authors@Fred Hutch

Neves, Alexandre

March 16, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 3 | Fred Hutchinson Cancer Research Center 
 
p27 Shows Its Dark Side In Breast Cancer 
March 16, 2015 
    A Neves 
Cell division is critical for normal development, and is known to go awry in many different types of 
cancer. It is a highly regulated process involving two major classes of proteins: cyclin-dependent 
kinases (Cdks) that trigger cell cycle transitions, and cyclin-dependent kinase inhibitors (CKI) that 
keep Cdks in check. p27 is a CKI that is known to function in the nucleus as a tumor suppressor. 
Unlike other tumor suppressors, however, p27 is rarely mutated in human cancers, and tumors that 
develop in p27 heterozygous mice do not lose the remaining allele. Moreover, previous studies 
showed that mice harboring a homozygous mutation in p27 that prevents binding to Cdks are more 
tumor prone than are mice lacking p27 entirely. Together, these studies paint a complex picture and 
collectively suggest that p27 may have oncogenic functions when localized to the cytoplasm. Indeed, 
localization of p27 to the cytoplasm correlates with poor prognosis, tumor grade and survival in some 
tumor types, including breast and ovarian cancer. However, how cytoplasmic p27 becomes 
oncogenic and its relevance for therapy resistance had remained unclear. 
A recent Fred Hutch study led by Research Associate Dr. Hui Zhao in the lab of Dr. Jim Roberts 
(Basic Sciences Division), in collaboration with the lab of Dr. Bruce Clurman, (Human Biology and 
Clinical Research Divisions) and published inOncotarget, sought to shed some light on these 
questions. As a first step to investigate the role of p27 in human breast cancer, the authors 
compared p27 levels using immunoblotting and immunofluorescence in quiescent and proliferating 
breast cancer cell lines that were stratified by Her2 status. Her2 is an epidermal growth factor 
receptor family member that is known to be amplified in 15-20% of breast cancers. In non-
transformed breast epithelial cells (MCF10A) or Her2- breast cancer cell lines (HCC38 and 
HCC1937), p27 levels markedly decreased following cell cycle entry. In contrast, the researchers 
found that p27 levels remained high in Her2+ cell lines (UACC893 and BT474) after cell cycle 
induction, but was located in the cytoplasm instead of the nucleus. 
Next, the authors examined the link between p27 localization and cell proliferation and found that 
cytoplasmic p27 correlated with increased proliferation. Lapatininb is a FDA-approved drug that 
inhibits the kinase activity of Her2. Treatment of Her2+ cells with lapatinib halted cell cycle 
progression and resulted in nuclear rather than cytoplasmic expression of p27. To determine 
whether p27 was required for lapatinib-induced cell cycle arrest, the investigators knocked down p27 
March 16, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 3 | Fred Hutchinson Cancer Research Center 
 
in Her2+ cells using short hairpin RNA (shRNA) and found that, in contrast to control cells, p27 
knockdown cells failed to undergo efficient cell cycle arrest in response to lapatinib. 
One outstanding question was whether localization of p27 to the cytoplasm is simply a mechanism 
for reducing nuclear p27 or if it has a more active, oncogenic role. To address this, the researchers 
knocked down p27 in Her2+ cells treated with lapatinib and assayed for viability. Strikingly, 
knockdown of p27 in Her2+ cells resulted in hypersensitivity to lapatibib and caused increased cell 
death, as visualized by staining with cleaved caspase-3. To corroborate these findings, the authors 
expressed a nuclear-excluded form of p27 (p27ΔNLS) in Her2+ cells where endogenous p27 was 
knocked down. Notably, p27ΔNLS expression reversed lapatibib sensitivity of Her2+ cells and 
reduced apoptosis. 
Overall, this study demonstrated that cytoplasmic p27 promotes resistance of Her2+ breast cancer 
cells to lapatibib. "Our results suggest that p27 localization may be useful as a predictive biomarker 
of therapeutic response in patients with Her2+ breast cancers", said Dr. Zhao. 
Zhao H, Faltermeier CM, Mendelsohn L, Porter PL, Clurman BE, Roberts JM. 2014. Mislocalization 
of p27 to the cytoplasm of breast cancer cells confers resistance to anti-HER2 targeted 
therapy. Oncotarget, 5(24), 12704-14. 
 
 
 
Image provided by Dr. Zhao 
Immunofluorescence staining of p27 in Her2+ human breast 
cancer cell lines 9UACC893 and BT474 cells). Cells were 
released into cell cycle for 72 hr after synchronization by serum 
starvation. The "honeycomb-like" staining of the cell sheet 
results from nuclear exclusion of p27. 
 


English

p27 Shows Its Dark Side In Breast Cancer

Frost, Sophia

April 20, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 4 | Fred Hutchinson Cancer Research Center 
 
For Flu Shot Compliance, Stick Beats Carrot 
April 20, 2015 
    SHL Frost 
Fever, cough, sore throat—aching body, runny nose. Influenza is never enjoyable, but those affected 
usually recover in less than two weeks. Some people may however develop life-threatening 
complications (e.g. pneumonia) related to the contagious respiratory disease; the flu is in fact a 
major cause of morbidity and mortality among immunosuppressed cancer patients or recipients of 
hematopoietic cell transplants. The nation's health protection agency, Centers for Disease Control 
and Prevention (CDC), recommends seasonal flu shots for anyone over six months of age. Still, the 
early season vaccination coverage for 2014-15 was estimated to barely 40% of the general 
population in the United States. For health care personnel the corresponding rate was on average 
64%, although it reached almost 79% among those working in hospitals. The latter may seem quite 
satisfactory at a first glance, but given the vulnerability of hospitalized high-risk individuals, all non-
vaccinated healthcare personnel pose a potential threat. Thus, the question arises: how can 
influenza vaccination rates be maximized at an institution where the baseline compliance already is 
high? 
This was examined in a collaborative study directed by Fred Hutch's Dr. Steven Pergam, including 
investigators from Seattle Cancer Care Alliance and University of Washington. At the center of 
attention was a large ambulatory cancer care center with historically high rates of flu vaccination. 
Approximately 85% of staff was vaccinated annually, but despite sophisticated campaigns and 
feasible, free-of-charge immunizations, the number had not increased in a few years. Setting the 
vaccine coverage goal at a commendable 95%, the investigators decided to try two different 
approaches to succeed, over the course of one year per strategy. 
In 2011, they launched an incentive-based campaign—the "carrot" approach—which allowed staff to 
passively decline the flu shot. At stake were $25 gift cards for all employees as a reward for reaching 
the pre-specified goal, and progress updates were distributed weekly showing the compliance for 
each department. This concept was superseded in 2012 by a penalty-based "stick" strategy, which 
forced employees who wanted to opt out to actively participate in a time-consuming face-to-face 
multistep declination process, after completion of enhanced influenza vaccine education. For Dr. 
Pergam and colleagues it was important to work towards changing people's minds about influenza 
vaccine rather than introducing a mandatory program. "The 'carrot' approach came from the 
business world—change minds through monetary interventions," Dr. Pergam explained. The "stick"  
April 20, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 4 | Fred Hutchinson Cancer Research Center 
 
