Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=tepm20

Download by: [73.163.185.95] Date: 15 February 2016, At: 07:21

Expert Review of Precision Medicine and DrugDevelopmentPersonalized medicine in drug development and clinical practice

ISSN: (Print) 2380-8993 (Online) Journal homepage: http://www.tandfonline.com/loi/tepm20

Challenges & Perspectives of ImmunotherapyBiomarkers & The HistoOncoImmune™Methodology

Christopher Ung & Mark M. Kockx

To cite this article: Christopher Ung & Mark M. Kockx (2016) Challenges & Perspectives ofImmunotherapy Biomarkers & The HistoOncoImmune™ Methodology, Expert Review ofPrecision Medicine and Drug Development, 1:1, 9-24

To link to this article: http://dx.doi.org/10.1080/23808993.2016.1140005

Accepted author version posted online: 12Jan 2016.Published online: 09 Feb 2016.

Submit your article to this journal

Article views: 58

View related articles

View Crossmark data

PERSPECTIVE

Challenges & Perspectives of Immunotherapy Biomarkers & TheHistoOncoImmune™ MethodologyChristopher Ung and Mark M. Kockx

HistoGeneX NV, Antwerp, Belgium

ABSTRACTThe confluence of the immunology and oncology tributaries has precipitated a biomarkerturbulence that is novel to the cancer community. On the heels of reversing the dismal outcomesfor numerous patients suffering from terminal disease, it has also left behind deep pools ofbiomarker questions – which ones, when to use them, how to use them and what cells expressthem – to name just a few. The tumor microenvironment, an arena potentially housing a complexmix of immune infiltrators in the midst of tumor cells, is a wildly heterogeneous terrain, one thatseverely challenges our single biomarker model. This article examines the recently FDA-approvedbiomarker assays for PD-1 therapeutics, considers the technical shortcomings of the singlebiomarker model, scans alternate biomarker technologies and proposes a model that mayallow investigators a method to systematically address these complexities.

ARTICLE HISTORYReceived 18 November 2015Accepted 6 January 2016Published online 9 February2016

KEYWORDSHistoOncoImmune;immunotherapy; biomarkers;PD-L1; diagnostics; inflamed;non-inflamed; infiltratingT-cells; immuno-oncology

Introduction

For an increasing subset of metastatic non-small celllung cancer (NSCLC) patients, checkpoint immunother-apy – a completely new mode of treatment – providesinvaluable options. Our recent ability to nudge theimmune system to actively mount a defense againstcancer cells has created options for physicians andpatients in desperate situations. In October 2015 alone,the US FDA rendered approval for two immune-oncol-ogy therapeutics that target the PD-1 receptors on Tcells, doing so in accelerated cadence and providing anear photo-finish drama. On 2 October, pembrolizumab(Keytruda™) earned the coveted position of the firstcheckpoint immunotherapy to be approved for meta-static NSCLC patients whose disease has progressed fol-lowing other treatments and whose tumors express aprotein called programmed death ligand 1 (PD-L1) [1].Exactly a week later, nivolumab (Opdivo™) also earnedthe right from the FDA to be used with the same patientgroup [2].

This October flurry of regulatory approvals has thepromising likelihood of continuing in multiple tumortypes as investigators and drug development compa-nies begin to unveil the potential of immune-oncologydrugs in other cancers. Previously, the FDA had alreadydeemed pembrolizumab [3] and nivolumab [4] to besafe and beneficial for patients with advanced mela-noma as single agents and even in combination with

other immuno-oncology therapeutics [5]. Nonetheless,the recent approvals in NSCLC are significant since bothare associated with PD-L1 biomarker assessment.

An emerging dilemma

Many studies across a diversity of tumor types identi-fied a relationship between the pretreatment expres-sion level of PD-L1 in the tumor microenvironment andthe likelihood of response to single-agent PD-1 path-way inhibitor therapy [6]. Taube et al. conducted anenlightening literature meta-analysis of solid tumorswhere they observed an increased response to check-point therapy when PD-L1 was expressed [7]. This pro-mising observation was rapidly tempered by numerousquestions as studies emerged with conflicting resultswhether PD-L1 served as a predictive biomarker [8].

The simultaneous FDA approval of two similar butdistinct biomarkers tests acknowledges the potentialrole of PD-L1 as a predictive biomarker while triggeringthe practical challenges of using these tools. It isinstructive to trace the array of questions that lurk atthe diagnostic–therapeutic interface in light of the reg-ulatory endorsement of these novel diagnostics. TheFDA approvals of pembrolizumab and nivolumab arrivewith different diagnostic guidelines for use despite thesimilarity of the intent to treat population group.Pembrolizumab, for instance, requires the use of a

CONTACT Christopher Ung christopher.ung@histogenex.com HistoGeneX NV, Lindendreef 1, B-2020 Antwerp, Belgium

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT, 2016VOL. 1, NO. 1, 9–24http://dx.doi.org/10.1080/23808993.2016.1140005

© 2016 Taylor & Francis

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

companion diagnostic test, the PD-L1 immunohisto-chemistry (IHC) 22C3 pharmDx test; ostensibly the firstFDA-approved test to detect PD-L1 expression in non-small cell lung tumors. The approval was based on datafrom the phase I KEYNOTE-001 trial, in which the overallresponse rate (ORR) with the drug was 41% among asubgroup of 61 patients with pretreated PD-L1-positiveadvanced NSCLC as determined by the 22C3 pharmDxdiagnostic test [9].

