Longitudinal machine learning uncouples healthy aging factors from chronic disease risks

To understand human longevity, inherent aging processes must be distinguished from known etiologies leading to age-related chronic diseases. Such deconvolution is difficult to achieve because it requires tracking patients throughout their entire lives. Here, we used machine learning to infer health trajectories over the entire adulthood age range using extrapolation from electronic medical records with partial longitudinal coverage. Using this approach, our model tracked the state of patients who were healthy and free from known chronic disease risk and distinguished individuals with higher or lower longevity potential using a multivariate score. We showed that the model and the markers it uses performed consistently on data from Israeli, British and US populations. For example, mildly low neutrophil counts and alkaline phosphatase levels serve as early indicators of healthy aging that are independent of risk for major chronic diseases. We characterize the heritability and genetic associations of our longevity score and demonstrate at least 1 year of extended lifespan for parents of high-scoring patients compared to matched controls. Longitudinal modeling of healthy individuals is thereby established as a tool for understanding healthy aging and longevity.

Early Infiltration of Innate Immune Cells to the Liver Depletes HNF4α and Promotes Extrahepatic Carcinogenesis

Multiple studies have identified metabolic changes within the tumor and its microenvironment during carcinogenesis. Yet, the mechanisms by which tumors affect the host metabolism are unclear. We find that systemic inflammation induced by cancer leads to liver infiltration of myeloid cells at early extrahepatic carcinogenesis. The infiltrating immune cells via IL6-pSTAT3 immune-hepatocyte cross-talk cause the depletion of a master metabolic regulator, HNF4α, consequently leading to systemic metabolic changes that promote breast and pancreatic cancer proliferation and a worse outcome. Preserving HNF4α levels maintains liver metabolism and restricts carcinogenesis. Standard liver biochemical tests can identify early metabolic changes and predict patients' outcomes and weight loss. Thus, the tumor induces early metabolic changes in its macroenvironment with diagnostic and potentially therapeutic implications for the host.

SIGNIFICANCE

Cancer growth requires a permanent nutrient supply starting from early disease stages. We find that the tumor extends its effect to the host's liver to obtain nutrients and rewires the systemic and tissue-specific metabolism early during carcinogenesis. Preserving liver metabolism restricts tumor growth and improves cancer outcomes. This article is highlighted in the In This Issue feature, p. 1501.

Functional enhancers at the gene-poor 8q24 cancer-linked locus

Multiple discrete regions at 8q24 were recently shown to contain alleles that predispose to many cancers including prostate, breast, and colon. These regions are far from any annotated gene and their biological activities have been unknown. Here we profiled a 5-megabase chromatin segment encompassing all the risk regions for RNA expression, histone modifications, and locations occupied by RNA polymerase II and androgen receptor (AR). This led to the identification of several transcriptional enhancers, which were verified using reporter assays. Two enhancers in one risk region were occupied by AR and responded to androgen treatment; one contained a single nucleotide polymorphism (rs11986220) that resides within a FoxA1 binding site, with the prostate cancer risk allele facilitating both stronger FoxA1 binding and stronger androgen responsiveness. The study reported here exemplifies an approach that may be applied to any risk-associated allele in non-protein coding regions as it emerges from genome-wide association studies to better understand the genetic predisposition of complex diseases.

Genome-wide scans of inherited genetic variation in the normal population have recently identified many sites (loci) associated with the predisposition to complex diseases such as cancer. Some of these cancer-associated loci, however, are devoid of genes (situated in so-called “gene deserts”) and the mechanism(s) of the association are not readily apparent. In the work reported here, we show that loci associated with several cancers in a gene desert found at chromosomal area 8q24 have embedded regulatory sequences affecting gene expression as enhancers, and in one case this activity is modulated by genetic variation. The results provide insight into the mechanism(s) governing genetic cancer risk.

Functional anatomy of polycomb and trithorax chromatin landscapes in Drosophila embryos

Polycomb group (PcG) and trithorax group (trxG) proteins are conserved chromatin factors that regulate key developmental genes throughout development. In Drosophila, PcG and trxG factors bind to regulatory DNA elements called PcG and trxG response elements (PREs and TREs). Several DNA binding proteins have been suggested to recruit PcG proteins to PREs, but the DNA sequences necessary and sufficient to define PREs are largely unknown. Here, we used chromatin immunoprecipitation (ChIP) on chip assays to map the chromosomal distribution of Drosophila PcG proteins, the N- and C-terminal fragments of the Trithorax (TRX) protein and four candidate DNA-binding factors for PcG recruitment. In addition, we mapped histone modifications associated with PcG-dependent silencing and TRX-mediated activation. PcG proteins colocalize in large regions that may be defined as polycomb domains and colocalize with recruiters to form several hundreds of putative PREs. Strikingly, the majority of PcG recruiter binding sites are associated with H3K4me3 and not with PcG binding, suggesting that recruiter proteins have a dual function in activation as well as silencing. One major discriminant between activation and silencing is the strong binding of Pleiohomeotic (PHO) to silenced regions, whereas its homolog Pleiohomeotic-like (PHOL) binds preferentially to active promoters. In addition, the C-terminal fragment of TRX (TRX-C) showed high affinity to PcG binding sites, whereas the N-terminal fragment (TRX-N) bound mainly to active promoter regions trimethylated on H3K4. Our results indicate that DNA binding proteins serve as platforms to assist PcG and trxG binding. Furthermore, several DNA sequence features discriminate between PcG- and TRX-N–bound regions, indicating that underlying DNA sequence contains critical information to drive PREs and TREs towards silencing or activation.

Although all cells of a developing organism have the same DNA, they express different genes and transmit these gene expression patterns to daughter cells through multiple rounds of cell division. This cellular memory for gene expression states is maintained by two groups of proteins: Polycomb-group proteins (PcG), which establish and maintain stable gene silencing, and trithorax group proteins (trxG), which counteract silencing and enable gene activation. It is unknown how this balance works and how exactly these proteins are recruited to their target sequences. By mapping the genome-wide distribution of PcG and trxG factors and proteins known to recruit them to chromatin, we found that putative PcG recruiters are not only colocalized at PcG binding sites, but also bind to many other genomic regions that are actually the binding sites of the Trithorax complex. We identified new DNA sequences important for the recruitment of both PcG and trxG proteins and showed that the differential binding of the recruiters PHO and PHOL may discriminate between active and inactive regions. Finally, we found that the two fragments of the Trithorax protein have different chromosomal distributions, suggesting that they may have distinct nuclear functions.

Comparison of the genome-wide distribution of PcG, trxG, and sequence-specific DNA binding proteins allowed the identification of key signals leading to Polycomb or Trithorax recruitment.