 
was modeled after a program at Harborview, which was believed to work on a larger scale in the 
population at hand. 
The outcomes were disclosed in American Journal of Infection Control, showing that high 
compliance (>95%) could be achieved without introducing mandatory immunization. More 
specifically, 87% (1,264 out of 1,446) in 2010 became 92% (1,453 out of 1,586) in 2011, further 
increasing to 96% (1,583 out of 1,641) in 2012. Thus, the pointy stick beat the crunchy carrot, in 
particular among nonclinical employees with little or no direct patient care responsibilities. When 
asked if this came as a surprise, Dr. Pergam's answer was clear: "No!" He explained that previous 
data have shown that penalty programs work better, "this is true in many fields, for example, seat 
belt and motorcycle helmet use improved once there were laws to change practices." 
One of the interesting findings was that the educational component of the penalty-driven approach 
alone resulted in no increase in compliance, suggesting that it was the added time burden and/or 
required face-to-face activity that made people less inclined to decline. Another discovery of 
significance was the "stick"-induced improvements in vaccination rates among staff without direct 
patient contact, of particular interest because they may unknowingly transmit disease through 
common areas. "[The study] demonstrates the importance of tailoring programs to nonclinical staff, 
who can be important sources of viral transmission in ambulatory care centers," Dr. Pergam said. 
The study showed that high compliance rates can be achieved without having to resort to mandatory 
vaccination policies, but the investigators are not settling yet. Each year they redesign the program, 
bit by bit, to ensure continuous development. Currently, the focus is on how to bring the persisting 
decliners on board. "We need to better understand the reasons and beliefs of those who decide not 
to get vaccinated," said Dr. Pergam, who is also working on improving the access to flu vaccine for 
patients and their families. "While we are happy to get staff vaccinated, vaccinating patients and their 
caregivers may be an even more critical step to improving respiratory virus prevention in the clinic." 
 
Podczervinski S, Stednick Z, Helbert L, Davies J, Jagels B, Gooley T, Casper C, Pergam SA. 2015. 
Employee influenza vaccination in a large cancer center with high baseline compliance rates: 
Comparison of carrot versus stick approaches. Am J Infect Control. 43(3), 228-33. 
 
\ 
April 20, 2015 SCIENCE SPOTLIGHT 
 
3 Volume 5, Issue 4 | Fred Hutchinson Cancer Research Center 
 
 
 
Image provided by Dr. Steven Pergam 
Acceptance of flu shots over the 60 first days of study. 
 


For Flu Shot Compliance, Stick Beats Carrot

Pattacini, Laura

April 20, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 4 | Fred Hutchinson Cancer Research Center 
 
Unveiling the Mystery of CXCL13 Production in 
HIV-Infected Subjects 
April 20, 2015 
     L Pattacini 
Chemokines are a family of small molecules that are able to drive the movement of cells toward the 
site where they are requested to operate. Because of their critical role, strict surveillance of their 
production is required for the immune responses to function properly. In HIV-infected subjects the 
secretion of one of these molecules, CXCL13, is more elevated than in HIV-negatives. While 
CXCL13 at physiological concentrations facilitates the co-migration of B cells and follicular helper T 
cells into B cell follicles and germinal centers, thus ensuring a correct B cell response, an altered 
secretion of the chemokine leads to a deregulated B cell response. The reasons for this modified 
production and secretion have been explored by Dr. Kristen Cohen of the Stamatatos Lab in the 
Vaccine and Infectious Disease Division (VIDD) at Fred Hutch, and have been published in 
the Journal of Immunology. 
The study unravels the mechanism through which HIV modifies the homeostatic production of 
CXCL13. The fundamental observation is that stimulation of peripheral blood mononuclear cells 
(PBMCs) with replication-competent HIV virions results in increased secretion of CXCL13. Moreover, 
monocytes constitute the main population responsible for this production, while both myeloid or 
plasmocytoid dendritic cells (pDCs) produce low amounts of the chemokine. 
Starting from these findings, we are led deeper into the mechanism driving this phenomenon. First, 
activation of the two toll-like receptors (TLR) 7 and 8, localized in the endosomal compartment and 
able to recognize HIV RNA sequences, induces increased secretion of CXCL13 from PBMCs. 
Interestingly, this is not true for isolated monocytes, where TLR stimulation can induce only a minor 
change in the chemokine secretion. Thus, another cell type has to be involved in induction of 
CXCL13 secretion by monocytes. Indeed, the effect of the two TLR ligands is not present when 
PBMCs are depleted from pDCs, which express both receptors and, upon engagement of TLR7 and 
TLR8, with their ligands, produce a hormone with known antiretroviral activity, IFN-alpha. 
Furthermore, the blockade of IFN-alpha function with a specific antibody inhibits the production of 
CXCL13. 
HIV causes yet another alteration of the immune system, by exploiting the body's mechanisms to 
fight viral infections. In fact, by binding TLR7 and TLR8, HIV RNA activates pDCs to produce IFN-
April 20, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 4 | Fred Hutchinson Cancer Research Center 
 