Nivolumab, on the other hand, is free of such aconstraint. The decision to prescribe nivolumab toadvanced NSCLC patients remains with the physician.Nonetheless, Richard Padzur, M.D. Director of the Officeof Hematology and Oncology in the FDA’s Center forDrug Evaluation and Research, acknowledged theimportance of PD-L1 as a biomarker in the approvalannouncement, ‘While Opdivo showed an overall survi-val benefit in certain non-small cell lung cancerpatients, it appears that higher expression of PD-L1 ina patient’s tumor predicts those most likely to benefit’[2]. Further, the FDA acknowledged that the level of PD-L1 IHC expression may help identify patients who aremore likely to live longer due to treatment with nivo-lumab by approving a corresponding test. Thus, PD-L1IHC 28-8 pharmDx test arrives with a ‘complementarydiagnostic’ designation, a test capable of helping phy-sicians determine which patients may benefit mostfrom treatment with nivolumab, but notably is not arequired test [10].

A physician who is considering the use of pembro-lizumab or nivolumab for NSCLC patients would needto ponder several use considerations of their respec-tively approved PD-L1 IHC biomarker, both of whichare inconveniently distinct from each other.Pembrolizumab use requires the use of a specificIHC PD-L1 test while the nivolumab use associatespossible benefits of its IHC PD-L1 test. The intrigueof this decision admixture deepens since it is thesame manufacturer, employing a similar technology,who makes both the pembrolizumab companiondiagnostic test and the nivolumab complementarydiagnostic test.

The near term NSCLC checkpoint therapy landscapepromises to complicate these decisions further.Atezolizumab and durvalumab, both of which are

investigational agents that target PD-L1, the ligandpartner of PD-1, are also likely to be tethered to indivi-dually specified PD-L1 IHC tests that are not only dif-ferent from each other, but also the ones associatedwith their recently approved PD-1 inhibitor counter-parts. The entry of institutional cost and logistics con-cerns will certainly convolute the decision-makingmatrix for the physician and pathologist. While thephysician suffers from ‘PD-L1 test confusion,’ thelaboratory pathologist is susceptible to the confusionof which PD-1 or PD-L1 checkpoint therapy is beingconsidered. The pathology laboratory will need to grap-ple with the cost of running multiple tests, wrestle withthe ethics of using up patient material for the multipletests, and payers may resist compensating for non-FDA-approved tests. Historical precedence does not offermuch relief to the confusion, since the survival ofapproved similar companion diagnostics for the samepopulation such as for BRAF mutations in melanoma isquestionable [11]. If the tandem of diagnostics asso-ciated with the PD-1 inhibitors of pembrolizumab andnivolumab has potentiated companion diagnostics tur-bulence, the arrival of a second pair with upcoming PD-L1 inhibitors will ensure that all keenly feel theturbulence.

Ongoing biomarker complexities in immuno-oncology drug development

Notwithstanding the FDA’s acceptance of PD-L1diagnostics in the NSCLC indication, the debateregarding PD-L1 as a biomarker continues to prolif-erate in an ever-expanding matrix of issues. Whilephysicians now possess tools to assess PD-L1 expres-sion in the NSCLC setting, drug developers anddiagnostic companies still need to overcome a longchecklist of challenges, many of which still confoundthe medical community precipitated byHercepTest™, which began the stream of FDA-approved targeted diagnostics [12]. Table 1 sum-marizes these challenges. It is instructive to reviewthese expressed challenges as we embark on thequest to refine a biomarker model for immune-oncology drug development.

Table 1. Current impact and challenges of PD-L1 IHC testing.Challenge Issue Impact

Technical Multiple methods and primary antibodies High variability in staining performance and multiple readoutsMultiple cutoffsPre-analytic conditions

Biological Tumor cells or immune cells Failure to adequately capture complexityPD-L1 IHC is dynamically heterogeneous

10 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

Multiple methods and primary antibodies

The results seen in the literature associating PD-L1expression with immunotherapeutic outcomes wereobtained with multiple different IHC assays. Multipleunresolved issues confound these multiple assays.Different antibodies, different staining protocols, differ-ent target cell assessment (tumor cells, tumor-infiltrat-ing immune cells, or both), different scoring methods(percentage of staining cells, IHC 1–3), and differentthresholds for defining a positive test result, all contri-bute to a rather unstable PD-L1 landscape [13].Therefore, each assay is likely to exhibit substantiallydifferent performance but without any arbitration todetermine its biological relevance or clinical utility. Wehave recently compared several PD-L1 clones using anautomated, standardized, SOP-driven protocol anddemonstrated rather interesting observations [14]. Forinstance, the E13LN clone from Cell SignalingTechnologies demonstrates a greater sensitivity on mel-anoma tumor cells compared to the SP142 clone fromSpring Bioscience. This feature highlights potentialdilemmas since tumor cell expression of PD-L1 IHC isa prominent readout for checkpoint immunotherapypatient selection.

Multiple cutoffs

The different assays employ different scoring methodsand unique thresholds to define a positive result.Table 2, restricted only to NSCLC-related PD-L1 IHCcutoffs, provides a hint of the impending complexityas the immunotherapy agents for other tumor typesmake their way into clinical settings.

Pre-analytic conditions

The powerful and valuable morphological informationfrom IHC is tempered by the requirement to be keenlyaware of the careful and meticulous considerations ofanalytical and pre-analytical conditions that are neces-sary to ensure relevant and reproducible outputs. Whilediagnostic manufacturers may remove many analyticalvariables through their manufacturing and automatedplatforms, the conditions prior to analysis remain out of

their reach. Lung cancer biopsies are on average verysmall which render them particularly vulnerable tothese pre-analytical conditions. The type of fixative,length of fixation times, the amount of time it takesfrom surgery before the specimen is placed into fixa-tive, cold ischemia time, and a variety of other issuesaffect the output of IHC [20]. Since checkpoint therapiessuch as nivolumab and pembrolizumab are likely tosecure pillar therapy roles, and therefore we can expectPD-L1 IHC testing to increase in prominence and fre-quency, pathologists and physicians will need to beaware of and account for the influence of pre-analyticconditions on their workflow and subsequent patientmanagement.

