Commentary on Some Recent Theses Relevant to Combating Aging: October 2018

Computational Pathology for Quantifying Spatial Heterogeneity in Digital Images of Tissue Sections from Solid Tumors

Luong Nguyen, PhD, University of Pittsburgh

Tumor heterogeneities have been linked to cancer patient outcomes and responses to therapies. With the increasing popularity of whole slide imaging systems and the development of hyperplexed immunofluorescence platforms, the volume of digital images of tissue sections from solid tumors is growing with an unprecedented rate. The era of big data poses a challenge to extracting useful information about tumor heterogeneity. This thesis presents computational pathology algorithms for modeling tumor heterogeneity in transmitted light and immunofluorescence digital images of breast and colon tissues.

In transmitted light images of hematoxylin and eosin stained tissue sections, we characterize tumor heterogeneity by the relative spatial arrangement of different histological structures in the tissues. To identify these structures, we developed two automated segmentation methods based on color statistics and internuclear distance distributions. After segmenting histological structures, we assign clinically relevant labels (e.g., invasive carcinoma, blood vessels, etc.) to them based on their cytological and architectural features. Finally, we quantify the spatial distributions of these histological structures, classify the whole image using the spatial characteristics, and validate our classification against pathologists' annotations. In immunofluorescence images, tumor spatial heterogeneity can be quantified using correlations between biomarker expressions of different cell types in a neighborhood of various sizes. We find that the biomarker correlations are superior to clinical histopathological covariates in terms of predicting recurrence status of colorectal cancer patients.

Comment: At the most fundamental level, cancer is an abnormal growth of cells—a deviation from the expected spatial organization of a given tissue. In theory, then, the disease can in general be diagnosed based simply on the presence of such an anomaly in histological images; yet this is far from a simple task, not least because the “expected” organization varies so greatly even in healthy tissue! With sufficient training and experience, oncologists can classify samples as benign or cancerous with good confidence. However, this approach is not truly scalable—as reflected by the perpetually short supply, and ever-growing cost, of such expertise—and the confidence achievable is bounded by the inability of any human, however gifted, to integrate all of the data available from rapidly advancing modern analytic methods. Fortunately, the field of machine learning is advancing with similar speed; it seems most probable that automated methods will be able to outperform human diagnosticians within the relatively near future (at least in cases where sufficient training data is available; thus, conveniently, the most common types of disease will tend to be the first to succumb). The results of this dissertation represent important progress along that trajectory—one which leads not only to faster and more reliable diagnoses for patients, but also liberates human talent to investigate rarer and more complex presentations.

Molecular Analysis of Amyotrophic Lateral Sclerosis and Frontotemporal Degeneration Brain Tissue Identifies Disease Mechanisms Associated with Repetitive DNA Elements

Elaine Liu, PhD, University of Pennsylvania

Aging-related neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS) and frontotemporal degeneration (FTD) are two fatal progressive neurodegenerative diseases that carry genetic and pathologic overlap: a hexanucleotide G₄C₂ repeat expansion in C9orf72 and loss of a nuclear RNA binding protein, TAR DNA binding protein-43 (TDP-43), into cytoplasmic aggregates. The C9orf72 expansion is the most common genetic cause of ALS/FTD and is associated with reduced C9orf72 expression and accumulation of toxic RNA and protein aggregates. In Chapter 2 of my thesis, using molecular analyses from human postmortem ALS/FTD brain, I show that the C9orf72 promoter is hypermethylated within a subset of expansion carriers. C9orf72 promoter hypermethylation is associated with reduced C9orf72 pathology and may be protective in these patients. Another key feature in ALS/FTD is the characteristic pathology of nuclear TDP-43 loss in degenerating neurons. TDP-43 is a ubiquitous nuclear RNA binding protein and is heavily involved in RNA processing. TDP-43 has been shown to bind genic elements and repetitive transposable elements such as long interspersed nuclear elements (LINE). Considering that TDP-43 is a ubiquitous RNA binding protein, I hypothesize that nuclear TDP-43 loss can lead to large transcriptomic changes and may contribute to alterations in LINE elements. For Chapters 3 and 4 of my thesis, I use a novel method of subcellular fractionation and fluorescent activated cell sorting (FACS) from postmortem ALS/FTD human brain to perform high-throughput sequencing analyses to study neuronal molecular changes. In Chapter 3 of my thesis, I use FACS coupled with RNA-seq on neuronal nuclei with and without TDP-43 to show that loss of nuclear TDP-43 is associated with large transcriptome changes and increased LINE accessibility. Furthermore, loss of nuclear TDP-43 leads to increased retrotransposition. I also extend this subcellular fractionation-FACS method to study the effects of the C9orf72 expansion in neuronal nuclei. In Chapter 4, I demonstrate that the C9orf72 expansion is linked to mild gene expression changes that reflect C9orf72 protein loss and not gain of toxic C9orf72 RNA. Through my work, I have shown disease mechanisms linked to repetitive DNA elements, in that I propose (1) the C9orf72 repeat expansion may contribute to disease primarily via a gain of toxic C9orf72 pathology (2) loss of neuronal nuclear TDP-43 may be associated with increase retrotransposon activity which may contribute to disease. Overall, my work has broadened the field of neurodegeneration in my implementation of cell-type specific molecular analyses on postmortem brain of ALS/FTD patients to identify disease mechanisms with the intent of discovering new therapeutics and biomarkers that can be extended into the clinic.