alpha, which in turn induces the transcription of CXCL13 in monocytes and causes the high protein 
concentration. This mechanism might be the reason of the disruption in lymphoid architecture, 
reduction of de novo antibody responses and hypergammaglobulinemia observed during the 
progression of untreated HIV infection. Thus, this study sheds new light on one of the reasons why 
HIV, although infecting CD4-positive T cells, also heavily affects the B cell compartment.  
Dr. Cohen concludes by highlighting the importance of the same pathways in other conditions: "Type 
I interferon-induction of CXCL13 may help explain the increase of CXCL13 during HIV-1 infection 
and other inflammatory conditions. Further research is still needed to fully understand the 
implications of increased CXCL13 during these conditions and its contribution to 
immunopathogenesis." 
Cohen KW, Dugast A-S, Alter G, McElrath MJ, Stamatatos L. 2015. HIV-1 Single-Stranded RNA 
Induces CXCL13 Secretion in Human Monocytes via TLR7 Activation and Plasmacytoid Dendritic 
Cell-Derived Type I IFN. Journal of immunology. 194(6):2769-2775. 
 
 
Image provided by Dr. Kristen Cohen 
Figure caption: Schematic representation of IFN-alpha-mediated 
up-regulation of CXCL13 following HIV infection. 
 


Unveiling the Mystery of CXCL13 Production in HIV-Infected Subjects

Kocarnik, Jonathan

May 18, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 5 | Fred Hutchinson Cancer Research Center 
 
Several Cancer Susceptibility Variants Also Affect 
Melanoma Risk 
May 18, 2015 
     JM Kocarnik 
Pleiotropy refers to a single genetic locus having multiple phenotypic effects.  Evidence of pleiotropy 
can suggest shared molecular mechanisms or disease susceptibility pathways between 
phenotypes.  Several regions of the genome demonstrate pleiotropic effects with multiple cancers, 
but the extent of these pleiotropic relationships has not been fully characterized.  To identify 
pleiotropic genetic variants that may impact melanoma risk, Drs. Jonathan Kocarnik and Ulrike 
Peters and colleagues in the Public Health Sciences Division evaluated a panel of genetic variants 
previously associated with other cancers for an association with melanoma.  As recently reported 
in PLoS One, the authors provided confirmatory evidence of such an association for two variants, 
while also providing additional suggestive evidence for a third variant that may have a male-specific 
effect. 
Melanoma is the most serious form of skin cancer.  While the largest risk factor for melanoma is 
ultraviolet radiation exposure, several genetic variants have also been found to impact risk.  Many of 
these variants are expected to affect melanoma risk through their impact on pigmentation.  By 
looking for pleiotropic associations with genetic variants associated with other cancers, additional 
mechanisms of melanoma susceptibility may be identified.  
As such, the authors evaluated 181 single nucleotide polymorphisms (SNPs) associated with various 
cancers for an additional association with melanoma.  For each SNP, the authors performed meta-
analyses across five studies, totaling over 2,000 melanoma cases and 20,000 controls.  Three of 
these studies were participants in the Population Architecture using Genomics and Epidemiology 
(PAGE) study, while the other two were collaborating cohort studies of health professionals.  
From these analyses, the authors found two SNPs associated with melanoma after correction for 
multiple comparisons.  These variants are located in the TERT/CLPTM1L locus, a well-known 
pleiotropic region that has been associated with several cancers, including melanoma and lung 
cancer.  Similar to previous findings, these variants previously associated with an increased risk of 
lung cancer were also found to be associated with a decreased risk of melanoma.  "These 
pleiotropic effects in opposite directions are interesting not only from the standpoint of elucidating  
 
May 18, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 5 | Fred Hutchinson Cancer Research Center 
 
 
tumor biology," said lead author Dr. Kocarnik, "but also from a risk interpretation and communication 
standpoint."  Whereas most genetic variants are generally thought of as only impacting risk for one 
phenotype or only increasing or decreasing risk, pleiotropic variants such as these will require a 
more nuanced understanding and discussion.  
Because previous melanoma studies have shown sex differences in anatomical location, risk of 
metastases, and survival, the authors also performed analyses stratified by sex.  One additional 
variant, a prostate cancer SNP between TPCN2 and MYEOV, showed a suggestive association with 
melanoma in males but not in females.  While not reaching statistical significance after multiple 
comparisons adjustment, several lines of evidence offer biological plausibility for the finding.  Two 
coding variants in the ion channel gene TPCN2 have been associated with hair pigmentation, 
while MYEOV is an oncogene that includes variants implicated for multiple cancers.  
The authors had previously reported a sex difference in the association with melanoma for another 
SNP in a solute-carrier gene associated with pigmentation.  Other studies have demonstrated the 
functional importance of ion and small molecule transport to melanogenesis and the pigmentation 
pathway, and that these processes can be up or down regulated by sex hormones.  "Together with 
our sex-specific findings, these lines of evidence raise questions regarding potential sex differences 
in the relationships between ion transport, pigmentation, and melanoma risk," said Dr. Kocarnik.  "It 
is possible that sex differences in circulating levels of these hormones impact ion exchange or 
tyrosinase activity in a way that modifies the effect of these variants on melanoma risk, perhaps 
through alterations to melanogenesis or skin pigmentation."  Future research will be needed to 
further explore these potential relationships. 
Other Public Health Sciences Division researchers contributing to this project were Drs. Chris 
Carlson and Charles Kooperberg. 
 
Kocarnik JM, Park SL, Han J, Dumitrescu L, Cheng I, Wilkens LR, Schumacher FR, Kolonel L, 
Carlson CS, Crawford DC, Goodloe RJ, Dilks HH, Baker P, Richardson D, Matise TC, Ambite JL, 
Song F, Qureshi AA, Zhang M, Duggan D, Hutter C, Hindorff L, Bush WS, Kooperberg C, Le 
Marchand L, Peters U. 2015. Pleiotropic and Sex-Specific Effects of Cancer GWAS SNPs on 
Melanoma Risk in the Population Architecture Using Genomics and Epidemiology (PAGE) 
Study. PLoS One. 10(3):e0120491. 
 
