAM) for 57% of patients (12 of 21) of whom 8 patients had received aromatase inhibitor (AI) after two years, while 43% of patients (9 of 21) had received AI as first line adjuvant hormonal therapy. All the patients had received aromatase inhibitor AI therapy in first or second line therapy with initially a variable period of good response. CONCLUSION: ESR1 mutation analysis could be determined in archived plasma samples using simple non-invasive methods. In the future, screening for mutation status could improve the therapeutic strategies in controlling ER signaling before the occurrence of wide spread disease metastasis.

Human Rab7 mutation mimics features of Charcot-Marie-Tooth neuropathy type 2B in Drosophila.

Charcot-Marie-Tooth disease type 2B (CMT2B) is an inherited axonal peripheral neuropathy. It is characterised by prominent sensory loss, often complicated by severe ulcero-mutilations of toes or feet, and variable motor involvement. Missense mutations in RAB7A, the gene encoding the small GTPase Rab7, cause CMT2B and increase Rab7 activity. Rab7 is ubiquitously expressed and is involved in degradation through the lysosomal pathway. In the neurons, Rab7 plays a role in the long-range retrograde transport of signalling endosomes in the axons. Here we developed the first animal model of CMT2B, modelling one of the mutations (L129F) in Drosophila melanogaster. Behavioural assays show that this model recapitulates several hallmarks of the human disease. Upon expression of mutant Rab7 in the sensory neurons, larvae present with a reduction of temperature and pain perception. Furthermore, the larvae exhibit a crawling defect when the mutant protein is expressed in the motor neurons. Analysis of axonal transport of Rab7 positive vesicles in sensory neurons of Drosophila larvae and in neurites of mammalian neuroblastoma cells demonstrates that mutant vesicles pause less than their wild-type counterparts. This latter finding indicates that alterations in vesicle transport might contribute to the pathomechanism of CMT2B.

HSPB1 facilitates the formation of non-centrosomal microtubules.

The remodeling capacity of microtubules (MT) is essential for their proper function. In mammals, MTs are predominantly formed at the centrosome, but can also originate from non-centrosomal sites, a process that is still poorly understood. We here show that the small heat shock protein HSPB1 plays a role in the control of non-centrosomal MT formation. The HSPB1 expression level regulates the balance between centrosomal and non-centrosomal MTs. The HSPB1 protein can be detected specifically at sites of de novo forming non-centrosomal MTs, while it is absent from the centrosomes. In addition, we show that HSPB1 binds preferentially to the lattice of newly formed MTs in vitro, suggesting that its function occurs by stabilizing MT seeds. Our findings open new avenues for the understanding of the role of HSPB1 in the development, maintenance and protection of cells with specialized non-centrosomal MT arrays.

The neuroinflammatory role of Schwann cells in disease.

Peripheral neuropathies are associated with a variety of clinical symptoms ranging from motor and sensory symptoms to autonomic dysfunction. The primary disease causes for peripheral nerve disorders are also very heterogeneous, including genetic causes, inflammation mediated damage and physical trauma. A common theme in these neuropathies is the important contribution of the immune system; leading either to a deterioration or an amelioration of the disease. Immune responses are typically mediated by immune cells such as antigen-presenting cells, macrophages or T-cells. However, also non-immune cells such as microglia in the central nervous system or Schwann cells in the peripheral nervous system might play a key role in innate and adaptive immune responses. Just like microglia, Schwann cells express a plethora of pattern recognition receptors that allows them to recognize exogenous as well as endogenous danger signals. Upon activation, Schwann cells initiate and regulate local immune responses by presenting antigens and by secreting cytokines and chemokines, which will further attract immune cells to the site of injury. By interacting with immune cells they contribute in shaping immune responses that can lead to inflammatory neuropathies. In hereditary neuropathies, the immune system has also been shown to aggravate the disease phenotype. Besides, a neuroprotective role for the immune system has been recognized that becomes mainly prominent in cases of acute nerve injury. The present review focuses on the recently recognized immune competent role of Schwann cells and its involvement in peripheral neuropathies.

Acute injury in the peripheral nervous system triggers an alternative macrophage response.