Comment: Although a large majority of dementia cases worldwide are due to Alzheimer's disease, several other neurodegenerative conditions have just as bleak of a prognosis—and in general their pathology is even less well understood. TDP-43 was first identified as disease-relevant in virtually all cases of ALS and in many cases of FTD (the second most prevalent dementia) just over a decade ago. Unfortunately, the protein has a far more complex network of interactions with other cellular machinery than amyloid-beta—particularly, binding to and regulating the production, splicing and transport of over 6000 RNA species—and thus it has proven especially difficult to unpick the pathogenic processes involved. Still, there has actually been remarkable progress over the last decade in describing core elements of the disease mechanism. While either over- or underexpression of TDP-43 can induce ALS/FTD-like symptoms, the loss of normal nuclear TDP-43 expression appears to precede the formation of mature cytoplasmic aggregates in disease models, and indeed correlates more closely with neurodegeneration than the presence of aggregates per se. This loss induces a significant dysregulation of autophagy, particularly the aggresome pathway and autophagosome-lysosome fusion, and thus very plausibly is permissive for the formation of the large inclusions of misfolded protein seen in clinical disease. In turn those inclusions have a well-established capacity to sequester wild-type TDP-43, creating a vicious cycle, and some evidence suggests they can be exported via exosomes to propagate the pathology to nearby cells. Less well-established is the cause of initial nuclear TDP-43 loss. One suggestion is based on the observation that the protein is a major component of stress granules, multiprotein complexes that sequester RNAs redundant for survival during periods of cellular stress; certainly, metabolic stress is associated with cognitive decline. ¹¹ TDP-43 in stress granules is detergent-resistant and prone to post-translational modification, both defining features of its proteinopathy, and under conditions of prolonged stress—conditions which of course become more frequent with increasing age—such granules might become the seed for pathological aggregates. However, attempts to identify an upregulation of stress granule activity in the pathogenesis of ALS/FTD have not been successful. This thesis presents a somewhat worrying alternative route by which the transient loss of nuclear TDP-43, consistent with merely background levels of cellular stress, might on occasion initiate a permanent pathological state: the derepression of LINEs and other transposable elements normally bound by TDP-43 is shown to lead to increased replication of those sequences, creating new binding sites with the potential to dilute proper function of the protein even once its nuclear abundance has returned to normal. Should this mechanism indeed prove to be central to the disease process, then direct repair is likely to be impractical—the only option may be to replace affected cells entirely ¹² —and the development of preventative approaches becomes all the more critical.