May 18, 2015 SCIENCE SPOTLIGHT 
 
3 Volume 5, Issue 5 | Fred Hutchinson Cancer Research Center 
 
 
 
 
Image from the publication. 
This pleiotropy-colored Manhattan plot shows the –log(10) p-value 
for the association between melanoma and SNPs previously 
associated with other cancers. The solid line denotes the 
Bonferroni-corrected significance threshold for this analysis 
(0.05/181 SNPs). 
 


Several Cancer Susceptibility Variants Also Affect Melanoma Risk

Zubair, Niha

November 15, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
Beyond the genome: epigenetic changes 
specific to prostate tumors 
November 15, 2015 
     N Zubair 
 
 
 
 
 
Volcano plot of DNA methylation in 
prostate cancer versus adjacent benign 
tissue. Differentially methylated CpGs are 
labeled in green or red. The 27 red-
labeled CpG sites had a mean 
methylation difference of at least 40% 
between cancer and benign tissue; all 
were hypermethylated. 
 
Image provided by Dr. Milan Geybels 
 
 
 
 
 
Methylation of DNA is a well-known epigenetic mechanism for the control of gene 
expression.  Methylation commonly occurs at CpG sites, or regions of DNA where a cytosine 
nucleotide occurs next to a guanine nucleotide. These sites tend to cluster into islands and are often 
found in gene promoter regions. CpG islands are typically unmethylated, thus allowing for 
uninhibited transcription of the gene, while hypermethylation of these regions can lead to 
transcriptional silencing. Losses and gains of DNA methylation have been associated with various 
cancers, including prostate cancer.  
While previous research has investigated methylation in relation to prostate cancer, the majority of 
studies have examined methylation in only a select number of genes. To address this caveat, Drs. 
November 15, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
Milan Geybels, Janet Stanford, and colleagues (Public Health Sciences Division) evaluated DNA 
methylation changes epigenome-wide in paired prostate cancer and adjacent benign tissue samples. 
This work, recently published in The Prostate, aimed to discover novel differences in methylation 
between cancerous and benign tissue samples and determine if previous findings from smaller 
studies would replicate in their larger study. 
Paired samples came from 20 radical prostatectomy patients, 18 European Americans and 2 African 
Americans. The authors conducted epigenome-wide profiling and found over 2,000 CpG sites with 
differential DNA methylation; the majority of these CpGs were hypermethylated in cancer versus 
benign tissue. The average hypermethylation difference was 26%.  
For verification of their results, the authors examined the 27 top-ranked differentially methylated CpG 
sites in 46 participants from the prostate cancer dataset in The Cancer Genome Atlas (TCGA). 
Confirming their initial findings, all 27 CpG sites were significantly hypermethylated in the cancerous 
tissue using TCGA. Next, the authors evaluated gene expression levels of these hypermethylated 
top-ranked regions. Of the 18 genes that had transcript data available, three genes (SCGB3A1, 
HIF3A, and AOX1) had reduced messenger RNA (mRNA) expression in cancer versus benign 
tissue. 
Dr. Geybels describes the impact of this work, "we identified specific DNA methylation changes in 
prostate tumor tissue compared to non-cancer prostate tissue. These DNA methylation changes 
were in specific genes, some of which have not been implicated in prostate cancer before," such as 
SCGB3A1, a gene suggested to function as a tumor-suppressor. Furthermore, "we also confirmed 
some previous findings of genes that are differentially methylated in prostate cancer, thereby 
providing further evidence for a role of these genes in prostate carcinogenesis. Our findings may 
therefore provide new insights into the development of prostate cancer."  
Additional Fred Hutch investigators contributing to this research were Shanshan Zhao, Chao-Jen 
Wong, and Dr. Michael Wu.  
Geybels MS, Zhao S, Wong C-J, Bibikova M, Klotzle B, Wu MC, Ostrander EA, Fan J-B, Feng Z, 
Stanford JL. 2015. Epigenomic profiling of DNA methylation in paired prostate cancer versus 
adjacent benign tissue. Prostate. 75(16):1941-50. 
Funding for this study was provided by the National Cancer Institute, Fred Hutchinson Cancer 
Research Center, Prostate Cancer Foundation, and the Dutch Cancer Society.  


Beyond the genome: epigenetic changes specific to prostate tumors

Koch, Lori

November 15, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
Filling in the gaps: centromere of stickleback 
fish identified 
November 15, 2015 
     L Koch 
 
 
 
Metaphase chromosome spread from 
threespine stickleback fish. The 
centromere repeat probe (green, right) 
hybridizes to the primary constriction on 
metaphase chromosomes (grey, left). 
 
Image provided by Jennifer Cech. 
 
 
 
 
 
 
 
 
Cell division is one of the most basic and important processes of living organisms.  Cells must 
replicate and accurately segregate their chromosomes into daughter cells in order to ensure their 
survival.  On the other hand, the regulation of cell division is often abnormal in cancer cells, which 
makes it an attractive target for therapeutic intervention.  Accurate separation of chromosomes by 
the mitotic spindle depends upon their connection to spindle fibers at a highly regulated location of 
each chromosome called the centromere.  
 