Multiple cell types can express PD-L1

PD-L1 can be expressed in multiple cell types, such astumor cells, macrophages, and lymphocytes [21,22].Figure 1 shows examples of PD-L1 IHC staining ontumor and immune cells from our laboratory. The clin-ical relevance of PD-L1 expression on the different celltypes remains a topic for further investigation.Assessment of PD-L1 on tumor cells has dominatedmost of the studies, as is the case for the recentlyapproved companion diagnostic test for pembrolizu-mab and the complementary diagnostic test for nivolu-mab. However, the assessment of PD-L1 is likely toexpand to tumor-infiltrating immune cells since theatezolizumab clinical trials are studying patients basedon their PD-L1 IHC expression on both tumor cells (TC)and immune cells (IC). Using the SP142 antibody, theexpression of PD-L1 IHC in the atezolizumab NSCLCstudies was captured using a dual scale of intensity,from low expression (TC0/IC0) to high expression(TC3/IC3). Using this combined immune-tumor cell PD-L1 expression readout, Spira et al. concluded that theefficacy of atezolizumab increased with higher PD-L1expression in the POPLAR trial [23]. This observationappears relevant in other tumor types, as studied byHerbst and colleagues [24]. While evaluating the effi-cacy of atezolizumab, he found that of the 175 efficacy-evaluable patients, confirmed objective responses wereobserved in 32 of 175 (18%), 11 of 53 (21%), 11 of 43

Table 2. Cutoff values of PD-L1 IHC assays associated with PD-1 and PD-L1 checkpoint agents for treatment of NSCLC [15].

Nivolumab [16]Pembrolizumab

[17]Atezolizumab

[18]Durvalumab

[19]

PD-L1 IHC antibody 28-8 22C3 SP142 SP263Cutoff values ≥1% to ≥10% ≥1% to ≥50% ≥1% to ≥50% (tumor cells)

≥1% to ≥10%(immune cells)(IHC 1, 2, or 3)

≥25%

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT 11

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

(26%), 7 of 56 (13%), and 3 of 23 (13%) of patients withall tumor types, NSCLC, melanoma, renal cell carcinoma,and other tumors (including colorectal cancer, gastriccancer, and head and neck squamous cell carcinoma).

PD-L1 IHC is dynamically heterogeneous

The sophisticated and dynamic biology of PD-L1 IHCexpresses itself through significant variation by tumortype across disease sites making it one of the morecomplicated biomarkers to guide potential clinical uti-lity. Multiple characteristics contribute to the heteroge-neity of PD-L1 expression. There are inter- andintratumoral differences [25], it is inducible anddynamic [26], it expresses on membrane and in thecytoplasm and, as described above, its expressionsvary by cell types.

Taube and colleagues illustrate this variability acrosstumor types, within the tumor itself, between the pri-mary and metastatic sites, and between cell types [27].The PD-L1 expression heterogeneity may silently influ-ence downstream patient management, potentially inthe form of false-negative results that result from asample that has not included the area of the tumorthat expresses PD-L1. Conceivably, because of expres-sion heterogeneity, the biopsy of one lesion may not bepredictive of an immune response ongoing at anothersite, offering a possible explanation for respondingpatients who test negative for PD-L1.

We participated in a study that investigated the PD-L1 expression heterogeneity of cell types, whichrevealed two distinct NSCLC subtypes, based on PD-L1expression on immune versus tumor cells. Strikingly,the tumors with high PD-L1 expressing tumor cells

BA

DC

E F

stroma

*

Figure 1. PD-L1 IHC staining on non-small cell lung cancer patient specimens. (A) Low density of immune cell infiltrates expressingPD-L1. (B) Moderate density of immune cell infiltrates expressing PD-L1. (C) High density of immune cell infiltrates expressing PD-L1.(D) Moderate density of immune cell infiltrates and tumor cells expressing PD-L1. (E) Magnification of boxed area in panel C. Denseinfiltrate of PD-L1 positive immune cells (arrowheads), both in the carcinoma as the stromal tissue. Other cells in the stroma (mostlyB-cells) as well as the tumor cells in this patient (arrows) are PD-L1 negative. (F) Magnification of boxed area in panel D. Tumor cellsare positive for membranous PD-L1 staining (arrows). In this patient, the immune cells are also PD-L1 positive (arrowheads), boththe dispersed cells and immune cell aggregates (white asterisk). Scale bar: A-D = 200µm, E-F = 100µm.

12 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

and the tumors with high PD-L1 expressing immunecells represented two distinct populations, with mini-mal overlap. Further examination revealed the dramatichistopathology and molecular differences for each sub-type such as the positioning of infiltrating immune cellsand the morphologic makeup of the tumor microenvir-onment [28].

Other biomarkers may prove useful

Numerous biomarker hypotheses focus on identifyingresponding populations and to deciphering themechanisms of action of responding patients who arebiomarker negative. A third investigational group intro-duces another round of exploratory biomarker hypoth-eses and analyses – the patient group who do notrespond to immunotherapies. These non-respondingindividuals may benefit from other treatment strategiesthat include combinations with other immunotherapies,targeted therapies, and even chemotherapies. Thesearch for biomarkers for all three patient groups iscertainly daunting but nonetheless intriguing possibili-ties exist.

Le et al. helmed an interesting preliminary study,indicating microsatellite instability measurement as abiomarker for gauging the response to immunotherapyin colorectal cancer patients. In their small study, thecohort of patients with mismatch-repair deficient color-ectal cancers exhibited an immune-related objectiveresponse rate of 40%. The corresponding responserate for the patient cohort with mismatch-repair profi-cient colorectal cancers was 0% [29]. Llosa et al. havedocumented similar observations showing vigorouscheckpoint activity in the immune microenvironmentof microsatellite instable colon cancers [30].