BACKGROUND: The activation of the immune system in neurodegeneration has detrimental as well as beneficial effects. Which aspects of this immune response aggravate the neurodegenerative breakdown and which stimulate regeneration remains an open question. To unravel the neuroprotective aspects of the immune system we focused on a model of acute peripheral nerve injury, in which the immune system was shown to be protective. METHODS: To determine the type of immune response triggered after axotomy of the sciatic nerve, a model for Wallerian degeneration in the peripheral nervous system, we evaluated markers representing the two extremes of a type I and type II immune response (classical vs. alternative) using real-time quantitative polymerase chain reaction (RT-qPCR), western blot, and immunohistochemistry. RESULTS: Our results showed that acute peripheral nerve injury triggers an anti-inflammatory and immunosuppressive response, rather than a pro-inflammatory response. This was reflected by the complete absence of classical macrophage markers (iNOS, IFN γ, and IL12p40), and the strong up-regulation of tissue repair markers (arginase-1, Ym1, and Trem2). The signal favoring the alternative macrophage environment was induced immediately after nerve damage and appeared to be established within the nerve, well before the infiltration of macrophages. In addition, negative regulators of the innate immune response, as well as the anti-inflammatory cytokine IL-10 were induced. The strict regulation of the immune system dampens the potential tissue damaging effects of an over-activated response. CONCLUSIONS: We here demonstrate that acute peripheral nerve injury triggers an inherent protective environment by inducing the M2 phenotype of macrophages and the expression of arginase-1. We believe that the M2 phenotype, associated with a sterile inflammatory response and tissue repair, might explain their neuroprotective capacity. As such, shifting the neurodegeneration-induced immune responses towards an M2/Th2 response could be an important therapeutic strategy.

Small heat-shock protein HSPB1 mutants stabilize microtubules in Charcot-Marie-Tooth neuropathy.

Mutations in the small heat shock protein HSPB1 (HSP27) are causative for Charcot-Marie-Tooth (CMT) neuropathy. We previously showed that a subset of these mutations displays higher chaperone activity and enhanced affinity to client proteins. We hypothesized that this excessive binding property might cause the HSPB1 mutant proteins to disturb the function of proteins essential for the maintenance or survival of peripheral neurons. In the present work, we explored this hypothesis further and compared the protein complexes formed by wild-type and mutant HSPB1. Tubulin came out as the most striking differential interacting protein, with hyperactive mutants binding more strongly to both tubulin and microtubules. This anomalous binding leads to a stabilization of the microtubule network in a microtubule-associated protein-like manner as reflected by resistance to cold depolymerization, faster network recovery after nocodazole treatment, and decreased rescue and catastrophe rates of individual microtubules. In a transgenic mouse model for mutant HSPB1 that recapitulates all features of CMT, we could confirm the enhanced interaction of mutant HSPB1 with tubulin. Increased stability of the microtubule network was also clear in neurons isolated from these mice. Since neuronal cells are particularly vulnerable to disturbances in microtubule dynamics, this mechanism might explain the neuron-specific CMT phenotype caused by HSPB1 mutations.

Mutant HSPB8 causes motor neuron-specific neurite degeneration.

Missense mutations (K141N and K141E) in the alpha-crystallin domain of the small heat shock protein HSPB8 (HSP22) cause distal hereditary motor neuropathy (distal HMN) or Charcot-Marie-Tooth neuropathy type 2L (CMT2L). The mechanism through which mutant HSPB8 leads to a specific motor neuron disease phenotype is currently unknown. To address this question, we compared the effect of mutant HSPB8 in primary neuronal and glial cell cultures. In motor neurons, expression of both HSPB8 K141N and K141E mutations clearly resulted in neurite degeneration, as manifested by a reduction in number of neurites per cell, as well as in a reduction in average length of the neurites. Furthermore, expression of the K141E (and to a lesser extent, K141N) mutation also induced spheroids in the neurites. We did not detect any signs of apoptosis in motor neurons, showing that mutant HSPB8 resulted in neurite degeneration without inducing neuronal death. While overt in motor neurons, these phenotypes were only very mildly present in sensory neurons and completely absent in cortical neurons. Also glial cells did not show an altered phenotype upon expression of mutant HSPB8. These findings show that despite the ubiquitous presence of HSPB8, only motor neurons appear to be affected by the K141N and K141E mutations which explain the predominant motor neuron phenotype in distal HMN and CMT2L.