Despite their functional importance and the conservation of the cell division machinery across 
eukaryotic organisms, there is a fascinating variability in the underlying DNA sequence of the 
centromere as well as in many of the proteins that bind to the centromere.  How can diverse 
organisms carry out this same process given such variability at the source of the connection 
between the DNA and the spindle?  This is an ongoing topic of research and has been called the 
"centromere paradox".  A recent study led by graduate student Jennifer Cech in the laboratory of Dr. 
Catherine Peichel (Basic Sciences) set out to identify and characterize the centromeric sequence of 
the threespine stickleback fish (Gasterosteus aculeatus), which is an important model system in 
evolutionary biology and genetics.  Their work was recently published in Chromosome Research.    
November 15, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
 
In addition to the variation in the sequence between organisms, another major obstacle to 
sequencing centromeres is the fact that they are highly repetitive and AT-rich regions, making 
aligning and assembling sequencing reads extremely difficult.  For this reason, we really don’t 
know the sequence of centromeres from many species, and centromeres have been called the dark 
matter of the genome’” , said Dr. Peichel (Basic Sciences).  To overcome this, the authors set out to 
identify the sequence of threespine stickleback CENP-A, a histone H3 variant that is known to be the 
epigenetic mark of the centromere across nearly all eukaryotes.  Using the sequence of the 
zebrafish CENP-A, the authors identified a homologous CENP-A gene in the threespine stickleback 
genome.    The gene encodes a 148 amino acid protein with the characteristic histone-fold domain in 
the C-terminus and a divergent N-terminal tail.  To confirm this protein marked the centromeric 
histones, they raised an antibody against the N-terminus of CENP-A, which varies considerably from 
the canonical histone H3 protein, and showed by immunofluorescence that the antibody recognizes 
distinct foci that localize to the primary constriction of each chromosome in nuclei and chromosome 
spreads, consistent with it marking the centromeric histone H3, CENP-A.    
 
Now the authors were ready to identify the centromeric DNA sequence of the threespine 
stickleback.  Using their stickleback CENP-A antibody they performed chromatin 
immunoprecipitation followed by high-throughput sequencing (ChIP-seq).  Next they used a cluster 
analysis pipeline developed in the laboratory of Dr. Steve Henikoff (Basic Sciences) in order to find 
sequences that were enriched in the immunoprecipitated DNA compared to total input DNA.  A 186 
base pair repeat was identified, bearing the characteristics of other known centromeric repeats such 
as being AT-rich and harboring a CENP-B box motif.  Furthermore, at 186 nucleotides, it is a length 
of DNA that could wrap a single nucleosome.     
 
In order to confirm this repeat marked the centromere of each chromosome, the authors performed 
FISH (fluorescence in situ hybridization) on the stickleback fish chromosomes using a probe against 
their GacCEN (G. aculeatus centromere) sequence.  They found that the sequence localized to the 
primary constriction of 41 out of 42 chromosomes in nuclei and chromosome spreads.  The authors 
hypothesized that the chromosome that was unlabeled by the GacCEN probe was the Y 
chromosome, given that the sequence of the Y centromere of mammals is known to be highly 
divergent from the centromere of the other chromosomes.  Indeed, the authors were able to identify 
the unlabeled chromosome as the Y by also using probes that distinguish the X and Y chromosomes 
of the stickleback fish.   
 
November 15, 2015 SCIENCE SPOTLIGHT 
 
3 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
"We are now working hard to identify the Y chromosome centromere sequence," said Dr. Peichel 
(Basic Sciences). "Once we do this, we will be able to see how it has evolved in other stickleback 
species, including two species that have a fusion between the Y chromosome and another 
chromosome. " It will certainly be exciting to see how these fusion chromosomes have contributed to 
the evolution of stickleback fish species.  The rapid evolution of centromeres has been hypothesized 
to be important for reproductive isolation of emerging species.   
 
Cech JN, Peichel CL.  2015.  Identification of the centromeric repeat in the threespine stickleback 
fish (Gasterosteus aculeatus).  Chromosome Research. 23(4):767-79. doi: 10.1007/s10577-015-
9495-3. 
 
This work was funded by National Science Foundation, National Institutes of Health and Fred 
Hutchinson Cancer Research Center.  
 
 


Filling in the gaps: centromere of stickleback fish identified

Herman, Jake

November 15, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
Pediatric core binding factor AML under the 
molecular lens 
November 15, 2015 
     J Herman 
 
 
 
 
 
 
 
(A)Principal component analysis of changes in 
transcription profiles for pediatric AML samples with 
different cytogenetic defects. Expression profiles of CBF-
AML samples are distinct from AML samples with normal 
karyotypes (blue) and segregate from each other based 
on the specific chromosome translocation, either Inv(16) 
(green) or t(8;21) (red). (B) While each translocation 
creates a unique transcription fingerprint, both types of 
CBF-AML are more similar to each other than AML 
samples without translocations, sharing nearly 200 
differentially expressed genes in common. 
 
Image provided by Jake Herman 
 
 
 
 
 
Acute myeloid leukemia (AML) affects hematopoietic progenitor and/or stem cells by altering key 
signaling pathways involved in differentiation, resulting in the clonal expansion of highly proliferative 
immature hematopoietic cells. When AML patient samples are analyzed at the molecular level it is 
clear that many genetic alterations can result in this disease phenotype, including chromosome 
translocation, single-nucleotide polymorphisms, and epigenetic alterations. Unsurprisingly, the 
molecular complexity of this disease results in dramatic variability in response to chemotherapy 
regimens. Consistently, there are indications that age-dependent genetic alterations also exist, as 
November 15, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
patient outcomes between pediatric and adult AML do not correlate. To better understand the 
genetic complexity of this cancer, there has been a large effort to understand the transcription profile 
of AML samples. Numerous studies, including The Cancer Genome Atlas, have used RNA-seq 
technology to characterize not just transcript numbers, but also genetic rearrangements and splicing 
alterations. However, to date these studies have all approached adult AML, while the transcription 
landscape of pediatric cases remains largely unexplored. 
The most common form of pediatric leukemia, core binding factor AML (CBF-AML), represents 25% 
of cases. At the molecular level CBF-AML arises from two distinct chromosomal translocations, 
t(8;21) and Inv(16), which disrupt the functions of core binding factor alpha and core binding factor 
beta, respectively. Core binding factor proteins are essential for development and function in a 
complex to regulate transcription of genes involved with hematopoiesis. These translocations result 
in the novel fusion proteins RUNX1-RUNX1T1 and CBFB-MYH11. While the presence of these 
chimeric proteins is currently used for risk stratification, to date there is no targeted therapy that 
exploits these unique genetic events. Moreover, it is known that both translocations lead to similar 
clinical outcomes despite affecting different core binding factors, yet the molecular similarities and 
differences between these pathways remain largely uncharacterized. The Meshinchi laboratory in 
the Clinical Research Division has begun to fill these gaps by performing comprehensive RNA-seq 
analysis on pediatric AML samples in a recent article published in PlosOne.  
RNA-seq analysis was performed on samples from 64 pediatric AML patients, these patients 
represented three distinct karyotypes: t(8;21), Inv(16), and Normal Karyotype (NK). The NK group 
was also subdivided by those lacking or containing a tandem duplication within the FLT3 gene 
(FLT3/ITD). FLT3/ITD is a commonly observed mutation in AML, however, this study revealed that 
the transcription profile of NK samples was largely unaffected by the presence of this mutation. 
Principal component analysis (PCA) was performed on the RNA-seq data for all coding genes in 
each sample. Using a two-component analysis, researchers were able to distinguish each unique 
cytogenetic subtype of AML. Despite their clinical similarities, t(8;21) and Inv(16) do in fact alter gene 
transcription in different fashions. While there are differences between the CBF-AML subtypes, there 
are also similarities. By comparing each group only to NK-AML nearly 200 genes were commonly 
altered by both translocations. Excitingly, this included the down regulation of nearly two-dozen HOX 
genes, which normally function in morphogenesis, particularly in skeletal development. Inhibiting 
these genes likely contributes to the lack of cellular differentiation observed specifically in CBF-AML, 
as they are not down regulated in NK-AML samples. Further analysis of unique expression changes 
within t(8;21) samples revealed decreased transcription of a network of genes involved in immune 
response, which only further emphasizes the unique nature of each cytogenetic subtype. 
November 15, 2015 SCIENCE SPOTLIGHT 
 