Mutational load is emerging as a crude global bio-marker for checkpoint therapy benefit fueled by agrowing acceptance that tumors with high mutationalload respond vigorously to checkpoint therapy, irre-spective of PD-L1 expression. Highly mutated tumorswith 1% or even nominal PD-L1 expression wouldrespond almost universally to the checkpoint inhibitors[31]. However, Rizvi also points out a group of patientswith highly mutated lung cancers whose tumors werePD-L1 negative who failed to respond to checkpointinhibitors. Thus, while mutational frequency serves asa ‘global’ measure across the tumors, in instanceswhere PD-L1 IHC expression with <1% of cells in thetumor microenvironment, it may still serve as a biomar-ker for non-response to checkpoint blockade when it isseen in a spatial relationship with host immune cells,thus preparing both physician and patient for thepotential need of additional countermeasures.

Several studies have linked the presence of tumorinfiltrating immune cells to prognostic and predictivebenefit [21]. Clinical trials with PD-L1 and PD-1 blockadesuggested that tumors with a high number of inflam-mation-causing T cells were more responsive to theimmunotherapy-based PD-L1 and PD-1 inhibitors.Tumors with low inflammation, or low numbers of Tcells, were less responsive, highlighting the potentialrole of cytotoxic T-cell biomarkers such as CD8 [32].

This in turn has stimulated the identification of mole-cular factors that inhibit immune infiltration intotumors. Gajewski provides an intriguing description ofthe T-cell inflamed versus non-inflamed solid tumor,where the former is an attempt by the host to organizean antitumor response while the latter often has thehallmark of not benefiting from checkpoint therapies[33]. Thus, unpacking the molecular detail of bothinflamed versus non-inflamed tumors may provideclues to checkpoint-relevant interrogative biomarkers.For example, Gajewski spotlighted beta-catenin as afactor that may distinguish immune inflamed versusnon-inflamed melanomas [34]. This premise also raisesthe intriguing role of beta-catenin to stimulate shuttlingof cytotoxic T cells into the tumor region. Interferon-gamma (IFN-γ) may provide additional definition to thisinflamed phenotype. In a small but interesting study,Ribas used a gene expression signature centered onIFN-γ to show ORR and progression-free survival asso-ciations for melanoma patients treated with pembroli-zumab [35]. Seiwert provided similar conclusions forhead and neck cancer [36]. Both investigators reliedon a combination of (1) ‘IFN- γ’, (2) ‘TCR Signaling’,and (3) ‘expanded immune’ gene expression signatures.

Epigenetic biomarkers may play an important rolewith the recent understanding that selective epigeneticreprogramming may increase immunotherapyresponses. A team from the University of Michiganreports that the invasion of the tumor microenviron-ment by T cells is impeded by selective gene methyla-tion and concludes that epigenetic reprogrammingmay condition tumors from poor T-cell infiltration torich T-cell infiltration, and ultimately potentiate cancertherapy [32].

Additional biomarkers, many of them available asIHC assays, continue to emerge and are worth monitor-ing. The CTLA-4, TIM3, and LAG3 receptors are part ofthe complex set of negative T-cell regulatory switchesand provide varying levels of response informationdepending on tumor types and therapeutic confronta-tion. Investigational inhibitors against these checkpointreceptors are advancing in the clinic, generating com-pelling biomarker data that provide further mechanisticinformation to predict response and resistance [26].

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT 13

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

Programmed death ligand 2 (PD-L2), a known ligand ofPD-1, has begun to attract investigational scrutiny,often in concert with its better-known counterpart,PD-L1. The recent observations from Yearley, that PD-L2 expression may be discordant to PD-L1, and that PD-L1-negative patients who respond to PD-1 inhibitorsmay have an active PD-L2 pathway, provide glimpsesof the potential role of PD-L2 in immunotherapy [37].Previously, Herbst and colleagues also showed thatNSCLC patients who have high PD-L2 expressionresponded to atezolizumab [24].

The multifactorial future

The two recently regulatory-approved PD-L1 IHC testswill sculpt their roles as companion and complementarydiagnostic for their respective PD-1 inhibitors, providingclinicians and NSCLC patients with a level of assurancewhere positive expressers have a strong probability ofresponse. As investigational agents atezolizumab anddurvalumab enter the market with their respectivediagnostics, the questions will increase and only addi-tional studies will provide guidance on how and whento use these diagnostics. The FDA convened a publicworkshop to surface these issues and has mandated acomparison of the different PD-L1 diagnostics [38].

Meanwhile, as the transition from chemotherapy tomolecular therapeutics, and now to immune therapeu-tics, continues to develop, investigators will launchmultifaceted studies to identify biomarkers with greaterresponse or resistance resolution to single or multi-agent therapy regimens. Many concede that it is impos-sible for a single biomarker to represent the complexand dynamic nature of the human immune system,much less when confronted in tandem with the tumormicroenvironment. Peters et al. express an increasinglyvalid perspective where ‘the complexity of immunesurveillance and escape might prevent us from identify-ing a simple and unique predictive biomarker’ [39]. Theemerging model is to review multiple immune check-points and other factors within the tumor that may becontributing to immune exclusion or susceptibility toimmune attack. By getting a multifactorial assessmentof what is happening within the tumor, investigatorsshould be better able to derive indicators to rationallydeliver individualized therapy, either monotherapy orcombinatorial immunotherapy.

We propose a systematic application of amethodologythat harnesses histopathology data of the tumor micro-environment with its correspondingmolecular signatures.The activity of infiltrating immune cells into the tumormicroenvironment is effectively captured through spatialand pictorial representations that viscerally inform of the

immune activity. A molecular profile meanwhile providesamultiplex picture, charting out genes that are associatedwith such activity, often unveiling underlying networks ofsignaling protein activity. This ‘histoinformatics’ approach,the integration of histopathology andmolecular mappingof the immune tumor microenvironment when applied ina systematic way, provides investigators a method tounderstand the tumor microenvironment activity and itsinterface with the immune system.