Neuroprotective Potential of the N-Terminal Beta Amyloid Peptide Fragment in the Neurodegeneration, Synaptic Dysfunction and Memory Deficits in Models of Alzheimer's Disease

Naghum Alfulaij, PhD, University of Hawai'i at Manoa

Beta amyloid (Aβ) plays a central role in the pathogenesis of Alzheimer's disease (AD). It is produced by the sequential cleavage of amyloid precursor protein (APP) by β- and γ-secretases. This amyloidogenic pathway produces peptides 38–42 amino acids in length, based on the sites cleaved by γ-secretase. Aβ_1–42 is the predominant peptide species found in neuritic plaques. Its accumulation and impaired clearance are associated with disease progression. An alternative pathway has been proposed wherein short, N-terminal Aβ fragments are produced. The N-terminal fragments are hydrophilic, making them soluble and less likely to aggregate into plaques. Aβ_1–15 is the N-terminal Aβ fragment of focus in this study. It is produced by the sequential cleavage of APP by β- and α-secretases followed by a carboxypeptidase. Aβ_1–15 is also upregulated in AD patients suggesting a possible compensatory switch between pathways to suppress Aβ_1–42 production. We have recently shown that Aβ_1–15 can enhance long-term potentiation (LTP) in wild-type hippocampal synapses at very low (fM) concentrations. Aβ_1–15 was also shown to reverse the LTP block incurred by Aβ_1–42, as well as rescue LTP deficits in APPswe mice. This fragment has a potent and effective signaling activity via nicotinic acetylcholine receptors. Taken together, these data support a neuromodulatory function and a possible neuroprotective action for Aβ_1–15. This study focuses on the effects of Aβ_1–15 on Aβ_1–42 toxicity in various neuronal models, including in vitro NG108-15 hybrid neuroblastoma cells, ex vivo hippocampal neuron cultures and in vivo wild-type and AD model-APP mice.

We explored Aβ_1–15 neuroprotection against Aβ-mediated neurodegeneration in vitro by incubating our differentiated neuroblastoma cell line with different treatment combinations and across different time points to examine both the potency of Aβ_1–15 as well as the different ways by which Aβ_1–15 might be affecting Aβ_1–42 toxicity. Specifically, we explored the possibilities of Aβ_1–15 priming by pretreating cells with Aβ_1–15 before the addition of the toxic Aβ_1–42, competition of the two fragments by combination treatment, or rescue experiments by adding Aβ_1–15 after addition of Aβ_1–42 to determine whether Aβ_1–15 can reverse or halt toxicity caused by Aβ_1–42. Cellular toxicity was assessed as oxidative stress (production of reactive oxygen species) and apoptotic cell death. Next, we investigated the neuroprotective effects of Aβ_1–15 ex vivo by establishing primary hippocampal neuron cultures to confirm neuroprotection in a more physiologically relevant neuronal model. We then explored the potential for the Aβ_1–15 to protect or reverse (rescue) synaptic dysfunction and memory deficits resulting from Aβ synaptotoxicity. Changes in synaptic plasticity were assessed by measuring LTP in acute mouse hippocampal slices. Furthermore, we inspected the potential for rescue by Aβ_1–15 of LTP in APPswe hippocampal slices known to have LTP deficits. Lastly, we explored the effect of Aβ_1–15 in different behavior paradigms on 5XFAD (familial Alzheimer's disease) mice, a model expressing APPswe, APP-London and APP-Florida mutant transgenes as well as two mutant presenilin (PS1) transgenes, which accumulates high levels of Aβ over an accelerated timeframe (months). The behaviors examined were those related to deficits observed AD, namely contextual fear conditioning, novel object recognition and elevated plus maze to examine effects on memory processing, recognition memory and anxiety.

We were able to show that the Aβ_1–15 protected against all measures of Aβ-triggered neurotoxicity and neuronal dysfunction: oxidative stress, DNA fragmentation, apoptotic cell death, synaptotoxicity and behavioral deficits. Notably, Aβ_1–15 prevented LTP inhibition caused by Aβ_1–42 treatment and, when injected into the hippocampus was able to rescue memory in contextual fear conditioning as well as decrease anxiety in the 5XFAD mice. To address preliminarily the possible molecular mechanisms underlying the rescue by Aβ_1–15 of memory deficits in the 5XFAD, we explored signaling pathways known to be involved in Aβ synaptotoxicity. We observed a substantial upregulation of the glutamate receptor GluR2 and phosphorylated CREB in mouse hippocampi injected with the Aβ_1–15, giving us insight into the specific actions of N-terminal fragment.

