See article, p. 34

  • Epigenetic silencing of STING impairs dsDNA sensing in LKB1-null KRAS-mutant lung cancer.

  • LKB1 loss results in SAM-mediated activation of DNMT1 and EZH2, which repress STING and PD-L1.

  • Restoring STING could enhance immune checkpoint blockade in KRAS-mutant/LKB1-null lung cancer.


KRAS-mutant/LKB1-null (KL) lung cancers are highly aggressive, PD-L1–negative tumors which do not respond well to immune checkpoint blockade. To elucidate the role of LKB1 loss in KL lung tumorigenesis, Kitajima and colleagues performed gene set enrichment analysis of KL lung cancer gene expression data. KL lung cancer exhibited upregulation of gene expression signatures associated with oxidative stress and downregulation of interferon signaling; specifically, TMEM173, which encodes for the dsDNA sensor STING, was highly downregulated in KL lung cancer. Similarly, STING expression was decreased in KL non–small cell lung cancer tumors and lung cancer cell lines. Given that DNMT1 expression was increased in KL cell lines with low STING expression and that LKB loss is known to drive the upregulation of the epigenetic enzyme substrate SAM, a panel of inhibitors targeting epigenetic modifiers was screened against KL cell lines. Inhibition of DNMT1 and EZH2 restored STING expression in STING-low KL cell lines in a partially SAM-dependent manner, and the binding of DNMT1 to the TMEM173 promoter was increased in STING-null KL cell lines compared to STING-low KL cell lines. Ectopic expression of LKB1 combined with inhibition of DNMT1 restored STING expression and pathway activation, and a STING agonist restored PD-L1 expression in KL cell lines. Together, these results identify LKB1 as a negative regulator of STING in KRAS-driven lung cancer and suggest potential therapeutic strategies for patients with KRAS-mutant/LKB1-null lung cancer.

See article, p. 46

  • The genomic landscapes of osteosarcomas were characterized by whole-genome sequencing.

  • Patient-derived osteosarcoma xenografts were established and genomically characterized.

  • Osteosarcomas harbor subclass-specific recurrent SCNAs that harbor therapeutically targetable variants.


Osteosarcoma is the most common form of bone malignancy in patients 30 years old and younger. Although somatic copy-number alterations (SCNA) have been identified in this disease, recurrent SCNAs in targetable cancer pathways have not been comprehensively described. To ascertain whether recurrent SCNAs in osteosarcoma harbor potential therapeutic targets in osteosarcoma subclasses, Sayles and colleagues performed whole-genome sequencing (WGS) of primary and post-treatment osteosarcomas as well as osteosarcoma metastases (30 total samples). Analysis of the 30 genomes and a previously published WGS cohort of 33 samples revealed that osteosarcoma genomes exhibit significant interpatient, but not intrapatient, genomic heterogeneity. MYC and CCNE1 amplifications were the most frequently amplified genes in osteosarcoma, followed by RAD21, VEGFA, AURKB, and CDK4 amplifications. Fifteen patient-derived xenografts (PDX) were established from primary and metastatic osteosarcoma and recapitulate parental tumor SCNA profiles. Integrated genomic analyses demonstrated that osteosarcoma can be divided into subclasses that exhibit gain of MYC, CDK4/FOXM1, CCNE1, AURKB, and VEGFA, and PTEN loss/AKT gain. Targeting of these pathways with small-molecule inhibitors was more effective against the respectively genomically matched osteosarcoma PDX subclasses compared to non–genomically matched PDXs. Taken together, these findings further characterize the genomic landscape of osteosarcoma, describe the establishment and characterization of preclinical osteosarcoma PDX models, and identify a potential precision medicine approach for patients with this disease.

See article, p. 64

  • Loss of the secreted protein HAPLN1 in dermal fibroblasts during aging alters ECM organization.

  • ECM changes induced by HAPLN1 loss facilitate melanoma cell invasion and metastasis in aged skin.

  • In contrast, HAPLN1 loss in aged skin impairs immune cell infiltration into the tumor-associated ECM.