The HistoOncoImmune™ system

We present a methodology that attempts to encompassthese elements, employing the most relevant technol-ogy to systematically capture multiple points ofdynamic biomarker dynamic data across a heteroge-neous terrain. There are several guiding principlesunderlying the design of this system (Figure 2):

● This methodology represents an organized plat-form that is dynamically adapted to optimize thebiomarker(s) selection for the investigationalimmune-oncology agent.

● The selected biomarkers are rigorously validated,SOP-encapsulated and deployed, operationally scal-able and can be deployed across multiple clinicaltrials of various sizes and geographical locations.

● The output (HistoMineTM) provides a representationof histopathology and molecular data.Histopathology evaluations and scores, high-resolu-tion whole slide images, and tissue patterns arereported in conjunctionwith numerical gene expres-sion and next generation sequencing (NGS) data.

The collection of biomarkers is organized in two groups:Category 1 serves as a histopathology-centered base-line panel that can be used across multiple trials todescribe the patient’s immune-tumor condition, whileCategory 2 provides supporting molecular data.

Category 1: description

In this histopathology driven mode, a qualified pathol-ogist first assesses the patient’s H&E stained tumorslides, confirming tumor presence and diagnoses,reviewing for tissue and staining quality, selectingregions of interest for high definition image analyses,and identifying extraction areas that are subsequentlyused for the molecular analyses in Category 2 testing.This H&E assessment is a critical step and anchors allHistoOncoImmuneTM biomarker activity. The data out-put from Category 1 consists of high-resolution imagescaptured from microscope slides, morphological

14 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

description of the immune and tumor activity, andsemi-quantitative measurements of targets of interest.

The Category 1 biomarkers provide a panoramicsurvey of the immune activity of the tumor microenvir-onment (Figure 3). These immunostains identify thepresence and location of the relevant immune-oncol-ogy target (e.g. PD-L1), cytotoxic T cells (CD8), tumor-associated macrophages (CD163), immune regulatorycells (FOXP3/CD3), and B cells (CD20).

Intentionally, regulatory-approved or the most widelyadopted versions of Category 1 biomarkers are used foranalyses to increase the likelihood that data acrossmultipleclinical trials can be compared and correlated. Furthermore,the validation of these biomarker assays includes an accu-racy assessment, assay linearity, assay precision acrossusers and platforms, and stability review of the targetantigens and assay reagents [40,41]. Additional experi-ments are also conducted to understand and refine assay,tissue, or antigen parameters (e.g. heterogeneous expres-sion, antigen stability, fixation sensitivities, etc.)

Further, we employ quantitative imaging techniquesand algorithms that enable us to systematically captureimmunostaining values that can be reported for regionsof interest from single cells to disease compartments.The CD8 biomarker is used for illustration.

Following pathology review and CD8 immunostain-ing on an automated IHC stainer, imaging scientistsdefine the pathologist-selected area into three possiblecompartments (Figure 4A): the tumor center (Cn), theperiphery of the tumor (Pe), and the invasive margin(IM). A selection of pixels, rather than the entire pixelset, is analyzed within each component in order toenable analysis in a timely way and to work with adata file that is manageable (Figure 4B). We accomplishthis by directing the software to select a random num-ber of pixels, optimized and validated such that incre-mental pixels are no longer informative. The ratio ofstained area versus total pixel area is captured for eachpixel, essentially providing macro-immunostainingvalues for each component that can be used for regres-sion and correlation across studies. This immunostain-ing value is available for all three compartments andcan be analyzed at various levels, providing granularinformation to understand biological phenomena suchas heterogeneity, simply by selecting the appropriateregion of interest, ostensibly even at the single celllevel. The relative area ratio was selected as a metricover cell counts since the deconvolution of the latterbecomes exceedingly complex in certain situationssuch as the heavy clustering of cells.

HistoOncoImmune

Panel

™Integrates histopathology

with molecular data for

immuno-oncology trials

Consistent biomarker

platform across trials

An adaptive methodology

enables quick biomarker

“kills” and substitutes in

new candidates

Pathology evaluation

High resolution whole slide image

Pattern analysis

Histopathology score

Imaging semi-quantitative readout

Gene expression & NGS data

This information is stored

in HistoMine™ - a dynamic,

searchable, database

Figure 2. The HistoOncoImmune™ Panel is a panel of biomarkers that enables systematic interrogation and integration ofhistopathological and molecular landmarks. This methodology represents an organized platform that is dynamically adapted tooptimize the biomarker(s) selection for the investigational immune-oncology agent. The selected biomarkers are rigorouslyvalidated and can be deployed across multiple clinical trials of various sizes and geographical locations. The output consists ofpathology evaluations and scores, high-resolution whole slide images and tissue patterns and are reported in conjunction withnumerical gene expression and next generation sequencing (NGS) data. Data is subsequently stored in a dynamic searchableHistoMine™ database.

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT 15

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

Figure 4C illustrates the immunostaining values thatare available by the relevant melanoma compartments.Thus, the immunostaining values for tumor centers,tumor periphery, and invasive margins can be com-pared at several levels, pre and post dose, patientswithin a study and tumors across multiple studies.These values are retained as metadata associated withthe image and can be retrieved as needed to providethe investigator with contextual morphology.

The morphological power of Category 1 biomarkersis tempered by the requirement that each biomarkerrequires a single slide. (The authors are evaluating tech-nologies that leverage the ability to align seriallystained slides to create a virtual multiplex slide.)Neither does it reveal the joint complexity of immuneand tumor signaling activity that lies behind the ‘sign-post’ Category 1 biomarkers. The Category 2 biomarkersassume these responsibilities.