3 Volume 5, Issue 11 | Fred Hutchinson Cancer Research Center 
 
Finally, using RNA-seq technology researchers were also able to identify previously uncharacterized 
splicing changes and fusions transcripts within all AML samples. Splicing alterations were 
particularly interesting, as both NK and t(8;21) samples commonly included alternative exons, while 
Inv(16) samples largely skipped and excluded exons. Despite these different changes to splicing, the 
set of genes that experienced alternative splicing was heavily enriched for acetylated proteins, and 
those involved in ribosome biogenesis. The fusion transcripts identified by this study were also 
intriguing, RNA-seq successfully identified 94% of CBF-AML fusions, as well as fusions unknown to 
exist in the samples. Particularly interesting was the fusion of NUP98, a nuclear pore complex 
protein, to three different proteins: NUP98-NSD1, NUP98-HOXD13, NUP98-HMGB3. The NUP98-
NSD1 fusion has previously been correlated to patient outcomes, while the other fusions are less 
known and mostly uncharacterized. Overall, this work has opened the field to many new and exciting 
questions. Understanding the functional consequences of these altered transcriptional profiles will 
hopefully lead to new therapies that exploit the unique behaviors of pediatric CBF-AML subtypes. 
Funding for this research was provided by the National Institutes of Health 
 
Hsu CH, Nguyen C, Yan C, Ries RE, Chen QR, Hu Y, Ostronoff F, Stirewalt DL, Komatsoulis G, 
Levy S, Meerzman D, Meshinchi S. 2015. Transcriptome profiling of pediatric core binding factor 
AML. PlosOne. 10(9):e0138782. doi: 10.1371/journal.pone.0138782. 
 
 


Pediatric core binding factor AML under the molecular lens

May 18, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 5 | Fred Hutchinson Cancer Research Center 
 
An Unpredicted Predictor of Outcome in Pediatric 
AML 
May 18, 2015 
    SHL Frost 
Acute myeloid leukemia (AML) is a cancer of the bone marrow that has large outcome variability 
among affected patients. The disease is relatively rare, accounting for only 1.2 % of all cancer 
deaths in the United States, but is expected to increase in incidence. A number of factors have been 
identified to facilitate diagnostic classification and risk assessment, but when it comes to predicting 
individual outcomes much less is known. The few predictive markers that have been identified so far 
are insufficient, and there is a need in the clinic for more sophisticated characterization methods for 
AML. 
Drs. George Laszlo and Roland Walter in the Clinical Research Division recently published results 
from discovery gene expression studies utilizing pediatric AML specimens. They found an 
association between patient outcomes and the expression level of MMRN1, which is a gene that 
encodes the protein multimerin-1, a member of the elastin microfibrillar interface protein 
(EMILIN)/multimerin family. To follow up on their findings, the Fred Hutch investigators initiated a 
retrospective study in which MMRN1 expression was quantified in frozen samples of bone marrow 
from patients enrolled in two Children's Oncology Group AML protocols. The data were then 
correlated with clinical outcomes to assess the potential use of MMRN1 as a predictive biomarker in 
childhood AML. 
The results were recently accepted for publication in Clinical Cancer Research, showing large 
variation in MMRN1 expression among AML patients. The researchers arbitrarily divided the patients 
into quartiles, making the cut point dividing the highest quartile from the lower three quartiles at a 
relative MMRN1 expression of 0.5. In the first part of the study 183 specimens were compared, 
showing that patients with a relative expression of ≥ 0.5 (n = 45) fared worse than those with a 
relative expression of < 0.5 (n = 138). While there was no statistically significant difference in overall 
survival (OS) in this training cohort, patients withMMRN1 ≥ 0.5 experienced shorter event-free 
survival (EFS; time from study entry to failed initial remission, relapse, or death), and higher relapse 
risk (RR). 
 