The structural organization of the skin extracellular matrix (ECM) changes with age, resulting in altered mechanical and physical properties. Changes in ECM stiffness and density have also been implicated in the regulation of tumor cell invasion, and recent work has suggested that factors secreted by dermal fibroblasts contribute to metastasis. Kaur and colleagues compared the secretomes of dermal fibroblasts from young and aged donors and identified hyaluronan and proteoglycan link protein 1 (HAPLN1), which cross-links and stabilizes ECM proteoglycans, as the most differentially expressed protein. HAPLN1 was abundantly secreted by young fibroblasts but was significantly downregulated during aging. In vitro cell-derived matrix assays and mouse experiments demonstrated that HAPLN1 expression was sufficient to mediate formation of a dense basket-weave matrix structure characteristic of that generated by young fibroblasts. Loss of HAPLN1 in dermal fibroblasts resulted in a less dense and more contractile matrix, which facilitated melanoma cell invasion in aged skin and enhanced lung metastasis. In contrast to its effects on melanoma cell migration, HAPLN1 expression promoted T-cell infiltration into the tumor microenvironment both in vitro and in vivo, which correlated with decreased primary tumor growth; this increase was potentially mediated by a decrease in the infiltration of polymorphonuclear myeloid-derived suppressor cells and regulatory T cells. Together, these findings provide further insight into the effects of age-related ECM changes on tumor progression and the immune microenvironment, with potential implications for therapeutic efficacy.

See article, p. 82

  • Lymphatic HAPLN1 expression is prognostic of improved overall survival in patients with melanoma.

  • HAPLN1 loss in aged fibroblasts promotes ECM degradation and enhances lymphatic permeability.

  • HAPLN1 treatment of aged lymph nodes decreases distant metastasis but increases lymphatic metastasis.


Clinical studies have suggested that although older patients with melanoma show a lower incidence of sentinel lymph node metastases, they have worse survival compared with younger patients. Ecker, Kaur, and colleagues analyzed sentinel node biopsies (SNB) from a large cohort of patients with melanoma and found that increasing age was associated with negative SNB but also with shorter distant metastasis–free survival and increased risk of distant recurrence. These findings were validated in an orthotopic mouse model of melanoma and suggested that aging enhances lymphatic vascular permeability, allowing for tumor cells to reach the systemic circulation. Consistent with this idea, aged lymphatic fibroblasts produced a less complex and more aligned extracellular matrix (ECM) compared with that of young fibroblasts, and these age-related ECM changes resulted in decreased VE-cadherin–dependent endothelial cell–cell adhesion and increased lymphatic vessel permeability. Similar to findings in the skin microenvironment, these changes in lymphatic vessel integrity were mediated by age-dependent loss of hyaluronan and proteoglycan link protein 1 (HAPLN1) secretion by lymphatic fibroblasts. HAPLN1 expression was reduced in sentinel lymph nodes of aged patients with melanoma, and higher HAPLN1 expression was prognostic of improved long-term survival. Furthermore, lymphatic injection of HAPLN1 into aged mice resulted in an increased rate of lymphatic metastases and decreased frequency of distant pulmonary metastases. These results support the notion that aging determines the route of melanoma metastasis via loss of HAPLN1-mediated regulation of ECM integrity and lymphatic permeability.

See article, p. 96

  • Circulating CD44+ breast tumor cell clusters promote cancer stemness and polyclonal metastasis.

  • Intercellular CD44 homophilic interactions recruit PAK2 and activate downstream FAK signaling.

  • Inhibition of CD44+ circulating tumor cell clusters may be a potential antimetastasis strategy.


Circulating tumor cells (CTC) are the tumor cells that disseminate into the vasculature; however, only a small CTC subpopulation forms distant metastases, and clustered cells are highly advantageous over single cells for polyclonal metastases. To elucidate the mechanisms underlying CTC cluster formation and polyclonal metastasis, Liu and colleagues evaluated blood samples and tissue sections from patients with breast cancer and mice implanted with patient-derived breast cancer xenografts (PDX). Intravital high-resolution two-photon microscopy identified CTC clusters in both human and mouse and showed that polyclonal cluster formation resulted from the dynamic aggregation of single tumor cells near sites of intravasation within the breast tumor and post-vascular aggregation of proximal CTCs within the lungs. Furthermore, CTCs that formed clusters exhibited a high potential of tumorigenesis and metastasis with increased expression of CD44, a well-established marker of breast cancer stem cells. CD44+, but not CD44, breast tumor cells formed clusters and polyclonal lung metastases; similarly, CD44 depletion in breast cancer PDX cells resulted in reduced tumor aggregation and metastasis. CD44-induced cell aggregation is independent of hyaluronan, but occurs through CD44 extracellular domain–mediated intercellular, homophilic interactions. Inhibition of CD44 or truncation of the CD44 N-terminal domain I suppressed breast tumor cell aggregation and lung metastasis. PAK2 was identified as a novel target and interactive partner of CD44, and inhibition or depletion of PAK2 blocked tumor cell aggregation and metastasis. These results show that metastasis is driven by homophilic CD44 interaction–mediated tumor cell aggregation.