Category 2: description

The biomarkers in this category correlate the signalsobtained from the Category 1 grouping and also providesupplemental pathway and actionable guidance for drug

development and patient selection (Figure 5). Importantly,the technologies espoused in this category use aminimumof patient material but elicit a large set of data.

We leverage the recent availability of platforms thatenable the simultaneous mapping of genes in formalin-fixed, paraffin-embedded materials to assess the geneexpression activity within a patient’s tumor microenvir-onment as indicators of tumor proliferation, susceptibil-ity of the disease to an investigational agent, the tumor’spathway addiction, and many other mechanistic reveals.

HistoOncoImmuneTM incorporates the nCounter®PanCancer Immune Profiling Panel (NanostringTechnologies®) as its tool to document the humanimmune response of various tumor types [42]. This770 gene panel combines markers for 24 differentimmune cell types and populations, 30 common cancerantigens and genes that represent all categories ofimmune response including key checkpoint blockadegenes. As discussed earlier, the assessment of PD-L1expression on tumor cells versus immune cells, or intandem, remains an issue requiring further investiga-tion. One limitation encountered with the nCounterplatform is its inability to differentiate between thePD-L1 gene expressions on tumor versus immune

Figure 3. The Category 1 panel of the HistoOncoImmune™ system is a baseline panel consisting of regulatory-approved or the most widelyadopted histopathology markers to increase the likelihood that data across multiple clinical trials can be compared and correlated. It startswith the H&E assessment by a trained pathologist, followed by a selected panel of markers that determine the immune activity of the tumormicroenvironment. These immunostains identify the presence and location of the relevant immune-oncology target (e.g. PD-L1), cytotoxicT-cells (CD8), tumor-associated macrophages (CD163), immune regulatory cells (FOXP3/CD3) and B-cells (CD20).

16 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

A

B

C

Pe

IM

Pe

IM

Figure 4. Pathology evaluation of a melanoma resection and subsequent tri-compartmentalization of the image for CD8 analyses.(A) A qualified pathologist first assesses the patient’s H&E stained tumor slides, confirming tumor presence and diagnoses,reviewing for tissue and staining quality, selecting regions of interest for high definition image analyses. Following CD8immunostaining on an automated IHC stainer, imaging scientists define pathologist-selected areas into 3 possible compartments:the tumor center (Cn), tumor periphery (Pe) and the invasive margin (IM). (B) Using imaging analysis software a selection of pixels,rather than the entire image is analyzed. The software is designed to select a random number of pixels, optimized and validated,such that incremental pixels are no longer informative. The ratio of stained area versus total pixel area is captured for each pixel. Inthis example 1,170 pixel tiles were analyzed and tile examples are shown to the right. (C) Example of immunostaining values thatare measured in the relevant melanoma compartments. Cn, the tumor center; Pe, tumor periphery; IM, invasive margin. Scale bar =5000µm.

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT 17

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

cells. As an aggregate, we observe that the PD-L1 geneexpression obtained via the nCounter platform corre-lates exceedingly well with our validated PD-L1 IHCassays producing robust R2 values (Figure 6, Schatset al., unpublished data). Furthermore, we are encour-aged by an equally strong correlation with other ortho-gonal platforms such as real-time PCR and mRNA in situhybridization.

Interestingly, there is an unraveling of this correlatewhen only the PD-L1 gene expression of immune cellswas compared to the IHC analogs, a demonstration of theneed for combinatorial techniques to unpack the partici-pating mechanisms of action. Further, while it is temptingto rely on a platform such as nCounter, which offersnumerical and operational simplicity compared to its IHCanalog, there needs to be an in-depth understanding of

Figure 5. The markers in the Category 2 panel of the HistoOncoImmune™ system complement the markers of Category 1. They areobtained using platforms that enable the simultaneous mapping of genes in formalin-fixed, paraffin-embedded tissue to assess thegene expression activity within a patient’s tumor microenvironment as indicators of tumor proliferation, susceptibility of the diseaseto an investigational agent, and many other mechanistic reveals. Ultimately, we aim to combine gene expression profiling using thenCounter® PanCancer Immune Profiling Panel (Nanostring Technologies®) with Next Generation Sequencing and the measurementof relevant moieties in plasma (liquid biopsies).

Figure 6. Analysis of metastatic melanoma samples using the nCounter Nanostring platform and IHC for PD-L1 (clone E1L3N) and CD8(clone C8/144B). PD-L1 and CD8 mRNA expression correlates exceedingly well with our validated IHC assays producing robust R2 values.

18 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

the frequency of discrepant cases between the platformsand their significance. The relevance of high PD-L1 geneexpression with concomitant low PD-L1 immune cellexpression cases remains to be understood in terms ofits frequency and relevance to PD-1/PD-L1 checkpointtherapy. The integrated analyses captured via theHistoOncoImmuneTM methodology allows for systematicreview of such cases and creating further hypotheses forclinical trials.

The targeted NGS component of HistoOncoImmuneTM

system acknowledges the need to identify actionablemutations while allowing for, and detecting, emergingor evolving resistancemutations. The diagnostic or clinicalmode of this NGS panel has its premise on an SNP panel(26 genes, 200+ amplicons) compiled by the FrenchInstitut National du Cancer who compared and analyzedseveral globally well-regarded panels that guide the iden-tification of driver and resistance mutations for multipletumor entities [43]. This panel combines a proprietarymultiplex PCR technology with subsequent NGS testing[44]. The investigational mode of our NGS panel, mean-while, will be used to detect indels, employing a combina-tion of split-read and paired-end approach to identifysmaller and larger variants, respectively.