 
May 18, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 5 | Fred Hutchinson Cancer Research Center 
 
 
Next, MMRN1 expression levels were correlated with clinical outcomes and disease characteristics 
in 740 patients, using the established threshold. In this cohort, a significant trend indicated higher 
likelihood for standard- or high-risk disease characteristics among patients with a high 
relative MMRN1 expression (≥ 0.5; n = 160); the same population experienced lower probability of 
low-risk disease. In addition, these patients were less likely to respond to initial chemotherapy, and 
had poorer long-term outcome in terms of OS, EFS, and RR. After adjusting for disease risk, age, 
bone marrow blast percentage, and treatment, it was demonstrated that MMRN1 expression could 
indeed be used for prediction of aforementioned clinical endpoints in childhood AML. Subgroup 
analyses to assess the predictive potential in low-, standard-, and high-risk groups were limited by 
the relatively small sample size, but indicated that patients with high MMRN1 expression in all three 
risk categories generally fared worse than those with lower relative expression. 
The mechanisms behind the observed correlation have yet to be fully revealed, as MMRN1 is 
previously unrecognized as a predictive biomarker for any human cancer. In fact, it "comes totally 
out of nowhere in terms of myeloid biology," according to Dr. Laszlo. "We stumbled upon MMRN1 as 
a gene that co-segregates with a proteasomal degradation complex component, called SOCS2, 
which also happens to have high transcript level in poor prognosis patients, but whereas there is 
rationale for SOCS2 expression being aberrantly high in some bad-acting pediatric AMLs, there is 
very little known about MMRN1 biology that would indicate what role it might have in leukemic cells." 
Besides expanding the analyses to include adult AML, the Walter Lab is thus interested in digging 
into the biology behind their findings to explore whether there is a biological basis for high 
multimerin-1 levels in myeloid cells and whether secreted multimerin-1 could serve as an AML 
biomarker. 
Further studies are needed to assertively conclude whether MMRN1 expression can refine the risk 
stratification and improve current predictive models for AML, but this unexpected finding may bring 
the scientists one step closer to advancing the development of individualized treatment plans - a 
sought-after goal. 
Laszlo GS, Alonzo TA, Gudgeon CJ, Harrington KH, Gerbing RB, Wang Y-C, Ries R, Raimondi SC, 
Hirsch BA, Gamis AS, Meshinchi S, Walter RB. 2015. Multimerin-1 (MMRN1) as novel adverse 
marker in pediatric acute myeloid leukemia: a report from the Children's Oncology Group. Clin 
Cancer Res. [Epub ahead of print] 
 
May 18, 2015 SCIENCE SPOTLIGHT 
 
3 Volume 5, Issue 5 | Fred Hutchinson Cancer Research Center 
 
 
See also: Laszlo GS, Ries RE, Gudgeon CJ, Harrington KH, Alonzo TA, Gerbing RB, Raimondi SC, 
Hirsch BA, Gamis AS, Meshinchi S, Walter RB. 2014. High expression of suppressor of cytokine 
signaling-2 predicts poor outcome in pediatric acute myeloid leukemia: a report from the Children's 
Oncology Group. Leuk Lymphoma. 55(12):2817-21. 
 
 
 
Image provided by Dr. George Laszlo 
Clinical outcome in patients with high (relative mRNA expression ≥ 
0.5) and low (relative mRNA expression < 0.5) MMRN1 expression 
in COG-AAML0531. 
 


An Unpredicted Predictor of Outcome in Pediatric AML

June 15, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 6 | Fred Hutchinson Cancer Research Center 
 
Mirkat Peeks at Associations between the 
Microbiome and Disease 
June 15, 2015 
   JM Kocarnik 
The microbiome refers to the community of microbes and bacteria that lives in and on each 
individual.  Recent advancements in high-throughput sequencing technology have allowed 
researchers to measure the presence and composition of the microbiome, and then evaluate these 
measurements for associations with disease.  Generally, the methods utilized for these analyses 
have relied on comparing phylogenetic distances between the species observed.  While useful, 
these methods have challenges that preclude extending microbiome analyses to alternative 
outcomes of interest, such as longitudinal or survival data.  To address these issues, Drs. Ni Zhao 
and Michael Wu in the Public Health Sciences Division developed a new statistical method called 
the Microbiome Regression-based Kernel Association Test (MiRKAT).  Presented in The American 
Journal of Human Genetics, this approach improves on current methods and allows researchers to 
extend microbiome research to investigations of alternative outcomes. 
Microbial communities can be profiled by performing targeted sequencing of the 16S rDNA 
sequences.  These sequences can identify the presence and relative abundance of microbial 
species contained in a sample.  After clustering these species into operational taxonomic units 
based on their 16S rDNA sequence similarity, distance metrics can be constructed to measure the 
phylogenetic or taxonomic dissimilarity between each sample.  These pairwise distances between 
each pair of samples can then be compared to the distribution of some outcome of interest in order 
to test for an association between microbiome diversity and disease.  
A key limitation to existing approaches is the need to choose an appropriate metric for measuring 
these pairwise distances.  UniFrac distances are the most popular (and can be weighted, 
unweighted, or generalized), but additional distances can also be used (such as Bray-Curtis or 
Euclidian).  Since these distances prioritize differences differently, choosing a non-optimal distance 
metric will reduce statistical power to detect a true association.  Choosing the optimal metric for a 
given project is difficult, however, and cherry-picking results from analyzing multiple metrics 
increases the chance of false positive findings.  In addition, current analytic approaches can be 
difficult to interpret, do not allow for easy covariate adjustment, and are challenging to apply to 
outcomes such as survival or multivariate data. 
"The need for more flexible analysis tools became apparent and motivated our development of 
June 15, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 6 | Fred Hutchinson Cancer Research Center 
 