See article, p. 114

  • Spatiotemporal loss of Nf1 in different cell lineages underlies the different neurofibroma subtypes.

  • Loss of Nf1 in a Hoxb7-expressing Schwann cell lineage induces both cutaneous and plexiform neurofibroma.

  • The Hippo pathway acts as a modifier of cutaneous and plexiform neurofibroma development.


Neurofibromatosis type 1 (NF1) is a genetic syndrome caused by heterozygous loss of the NF1 gene, which predisposes to the development of benign nerve sheath tumors called neurofibromas. Efforts to model NF1 through conditional inactivation of Nf1 has successfully recapitulated the development of plexiform neurofibromas (pNF) and malignant peripheral nerve sheath tumors, which can develop from pNFs, but no mouse model to date has successfully recapitulated the painful and disfiguring cutaneous neurofibromas (cNF) that develop at skin nerve terminals in every patient with NF1 starting in puberty and increasing through adulthood. Hypothesizing that NF1 loss needs to occur at a specific time and in specific cell types during development to generate cNF, Chen and colleagues used lineage tracing experiments to identify a putative Hoxb7-expressing cNF cell of origin that gave rise to derivatives populating both peripheral nerves and dermal nerve endings. Biallelic inactivation of Nf1 in Hoxb7-expressing cells not only faithfully recapitulated the features of diffuse cNF but also gave rise to pNF, unlike other NF1 models. The Hippo pathway was further identified as a modifier of cNF development, as co-deletion of Lats1 and Lats2 (negative Hippo pathway regulators) with Nf1 in Hoxb7-expressing cells promoted the widespread development of both diffuse and nodular cNFs in association with increased MAPK signaling and decreased overall survival compared with Nf1 deletion alone. The development of a more faithful NF1 model can provide insights into the etiology of this syndrome and facilitate investigation of potential therapies.

See article, p. 130

  • Lineage tracing shows PRSS56+ boundary cap cells give rise to cells at nerve roots and skin nerve terminals.

  • Conditional inactivation of Nf1 in PRSS56+ cells leads to the development of both cutaneous and plexiform neurofibroma.

  • Cutaneous neurofibroma development occurs in a progressive manner and can be accelerated by skin injury.


Cutaneous neurofibromas (cNF) are benign growths arising from the nerve terminals of the skin that occur in all patients with the tumor predisposition syndrome neurofibromatosis type 1 (NF1), caused by loss of NF1. cNFs can cause significant morbidity and are typically treated by surgical or laser ablation. Genetically engineered mouse models of NF1 have provided important insights into the origin and development of less common forms of neurofibroma such as plexiform neurofibromas (pNF) affecting nerve tracts and roots, but no model has recapitulated cNF. Radomska and colleagues showed that biallelic inactivation of Nf1 in Prss56-expressing neural crest–derived boundary cap (BC) cells, derivates of which migrate into the nerve roots and along peripheral nerves into the nerve terminals of the skin, led to the development of both pNF and diffuse cNF. Analysis of skin from embryonic day E13.5 when BC-derived cells first migrate to the skin to 6 months after birth when mice are still asymptomatic indicated that Nf1 deletion leads to progressive Schwann cell hyperplasia and the formation of cNF-like microlesions by 6 months, suggesting that cNF does not develop suddenly in adult skin but evolves from smaller lesions. Skin injury also promoted the proliferation and migration of Schwann cells and accelerated the development of cNFs in Nf1 knockout mice, a finding with potentially important implications for current cNF treatment. The insights provided into the origin and progression of cNF by more faithful models have the potential to facilitate study of new treatments for this debilitating disease.

Note:In This Issue is written by Cancer Discovery editorial staff. Readers are encouraged to consult the original articles for full details.