Combined, the nCounter platform, targeted NGS,and indel NGS provides a molecular panorama of theimmune tumor environment using a standard set ofpanels to accommodate significant evolving and emer-ging resistance genes and mutations, a research modeadmits the novel targets into a rigorous validation

process, containing well-validated orthogonal methods.Once complete, these targets are then included into thegene expression codeset or the appropriate NGS panel.

A key step to molecular analysis especially for profilinghuman immune response to immunotherapeutics, suchas checkpoint blockade, is the proper selection of tissuearea for analysis. Critically, the selection of the regionshould include, where possible, the interface areabetween tumor and normal tissue since the infiltratingimmune cells tend to congregate there. Improper selec-tion may lead to misleading molecular data. On the onehand, a cursory extraction of the entire section withoutproper pathology selection would dilute active geneexpression information by the inclusion of too much nor-mal material. On the other hand, an overly enthusiastic ofa tumor-only sectionwould sacrifice the valuable informa-tion of immune entities gathered at the interface.

The final component of Category 2 testing accom-modates characteristics specific to tumor types or par-ticular immune features. As discussed earlier,microsatellite instability, other cancer testis agents, epi-genetic events, certain oncogenic viruses, and other IHCbiomarkers may provide data that bring focus to themolecular panorama of a complex terrain. We will alsoleverage the intriguing findings of Gajewski and collea-gues, and incorporate markers such as beta-catenin IHCin concert with T-cell activity signatures such as IFN-γand granzyme B to distinguish the relevance ofinflamed and non-inflamed tumors [33]. Figure 7demonstrates PD-L1 expression of an NSCLC

stroma

Figure 7. PD-L1 IHC staining of a NSCLC specimen showing distinguishing morphological features of homogenous, intense andcomplete membrane patterns on clustered tumor cells only and no immune cell expression. The stroma is fibrous withoutinfiltration of immune cells. Scale bar = 100µm

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT 19

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

specimen with distinguishing morphological features.Its PD-L1 expression has a high-intensity HER2-likehomogeneity and is restricted only to tumor cellswith no immune cell staining. In addition to thegene signatures mentioned above, as outcome databecome available, we can begin to examine if othermechanisms such as PD-L1 amplification are asso-ciated with such staining patterns [45]. In all cases,these tests are vigorously validated for specificity,sensitivity, and reproducibility.

We have also begun validation exercises for exam-ining relevant moieties in plasma, taking advantageof the vigorous innovations transpiring with the cur-rent interest in liquid biopsies. Nyens recently illu-strated the potential utility of detecting circulatingtumor DNA (ctDNA) as a monitoring tool for patientswith advanced BRAFV600 mutant melanoma [46].Significantly, their study albeit with a small patientsubset highlights the superiority of the patient’sBRAFV600 mutant fraction as a predictor of prolifera-tive tumor burden, rather than the patient’s tumormass.

Conclusion

The range of information required to effectively selectthe best therapeutic combination for a patient hasexpanded enormously with the addition of immune-oncology agents. Patient selection by single biomarkersin this setting is frustratingly inadequate as markedlydemonstrated by IHC PD-L1 assessment which hints atpatient selection strategies yet leaves many questionsunanswered. Meanwhile, questions abound regardingthe responding immune system and its interactionwith the tumors. The additive effect of having toprobe the immune system within the tumor microen-vironment with multivariate biomarker analyses forcesquestions on several dimensions. First, what are the setof relevant biomarkers? Next, what platforms should weinterrogate with? Finally, how do we integrate thisinformation?

We designed the HistoOncoImmuneTM system tosystematically integrate each patient’s morphologicaland molecular information in global clinical trials thatcan subsequently be correlated to patient outcomesand be viewed according to tumor types. A mindfulapplication of the HistoOncoImmuneTM system thatrespectfully considers the scarcity of patient material isequally critical and we present a workflow we currentlyemploy to mine the information with intelligent con-sideration to the patient’s biological material (Figure 8).Fundamentally, the HistoOncoImmuneTM will enable

the collection of data that lead to intelligently designedhypothesis generation with the adaptive ability toincorporate distinguishing biomarkers and extinguishnon-yielding ones. Future reports will delineate thecollection of such data and also review the system’sefficacy for patient selection.

Expert commentary & five-year view

Many of us in the pathology community have navigatedthe challenges that PD-L1 IHC has presented, acquiringlessons given to us from previous IHC companion diag-nostics. The technical, standardization and interpretationissues mentioned in this review are ones we have encoun-tered and debated extensively. We established standardi-zation guidelines and as a result have also improvedpathology laboratory systems and processes globally.One of the authors helmed the commercialization ofHercepTest™ and observes the repeating themes, nowbeing presented and debated with greater sophistication,for example, the early engagement of the FDA to createawareness of the different PD-L1 IHC tests through itstown hall meetings and the ongoing comparison study.Notwithstanding these efforts, the pathology laboratorywill again have to immerse themselves in numerous risk–benefit decisions that intertwine best and accurate patientcare with costs and logistics decision factors, all whilesifting through emerging streams of conflicting data. It isdifficult to envision an average pathology laboratory com-mitting to multiple automated staining platforms withmultiple PD-L1 IHC tests on each. They will need to jugglemultiple interpretation schemes, laboring from one toanother depending on the tumor and intended drug oftreatment in question. Many will clamor for standardiza-tion but will do so in the flavor that best suits their envir-onment and needs, including the use of non-proprietarycommercial PD-L1 IHC antibodies in a Laboratory-Developed Test format, an arguably unstandardizedtrend that competes vigorously with preceding compa-nion diagnostics. Since these challenges are familiar, per-haps some of the familiar solutions might also beinstructive. In the case of HercepTestTM, laboratoriesresponded positively to education and dialog. Perhaps aswe sort out the similarities and differences of the regula-tory-branded PD-L1 IHC diagnostics, some energy shouldbe rationed for detailed pathology education of PD-L1 IHCexpression in various tumor types. We have, as haveothers, begun to establish digital libraries of PD-L1 IHC invarious tumor types as part of our HistoOncoImmuneTM

effort so that we can chart this inducible biomarker invarious settings. IHC of immune cell activity has greatpotential value in patient selection and it too deserves

20 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

systematic cataloging and education. This acquirednuanced knowledge of PD-L1 and related expression mar-kers, steadily accumulated by pathologists, may be themost practical and effective solution to enable therapy-guiding benefits and it will position the pathologist wellfor dialogs with oncologists who will also gain their ownnuanced set of checkpoint therapy data.