In summary, the data show that Aβ_1–15 fully protected against Aβ_1–42—induced cellular toxicity, synaptotoxicity and behavioral deficits. Taken together, the data support our hypothesis that the N-terminal fragment (Aβ_1–15) is not only neuroprotective against acute Aβ_1–42 toxicity, but also has the ability to rescue memory in 5XFAD mice, potentially introducing a new avenue for AD therapeutics.

Comment: Although the aggregation of amyloid-beta protein into insoluble plaques is undeniably linked to the progression of AD, therapies designed on the basis that those plaques are the direct cause of pathology—such as vaccines against the protein, designed to promote their turnover by the immune system—have largely failed in the clinic despite promising results in animal models. The explanation for this disparity remains unclear, but most contemporary theories suggest either that plaques are merely an epiphenomenon of the disease process (and indeed that the true process may be unrepresented in models engineered to display intense plaque pathology); or that their effects are in fact pleiotropic, such that removal of the peptide en masse yields no net benefit. This thesis lends substantial weight to the latter possibility by identifying a potent neuroprotective and restorative role for the Aβ_1–15 sub-fragment, which has to date been studied primarily as an alternative to full-length amyloid-beta in the context of vaccination. Whether these beneficial properties will prove to translate to human disease remains to be seen—as alluded above, there is ever-increasing doubt over the reliability of rodent models of AD—although intriguingly a decline in Aβ_1–15 abundance has already been reported in several other neurodegenerative conditions, particularly Parkinson's disease, supporting the notion of some physiological role. Should its neuroprotective effect be real then (while its use as a vaccine would then seem rather unwise) a new therapeutic approach presents itself: by deploying a sufficiently specific protease to liberate the fragment from aggregates in situ, we might convert amyloid-beta into its own treatment!

pH-triggered Self-Assembly of a PEGylated Peptide Amphiphilic Contrast Agent

Ashley Wallace, PhD, The Ohio State University

Self-assembling peptide amphiphiles (PAs) have gained significant interest in the area of “smart” nano-diagnostics and therapeutics due to their ability to form different morphologies in response to various physiological stimuli. For cancer targeting, it is ideal that these PAs respond to a broad hallmark associated with tumors and not specific biomarkers that evolve constantly. Developing PA based dynamic imaging agents that transform into a larger, more slowly-diffusing morphology in response to a generic cancer hallmark can allow for the selective accumulation of such agents in high concentrations at the tumor site to enhance the sensitivity and resolution of traditional imaging techniques.

One particularly attractive cancer hallmark is the acidic extracellular microenvironment (pH_e = 6.6–7.4) that results from the high rate of glycolytic metabolism in tumor tissues, as well as the enhanced permeation and retention of nano-sized diagnostic and therapeutic agents (ideally 10–30 nm) arising from their leaky vasculature system. We have previously developed PAs that respond to the differences in the pHe and physiological pH via undergoing a morphological transition from spherical micelles (10–15 nm) to nanofibers (>1 μm in length, 10–15 nm) in isotonic salt solutions simulating the ionic strength of blood serum. Using the solid phase technique, each PA was synthesized to incorporate four specific regions: a hydrophobic (alkyl) tail, a β-sheet region, a region of charged amino acids, and a magnetic resonance imaging (MRI) moiety that consists of a Gd-chelator, to enable this transition in the desired pH range. The small diameter of the micelles is expected to maximize tumor penetration, and the much larger, much more slowly diffusing nanofibers formed in the tumor's interstitial microenvironment will maximize retention, thereby allowing cancer detection with high signal and sensitivity.

PEGylation, or addition of polyethylene glycol units, to diagnostic and therapeutic vehicles has been extensively used to increase their blood circulation half-life. Covalent linkage of these units work by (1) providing a protective coat that reduces recognition, degradation, and elimination from the bloodstream by the body's immune system and (2) increasing the apparent size of the molecule as to reduce renal clearance. Previous Research have shown that the addition of these units modifies physical features of a molecule but do not affect their functions. Here, we investigate the effects of various PEGylation ratios on the pH-triggered self-assembly. Circular dichroism, critical aggregation concentration and transmission electron microscopy measurements will be used to characterize morphology transitions at various concentrations and pH values of the PA. Fluorescence anisotropy using a fluorophore-labeled PA will be used to probe self-assembly behavior in blood serum. Finally, the rate of water relaxation will be investigated to understand the efficiency of these agents to be used for MRI.