MiRKAT," said senior author Dr. Wu. "MiRKAT arose, in part, because my collaborators wanted to 
understand the role of the microbiome in a set of clinical samples, but the standard analysis 
approaches did not easily accommodate aspects of the study design.  Our work makes a significant 
effort to close this gap and to help researchers better analyze their data and obtain more reliable and 
meaningful results." 
In short, the MiRKAT method uses the kernel machine framework to compare pairwise similarity in 
the outcome variable to pairwise similarity in the microbiome profiles.  High correspondence 
between these measurements suggests an association.  Mathematically, this resembles a linear or 
logistic regression model with an added kernel function (see figure), which also allows controlling for 
potential confounders.  The kernel measures the similarity between different individuals, and 
different kernels can be chosen to evaluate more advanced underlying models.  A score statistic 
then allows researchers to obtain a p-value for covariate-adjusted associations between microbiome 
diversity and an outcome of interest.  
An extension of their method, "Optimal" MiRKAT, alleviates the issue of having to choose the best 
distance metric by simultaneously considering several kernel options and correcting for these 
multiple iterations.  Both methods performed well in simulated and real datasets, demonstrating the 
power, efficiency, and utility of this new method for microbiome-profiling studies.  For example, the 
authors used MiRKAT in existing data to replicate an association between microbiome profiles in the 
lung and smoking status, but were also able to further control for potential confounders not possible 
in the original research. 
"MiRKAT opens the doors for a lot of new research," said Dr. Wu.  "As a generic tool which is more 
flexible than what is already out there, it will enable the conduct of new studies and facilitate 
research into the role of microbiome across a wide range of diseases.  This is particularly the case 
as researchers start to investigate the relationship between the microbiome and more complex 
outcomes such as time to event outcomes and longitudinal measurements.  MiRKAT also opens 
new areas of methodological research which we hope to pursue.  The statistical underpinnings of the 
approach are similar to what is done in genetic association studies; consequently, we hope to exploit 
these similarities to enable joint analysis of microbiome data with other complex -omics data such as 
genotyping or metabolomics.  This will be a major focus of my group’s efforts in the near future." 
Citation:  
Zhao N, Chen J, Carroll IM, Ringel-Kulka T, Epstein MP, Zhou H, Zhou JJ, Ringel Y, Li H, Wu MC. 
2015. Testing in microbiome-profiling studies with MiRKAT, the Microbiome Regression-based 
Kernel Association Test. Am J Hum Genet. 96(5):797-807. doi: 10.1016/j.ajhg.2015.04.003. 
June 15, 2015 SCIENCE SPOTLIGHT 
 
3 Volume 5, Issue 6 | Fred Hutchinson Cancer Research Center 
 
 
 
 
Image provided by Dr. Jonathan Kocarnik 
Mathematical representation of the MiRKAT regression 
framework. The association between the microbiome and an 
outcome of interest can be calculated through a linear kernel 
machine model for continuous outcomes (A) or a logistic kernel 
machine model for binary outcomes (B). The kernel measures 
the phylogenetic or taxonomic distance (C), and a score 
statistic (D) is calculated to generate a p-value for the 
association. 
 


Mirkat Peeks at Associations between the Microbiome and Disease

October 19, 2015 SCIENCE SPOTLIGHT 
 
1 Volume 5, Issue 10 | Fred Hutchinson Cancer Research Center 
 
T-cell ligands take the teeth out of BiTE 
antibodies 
October 19, 2015 
     JA Herman 
 
 
 
 
 
AMG 330-induced cytotoxicity displays dose-
dependent response in CD33+ human AML cells. 
Expression of inhibitory T-cell ligands, PDL-1 and 
PD-L2, reduces the activity of AMG 330, while 
stimulatory T-cell ligands, CD80 and CD86, 
sensitize cells to AMG 330 activity. 
 
Image provided by Dr. George Laszlo 
 
 
 
 
Antibody based therapeutics have heralded major advances in cancer treatment. A long -pursued 
therapeutic approach in this class is bi-functional antibodies that mediate the interaction between 
tumor and immune cells. The goal of such a therapy is to induce cytotoxicity, but more importantly to 
do so in targeted manner as an antigen-specific immunotherapy. Such bispecific antibodies 
exhibited limited clinical success until miniaturized molecules containing only minimal binding 
domains were generated. These streamlined molecules are termed BiTE (Bispecific T-cell engaging) 
antibodies and have already seen clinical success. These molecules generally bind T-cells by 
targeting the invariant epsilon subunit, CD3, however the tumor-specific marker can be vastly 
different depending on the cancer to be targeted. For example, Blinatumomab has seen clinical 
success in refractory and relapsed acute lymphoblastic leukemia (ALL) using the tumor-specific 
October 19, 2015 SCIENCE SPOTLIGHT 
 
2 Volume 5, Issue 10 | Fred Hutchinson Cancer Research Center 
 
marker, CD19. AMG 330 is another BiTE therapy, which was developed due to Blinatumomab’s 
success in patients. AMG 330 is designed to bind CD33 rather than CD19 because it targets tumors 
in patients with acute myeloid leukemia (AML). Patients are currently being enrolled for a phase I 
clinical trial thanks, in part, to preclinical successes. In their recent study in Blood Cancer Journal, 
Dr. George Laszlo and colleagues in the Walter Laboratory (Clinical Research Division), 
demonstrate that the entire composition of T-cell receptors expressed in AML patients affects 
response to AMG 330. 
Two AML cell lines were engineered to amplify expression of either inhibitory (PD-L1/2) or activating 
(CD80/86) T-cell ligands. Using these derived cell lines, Dr. Laszlo and colleagues found that AMG 
330 cytotoxicity was strongly repressed by either PD-L1 or PD-L2 expression. The resistance to 
AMG 330 activity is due to specific PD-L1/2 activity as antibodies targeting those ligands sensitized 
cells to AMG 330 treatment. This finding is important both for understanding heterogeneity in patient 
responses to AMG 330 and for addressing mechanisms of drug resistance that may evolve. Already, 
it appears, the question of drug resistance will be a more pertinent application of this study, “ Our 
PD-L1 phenotyping studies on patient specimens revealed little in the way of PD-L1 expression in 
previously frozen AML specimens, however another study recently published in Leukemia 
demonstrated PD-L1 expression due to prolonged exposure of AML patient specimens to AMG 
330” , said Dr. Laszlo. 
Understanding the mechanism of AMG 330 resistance and patient response in AML may provide the 
answer for overcoming those challenges. Dr. Roland Walter hopes this and future work will impact 
the application of BiTE antibodies “ 2-fold: first, expression of T-cell ligands could be explored as 
biomarkers for response (e.g. to select patients most appropriate to receive drug); second: if ligands 
can be documented, then combination therapies could be explored.”  
 
Laszlo GS, Gudgeon CJ, Harrington KH, Walter RB. 2015. T-cell ligands modulate the cytolytic 
activity of the CD33/CD3 BiTE antibody construct, AMG 330. Blood Cancer J. 5:e430. 
Funding for this research was provided by the NIH and Amgen Inc 
 
 
 


Not a member yet