Deep Learning Applied to Raman Spectroscopy for the Detection of Microsatellite Instability/MMR Deficient Colorectal Cancer.

Defective DNA mismatch repair is one pathogenic pathway to colorectal cancer. It is characterised by microsatellite instability which provides a molecular biomarker for its detection. Clinical guidelines for universal testing of this biomarker are not met due to resource limitations; thus, there is interest in developing novel methods for its detection. Raman spectroscopy (RS) is an analytical tool able to interrogate the molecular vibrations of a sample to provide a unique biochemical fingerprint. The resulting datasets are complex and high-dimensional, making them an ideal candidate for deep learning, though this may be limited by small sample sizes. This study investigates the potential of using RS to distinguish between normal, microsatellite stable (MSS) and microsatellite unstable (MSI-H) adenocarcinoma in human colorectal samples and whether deep learning provides any benefit to this end over traditional machine learning models. A 1D convolutional neural network (CNN) was developed to discriminate between healthy, MSI-H and MSS in human tissue and compared to a principal component analysis-linear discriminant analysis (PCA-LDA) and a support vector machine (SVM) model. A nested cross-validation strategy was used to train 30 samples, 10 from each group, with a total of 1490 Raman spectra. The CNN achieved a sensitivity and specificity of 83% and 45% compared to PCA-LDA, which achieved a sensitivity and specificity of 82% and 51%, respectively. These are competitive with existing guidelines, despite the low sample size, speaking to the molecular discriminative power of RS combined with deep learning. A number of biochemical antecedents responsible for this discrimination are also explored, with Raman peaks associated with nucleic acids and collagen being implicated.

Machine Learning of Raman Spectroscopy Data for Classifying Cancers: A Review of the Recent Literature.

Raman Spectroscopy has long been anticipated to augment clinical decision making, such as classifying oncological samples. Unfortunately, the complexity of Raman data has thus far inhibited their routine use in clinical settings. Traditional machine learning models have been used to help exploit this information, but recent advances in deep learning have the potential to improve the field. However, there are a number of potential pitfalls with both traditional and deep learning models. We conduct a literature review to ascertain the recent machine learning methods used to classify cancers using Raman spectral data. We find that while deep learning models are popular, and ostensibly outperform traditional learning models, there are many methodological considerations which may be leading to an over-estimation of performance; primarily, small sample sizes which compound sub-optimal choices regarding sampling and validation strategies. Amongst several recommendations is a call to collate large benchmark Raman datasets, similar to those that have helped transform digital pathology, which researchers can use to develop and refine deep learning models.

Engineering transplantable jejunal mucosal grafts using patient-derived organoids from children with intestinal failure.

Intestinal failure, following extensive anatomical or functional loss of small intestine, has debilitating long-term consequences for children1. The priority of patient care is to increase the length of functional intestine, particularly the jejunum, to promote nutritional independence2. Here we construct autologous jejunal mucosal grafts using biomaterials from pediatric patients and show that patient-derived organoids can be expanded efficiently in vitro. In parallel, we generate decellularized human intestinal matrix with intact nanotopography, which forms biological scaffolds. Proteomic and Raman spectroscopy analyses reveal highly analogous biochemical profiles of human small intestine and colon scaffolds, indicating that they can be used interchangeably as platforms for intestinal engineering. Indeed, seeding of jejunal organoids onto either type of scaffold reliably reconstructs grafts that exhibit several aspects of physiological jejunal function and that survive to form luminal structures after transplantation into the kidney capsule or subcutaneous pockets of mice for up to 2 weeks. Our findings provide proof-of-concept data for engineering patient-specific jejunal grafts for children with intestinal failure, ultimately aiding in the restoration of nutritional autonomy.

Rapid and complete paraffin removal from human tissue sections delivers enhanced Raman spectroscopic and histopathological analysis.

Incomplete removal of paraffin and organic contaminants from tissues processed for diagnostic histology has been a profound barrier to the introduction of Raman spectroscopic techniques into clinical practice. We report a route to rapid and complete paraffin removal from a range of formalin-fixed paraffin embedded tissues using super mirror stainless steel slides. The method is equally effective on a range of human and animal tissues, performs equally well with archived and new samples and is compatible with standard pathology lab procedures. We describe a general enhancement of the Raman scatter and enhanced staining with antibodies used in immunohistochemistry for clinical diagnosis. We conclude that these novel slide substrates have the power to improve diagnosis through anatomical pathology by facilitating the simultaneous combination of improved, more sensitive immunohistochemical staining and simplified, more reliable Raman spectroscopic imaging, analysis and signal processing.

Activation of G Proteins by Guanine Nucleotide Exchange Factors Relies on GTPase Activity.