Comment: With over a hundred distinct types of cancer recognized (based on cell of origin; other approaches give even higher figures) and the infamous mutability of cancerous cells, there is an urgent need for diagnostic and/or therapeutic agents which can be applied across as wide as possible a range of disease types and stages. Microenvironmental acidity and vascular permeability are near-ubiquitous features of solid tumors and, while neither is unique to the cancerous stroma, their combined presence in the adult body signals neoplasia with high specificity relative to healthy tissue. (One significant “false positive” for this approach occurs in the context of non-cancerous inflammation, which also produces a leaky, acidic microenvironment. However, in this case the acidity originates from specific macrophage activity which could be transiently inhibited to further improve precision.) The rational design of self-assembling agents able to exploit this specificity to reliably accumulate in tumors—while having the flexibility to incorporate a range of cargo moieties without compromising that behavior—has proven challenging thanks to the extraordinary computational resources required to model groups of flexible molecules in solution. Yet the value of such pan-cancer targeting vehicles is difficult to overstate—in particular given their potential use in non-invasive screening for arbitrary presymptomatic disease, thereby enabling early treatment and the associated dramatic improvement in prognosis for patients across the entire spectrum of solid cancers. Thus we applaud the ingenious groundwork laid in this thesis, and look forward to further development of the approach described.

Quantitative Approaches for Profiling the T Cell Receptor Repertoire in Human Tissues

Boris Grinshpun, PhD, Columbia University

The study of B and T cell receptor repertoires from high throughput sequencing is a recent development that allows for unprecedented resolution and quantification of the adaptive immune response. The immense diversity and long tailed distribution of these repertoires has up until now limited such studies to expanded clonal signatures or to analysis of imprecise signals with limited dynamic range collected by techniques such as radioactive and fluorescent labeling. This thesis presents a number of quantitative methods to characterize the repertoire and examine the questions of sequence diversity and inter-repertoire divergence of T cell repertoires. These approaches attempt to accurately parametrize the inherent distribution of T cell clones drawing from statistical tools derived from ecological literature and information theory.

The methods presented are applied to T cell analyses of various tissue compartments of the human body, including peripheral blood mononucleocytes, thymic tissues, spleen, inguinal lymph nodes, lung lymph nodes and the brain. A number of applications are explored with strong implications for translational use in medicine. Novel insights are made into the mechanism of maintenance and compartmentalization of naïve T cells from human donors of many different ages. Diversity and divergence of the tumor infiltrating sequence repertoire is measured in low grade gliomas and glioblastomas from cancer patients, and potential sequence based biomarkers are assessed for studying glioma phenotype progression. A careful investigation of the immune response to allogeneic stimulus reveals the effect of HLA on sequence sharing and diversity of the alloresponse, and quantifies for the first time using sequence data the fraction of T cells in a repertoire that are alloreactive.

The use of repertoire sequencing and mathematical models within immunology is a new and emerging concept within the rapidly expanding field of systems immunology and will undoubtedly have a profound impact on the future of immunology research. It is hoped that the tools presented in this thesis will give insight into how to quantitatively explore the breadth and depth of the T cell receptor repertoire, and provide future directions for TCR repertoire analysis.