Toll-like receptor expression in the peripheral nerve.

Toll-like receptors comprise a family of evolutionary conserved pattern recognition receptors that act as a first defense line in the innate immune system. Upon stimulation with microbial ligands, they orchestrate the induction of a host defense response by activating different signaling cascades. Interestingly, they appear to detect the presence of endogenous signals of danger as well and as such, neurodegeneration is thought to trigger an immune response through ligation of TLRs. Though recent data report the expression of various TLRs in the central nervous system, TLR expression patterns in the peripheral nervous system have not been determined yet. We observed that Schwann cells express relatively high levels of TLRs, with especially TLR3 and TLR4 being prominent. Sensory and motor neurons hardly express TLRs at all. Through the use of NF-κB signaling as read-out, we could show that all TLRs are functional in Schwann cells and that bacterial lipoprotein, a ligand for TLR1/TLR2 receptors yields the strongest response. In sciatic nerve, basal levels of TLRs closely reflect the expression patterns as determined in Schwann cells. TLR3, TLR4, and TLR7 are majorly expressed, pointing to their possible role in immune surveillance. Upon axotomy, TLR1 becomes strongly induced, while most other TLR expression levels remain unaffected. Altogether, our data suggest that similar to microglia in the brain, Schwann cells might act as sentinel cells in the PNS. Furthermore, acute neurodegeneration induces a shift in TLR expression pattern, most likely illustrating specialized functions of TLRs in basal versus activated conditions of the peripheral nerve.

Increased monomerization of mutant HSPB1 leads to protein hyperactivity in Charcot-Marie-Tooth neuropathy.

Small heat shock proteins are molecular chaperones capable of maintaining denatured proteins in a folding-competent state. We have previously shown that missense mutations in the small heat shock protein HSPB1 (HSP27) cause distal hereditary motor neuropathy and axonal Charcot-Marie-Tooth disease. Here we investigated the biochemical consequences of HSPB1 mutations that are known to cause peripheral neuropathy. In contrast to other chaperonopathies, our results revealed that particular HSPB1 mutations presented higher chaperone activity compared with wild type. Hyperactivation of HSPB1 was accompanied by a change from its wild-type dimeric state to a monomer without dissociation of the 24-meric state. Purification of protein complexes from wild-type and HSPB1 mutants showed that the hyperactive isoforms also presented enhanced binding to client proteins. Furthermore, we show that the wild-type HSPB1 protein undergoes monomerization during heat-shock activation, strongly suggesting that the monomer is the active form of the HSPB1 protein.

Purinergic signaling in the pulmonary neuroepithelial body microenvironment unraveled by live cell imaging.

Pulmonary neuroepithelial bodies (NEBs) are densely innervated groups of complex sensory airway receptors involved in the regulation of breathing. Together with their surrounding Clara-like cells, they exhibit stem cell potential through their capacity to regenerate depopulated areas of the epithelium following lung injury. We have employed confocal live cell imaging microscopy and novel electrophysiological techniques in a new ex vivo lung slice model to unravel potential purinergic signaling pathways within the NEB microenvironment. Quinacrine histochemistry indicated high amounts of vesicular ATP in NEB cells. Using a "reporter-patching" method adapted to create a uniquely sensitive and selective biosensor for the direct detection of ATP release from NEBs ex vivo, we demonstrated quantal ATP release from NEBs following their depolarization. Enhancing enzymatic extracellular ATP hydrolysis or inhibiting P2 receptors confirmed the central role of ATP in paracrine interactions between NEB cells and Clara-like cells. Combined calcium imaging, pharmacology, and immunohistochemistry showed that ligand-binding to functional P2Y(2) receptors underpins the activation of Clara-like cells. Hence, NEB cells communicate with their cellular neighbors in the NEB microenvironment by releasing ATP, which rapidly evokes purinergic activation of surrounding Clara-like cells. Besides ATP acting on the P2X(3) receptor expressing vagal sensory nerve terminals between NEB cells, local paracrine purinergic signaling within this potential stem cell niche may be important to both normal airway function, airway epithelial regeneration after injury, and/or the pathogenesis of small cell lung carcinomas.