G proteins are an important family of signalling molecules controlled by guanine nucleotide exchange and GTPase activity in what is commonly called an 'activation/inactivation cycle'. The molecular mechanism by which guanine nucleotide exchange factors (GEFs) catalyse the activation of monomeric G proteins is well-established, however the complete reversibility of this mechanism is often overlooked. Here, we use a theoretical approach to prove that GEFs are unable to positively control G protein systems at steady-state in the absence of GTPase activity. Instead, positive regulation of G proteins must be seen as a product of the competition between guanine nucleotide exchange and GTPase activity--emphasising a central role for GTPase activity beyond merely signal termination. We conclude that a more accurate description of the regulation of G proteins via these processes is as a 'balance/imbalance' mechanism. This result has implications for the understanding of intracellular signalling processes, and for experimental strategies that rely on modulating G protein systems.

Modelling the effects of inhibitors of guanine nucleotide synthesis: implications for studies of cellular differentiation pathways.

Mizoribine induces the differentiation of promyelocytes by an unknown mechanism that relies on compromised guanine nucleotide synthesis. I have found that mizoribine also perturbs adenosine nucleotide levels in HL-60 promyelocytes, particularly ATP. To reconcile these observations with the known actions of mizoribine I have adapted an existing model of human purine metabolism composed as an S-system familiar from Biochemical Systems Theory. Mizoribine's actions were then simulated and compared with experimental data.

Developmentally acquired PKA localisation in mouse oocytes and embryos.

Localisation of Protein Kinase A (PKA) by A-Kinase Anchoring Proteins (AKAPs) is known to coordinate localised signalling complexes that target cAMP-mediated signalling to specific cellular sub-domains. The cAMP PKA signalling pathway is implicated in both meiotic arrest and meiotic resumption, thus spatio-temporal changes in PKA localisation during development may determine the oocytes response to changes in cAMP. In this study we aim to establish whether changes in PKA localisation occur during oocyte and early embryo development. Using fluorescently-labelled PKA constructs we show that in meiotically incompetent oocytes PKA is distributed throughout the cytoplasm and shows no punctuate localisation. As meiotic competence is acquired, PKA associates with mitochondria. Immature germinal vesicle (GV) stage oocytes show an aggregation of PKA around the GV and PKA remains co-localised with mitochondria throughout oocyte maturation. After fertilisation, the punctuate, mitochondrial distribution was lost, such that by the 2-cell stage there was no evidence of PKA localisation. RT-PCR and Western blotting revealed two candidate AKAPs that are known to be targeted to mitochondria, AKAP1 and D-AKAP2. In summary these data show a dynamic regulation of PKA localisation during oocyte and early embryo development.

The differential regulation of phosphatidylinositol 4-phosphate 5-kinases and phospholipase D1 by ADP-ribosylation factors 1 and 6.

Phosphatidylinositol 4-phosphate 5-kinases [PtdIns4P5Ks] synthesise the majority of cellular phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P(2)] and phospholipase D1 (PLD1) synthesises large amounts of phosphatidic acid (PtdOH). The activities of PtdIns4P5Ks and PLDs are thought to be coupled during cell signalling in order to support large simultaneous increases in both PtdIns(4,5)P(2) and PtdOH, since PtdOH activates PtdIns4P5Ks and PLD1 requires PtdIns(4,5)P(2) as a cofactor. However, little is known about the control of such a system. Membrane recruitment of ADP-ribosylation factors (Arfs) activates both PtdIns4P5Ks and PLDs, but it is not known if each enzyme is controlled in series by different Arfs or in parallel by a single form. We show through pull-down and vesicle sedimentation interaction assays that PtdIns4P5K activation may be facilitated by Arf-enhanced membrane association. However PtdIns4P5Ks discriminate poorly between near homogeneously myristoylated Arf1 and Arf6 although examples of all three known active isoforms (mouse alpha>beta, gamma) respond to these G-proteins. Conversely PLD1 genuinely prefers Arf1 and so the two lipid metabolising enzymes are differentially controlled. We propose that isoform selective Arf/PLD interaction and not Arf/PtdIns4P5K will be the critical trigger in the formation of distinct, optimal triples of Arf/PLDs/PtdIns4P5Ks and be the principle regulator of any coupled increases in the signalling lipids PtdIns(4,5)P(2) and PtdOH.

Two distinct chloride ion requirements in the constitutive protein secretory pathway.

The role of chloride ions in regulated secretion is well described but remains poorly characterised in the constitutive system. In the liver, newly synthesised proalbumin is transported to the trans Golgi network where it is converted to albumin by a furin protease and then immediately secreted. We used this acid-dependent hydrolysis and the measurement of specific protein secretion rates to examine the H+ and Cl- ion dependence of albumin synthesis and secretion, a major constitutive protein secretory event in all mammals. Using permeabilised primary rat hepatocytes we show that ordinarily chloride ions are essential for the processing of proalbumin to albumin. However Cl- is not required for transport which continues but releases solely proalbumin. Prior treatment of the cells with Tris (used as a membrane-permeable weak base to neutralise Golgi luminal pH) both eliminated the formation of albumin and very greatly reduced secretion. After washing out Tris, both authentic secretion and processing could be restarted if Cl-, ATP, GTP, cAMP, Ca2+ and cytosolic proteins were added. Hence a requirement for chloride in transport, in addition to processing, can be uncovered by first neutralising pH gradients. Furthermore, the chloride channel blocker DIDS (4,4-diisothiocyanostilbene 2,2-disulphonic acid) reversibly inhibited the constitutive secretory pathway. However, the total mass of proalbumin detectable in DIDS-treated cells fell to 36% of control while the fraction processed to albumin remained almost constant. This clearly dissociates a large part of the Cl- requirement of the constitutive protein secretory pathway from the function of known liver Golgi Cl- channels.