Comment: Immunosenescence—the age-related decline in the function of the immune system—is a ubiquitous feature of human aging (although at present unfortunately considered a clinical indication only in extreme cases) which leads both to a decrease in resistance to novel pathogens, and also to the destabilization of the finely-controlled surveillance mechanisms that simultaneously stem cancerous growth and block autoimmune attacks against healthy cells. These normal functions depend on the tremendous diversity of molecular receptors expressed by adaptive immune cells, and a major contributor to immunosenescence is the contraction of that diversity due to a dwindling production of naïve lymphocytes and the accumulation of dysfunctional (anergic) effector cells. There is good evidence that the latter factor is driven at least in part by chronic viral infections, but overall our understanding of the relationship between repertoire shrinkage and the loss of immunological resilience remains very limited for precisely the reasons discussed in the abstract above. The methods introduced in this thesis will enable a much more comprehensive analysis of how aging impacts the entire T cell population, and likewise how successfully any given therapy counteracts that effect, thus facilitating the recovery of youthful immune function (at least in conjunction with a technique for restoring naïve T cell production, an avenue of research which has made excellent progress over recent years ¹³ ). Further down the line, the periodic analysis of TCR diversity in a given patient may also have the potential to signal the presence of cancer, absent any other symptoms, simply by virtue of a sudden unexplained skew in repertoire resulting from the “immune editing” performed by growing tumors.

Regulation and Repair of Neural Stem Cells and the Neurogenic Niche

Michael Fatt, PhD, University of Toronto

In the adult mammalian brain, numerous complex regulatory networks exist to regulate adult neural precursor cell (NPC) maintenance throughout life. While many of these networks have been described, others remain less well understood. Here, we have asked about three of these pathways (specifically the p53 family, cellular senescence, and AMPK), using genetic and pharmacological tools to determine their roles in precursor maintenance and neurogenesis in vivo. First, we show that p53, p63, and p73 functionally interact in NPCs to regulate cell fate. When these interactions are perturbed due to combined p63;p73 heterozygosity, precursor pools and neurogenesis are rapidly depleted due to an increase in p53-mediated senescence. Second, we demonstrate that senescent NPCs rapidly accumulate in the subgranular zone (SGZ) with age, and this coincides with a decrease in neurogenesis and precursor numbers. When senescent cells in the SGZ are reduced either pharmacologically or genetically, neurogenesis and stem cell proliferation are dramatically enhanced, demonstrating a cell non-autonomous regulatory network between senescent and non-senescent precursors in the hippocampus. Last, we show that AMPK agonist metformin acts on two distinct pathways to regulate cell fate in adult neural precursors. Specifically, we demonstrate that metformin acts to enhance NPC proliferation and self-renewal via a TAp73-dependent mechanism. However, metformin can also promote NPC differentiation by activating an aPKC-CBP pathway, indicating that AMPK plays an important role in regulating cell fate decisions. Taken together, the results of this thesis demonstrate that complex regulatory networks including the p53 family, cellular senescence and AMPK act to maintain adult NPCs, and manipulation of these networks can drastically alter neurogenesis and precursor pools in vivo.

Comment: Of the seven strands of the SENS platform the ablation of senescent cells (ApoptoSENS) is by far the closest to clinical deployment, with multiple human clinical trials of senolytic drugs now either in progress or scheduled to start within the coming year. In no small part this surge has been prompted by the demonstration in model systems that the toxic effect of those cells—driving inflammation, suppressing progenitor cell functions, facilitating tumorigenesis, and more—accounts for a surprisingly large proportion of the total burden endured by aged tissue, such that eliminating this single type of age-related damage yields very broadly positive effects on health and survival. Nonetheless, there is still a great deal of work to be done; the early hope that senescence was a single well-defined pathway has been quashed by more recent data indicating that the phenotype has a modular composition and can be heterogeneously expressed even within a nominally clonal population, such that no single senolytic agent is ever likely to effectively clear all such cells from the body. To be confident that we have truly blocked the route to age-related disease via senescent cell accumulation, a full list of such cell types and their vulnerabilities to various senolytic mechanisms will be required (fortunately, at least one ambitious company is already working on just such a resource) in addition to an adequate repertoire of interventions. Meanwhile, to enable regulatory approval for therapies based on that information, the relationship between specific senescent populations and diseases will also need to be determined (at least pending the recognition of aging per se as a treatable indication). This thesis is an excellent example of just this type of research—identifying a novel senescent population that accumulates with age, and clearly linking its presence to pathology; in this case the age-related failure of hippocampal neurogenesis that is thought to underpin the cognitive decline seen even in the “healthy elderly” (and mitigated only partially by traditional methods such as exercise ¹⁴ ), and which crucially is also a feature of several recognized neurodegenerative diseases.