Xenopus apyrase (xapy), a secreted nucleotidase that is expressed during early development.

We have characterized a cDNA encoding a Xenopus laevis apyrase (XAPY) that is expressed during embryogenesis. XAPY is highly homologous to two recently described mammalian apyrases, human SCAN-1 and rat Ca2+-NDPase, and to a lesser extent the salivary apyrase of the blood-feeding arthropod Cimex lectularis. RT-PCR analysis shows that Xapy is expressed at all the developmental stages tested, from oocytes through to tadpoles. Xapy transcripts are widely distributed in the embryo, but from late neurulae through to late tailbud stages they are highly enriched in the cement gland, an adhesive organ in the epidermis of the head. When expressed in HEK 293 cells, XAPY is largely retained in the endoplasmic reticulum, although some is also secreted. XAPY conditioned media hydrolyses UDP and UTP, confirming that it is a functional apyrase.

Constitutive secretion of serum albumin requires reversible protein tyrosine phosphorylation events in trans-Golgi.

Serum albumin secretion from rat hepatocytes proceeds via the constitutive pathway. Although much is known about the role of protein tyrosine phosphorylation in regulated secretion, nothing is known about its function in the constitutive process. Here we show that albumin secretion is inhibited by the tyrosine kinase inhibitor genistein but relatively insensitive to subtype-selective inhibitors or treatments. Secretion is also blocked in a physiologically identical manner by the tyrosine phosphatase inhibitors pervanadate and bisperoxo(1,10-phenanthroline)-oxovanadate. Inhibition of either the kinase(s) or phosphatase(s) leads to the accumulation of albumin between the trans-Golgi and the plasma membrane, whereas the immediate precursor proalbumin builds up in a proximal compartment. The trans-Golgi marker TGN38 is rapidly dispersed under conditions that inhibit tyrosine phosphatase action, whereas the distribution of the cis-Golgi marker GM130 is insensitive to genistein or pervanadate. By using a specifically reactive biotinylation probe, we detected protein tyrosine phosphatases in highly purified rat liver Golgi membranes. These membranes also contain both endogenous tyrosine kinases and their substrates, indicating that enzymes and substrates for reversible tyrosine phosphorylation are normal membrane-resident components of this trafficking compartment. In the absence of perturbation of actin filaments and microtubules, we conclude that reversible protein tyrosine phosphorylation in the trans-Golgi network is essential for albumin secretion and propose that the constitutive secretion of albumin is in fact a regulated process.

Interaction of neuronal calcium sensor-1 and ADP-ribosylation factor 1 allows bidirectional control of phosphatidylinositol 4-kinase beta and trans-Golgi network-plasma membrane traffic.

We have identified a novel Ca(2+)-dependent interaction between neuronal calcium sensor-1 (NCS-1) and the GTPase ARF1. Both of these proteins are localized to the Golgi complex, and both regulate phosphatidylinositol 4-kinase IIIbeta (PI(4)Kbeta). Spatial and temporal control of phosphatidylinositol 4-phosphate levels through activation of PI(4)Kbeta is important for the recruitment of trafficking complexes to the trans-Golgi network (TGN) and vesicular traffic from this organelle. The NCS-1-ARF1 interaction and its specificity have been demonstrated through in vitro binding assays, in vitro enzyme assay, and through functional cellular assays. We show that NCS-1 can exert bidirectional effects to activate PI(4)Kbeta on its own or inhibit the activation by ARF1. NCS-1 was shown to modulate the effects of expression of ARF mutants that disrupt Golgi morphology and to recruit GDP-loaded ARF to the Golgi complex in a Ca(2+)-dependent manner. We demonstrate antagonist effects of NCS-1 and ARF on constitutive and regulated exocytosis. The NCS-1-ARF1 interaction provides evidence for functional cross-talk between Ca(2+)-dependent and ARF-dependent pathways in TGN to plasma membrane traffic.

Preparation of myristoylated Arf1 and Arf6.

Arf proteins are members of the Arf family of small Ras-like GTP binding proteins. Six Arfs, grouped into three classes, have been identified in mammalian cells and three members have been identified in yeasts. Arf1 and Arf6, more extensively studied than other Arfs, have been found to affect membrane traffic and actin remodeling. A structural feature that distinguishes Arfs from other Ras superfamily members is an N-terminal alpha-helix, extending from the basic G-protein fold, which is cotranslationally myristoylated. Both the helix and the myristate affect biochemical properties of Arfs, including nucleotide exchange, membrane association, and interaction with some effector proteins. Preparation of myristoylated Arf for in vitro studies of Arf function requires consideration of both the reaction yielding myristoylated protein and the properties of the modified Arfs. Here, we describe methods that yield homogeneous preparations of myristoylated Arf1 and Arf6.