Financial & competing interests disclosure

This work was supported by the Agency for Innovation byScience and Technology (IWT O&O grant 130894).The authorshave no other relevant affiliations or financial involvement withany organization or entity with a financial interest in or finan-cial conflict with the subject matter or materials discussed in themanuscript apart from those disclosed.

Key issues

● Cancer cells can escape the immune system by overexpressing immunosuppressive factors. The PD-1/PD-L1 checkpoint axis plays a key roleherein. Expression of PD-L1 on the tumor cells protects the tumor cells from cytotoxic CD8+ T cells.

● The development of checkpoint blocking antibodies, such as those directed against programmed death-ligand 1 (PD-L1) or its receptor PD-1, hasdemonstrated significant recent promise in the treatment of an expanding list of malignancies.

● Anti-PD-1 immunotherapeutics nivolumab (Opdivo™) and pembrolizumab (Keytruda™) received FDA approval for the treatment of advancedmelanoma and metastatic NSCLC patients whose disease has progressed after other treatments and whose tumors express PD-L1. Anti-PD-L1agents atezolizumab and durvalumab are under intense investigation.

● The FDA approval of two similar but distinct PD-L1 IHC tests (22C3 pharmDx and 28-8 pharmDx) acknowledges the potential of PD-L1 as apredictive biomarker and helps physicians decide which checkpoint therapy to prescribe. Unfortunately, not all patients respond to thesetherapies, and evaluation of biomarkers associated with clinical outcomes is crucial and ongoing.

● We propose a systematic application of a methodology that harnesses histopathology data of the tumor with its corresponding molecularsignatures. The selected biomarkers are rigorously validated and can be implemented across multiple clinical trials of various sizes andgeographical locations.

● The HistoOncoImmuneTM system is the integration of histopathology and molecular technology and provides investigators and physicians amethod to understand the tumor microenvironment activity and its interface with the immune system. It offers a methodology to attain abiomarker profile that predicts for response or resistance to checkpoint therapy.

Figure 8. The HistoOncoImmune workflow acknowledges the scarcity of available tissue material through two decision points. Non-inflamed tumors are directly exposed to gene expression analysis and subsequently, only the tumors that may benefit from furtherCategory testing are advanced. The morpholocators are unique and noteworthy pathology features that are captured in annotatedand numeric forms using image analyses tools.

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT 21

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

References

Papers of special note have been highlighted as:• of interest•• of considerable interest

1. FDA approves Keytruda for advanced non-small celllung cancer [Internet]. [cited 2015 Nov 9]. Availablefrom: http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm465444.htm

2. FDA expands approved use of Opdivo in advancedlung cancer [Internet]. [cited 2015 Nov 9]. Availablefrom: http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm466413.htm

3. FDA approves Keytruda for advanced melanoma [Internet].[cited 2015 Nov 9]. Available from: http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm412802.htm

4. FDA approves Opdivo for advanced melanoma [Internet].[cited 2015 Nov 9]. Available from: http://www.fda.gov/NewsEvents /Newsroom/PressAnnouncements /ucm427716.htm

5. BMS Receives FDA Approval for Opdivo (nivolumab) +Yervoy (ipilimumab) Regimen in BRAF V600 Wild-TypeMelanoma [Internet]. [cited 2015 Nov 9]. Available from:http://www.drugs.com/newdrugs/bms-receives-fda-approval-opdivo-nivolumab-yervoy-ipilimumab-regimen-braf-v600-wild-type-melanoma-4271.html

6. Mahoney KM, Atkins MB. Prognostic and predictive mar-kers for the new immunotherapies. Oncology (WillistonPark). 2014;28 suppl 3:39–48.

7. Sunshine J, Taube JM. PD-1/PD-L1 inhibitors. Curr OpinPharmacol. 2015;23:32–38. doi:10.1016/j.coph.2015.05.011.

8. Carbognin L, Pilotto S, Milella M, et al. Differential activityof nivolumab, pembrolizumab and MPDL3280A accord-ing to the tumor expression of programmed death-ligand-1 (PD-L1): sensitivity analysis of trials in mela-noma, lung and genitourinary cancers. PLoS One.2015;10(6):e0130142. doi:10.1371/journal.pone.0130142.eCollection 2015

• Meta-analysis of 20 trials (1475 patients) reportingPD-L1 expression and overall response rate to check-point inhibitor therapy, showing benefit even in theabsence of PD-L1 expression.

9. Garon EB, Rizvi NA, Hui R, et al. Pembrolizumab for thetreatment of non-small-cell lung cancer. N Engl J Med.2015;372(21):2018–2028. doi:10.1056/NEJMoa1501824.

• KEYNOTE-001 clinical trial in which the efficacy andsafety of pembrolizumab in patients with advancedNSCLC was assessed.

10. Philips T, Simmons P, Inzunza HD, et al. Development ofan automated PD-L1 immunohistochemistry (IHC) assayfor non-small cell lung cancer. Appl ImmunohistochemMol Morphol. 2015;23(8):541–549. doi:10.1097/PAI.0000000000000256.

11. Rimm DL. Comment to the FDA public workshop 2015 –complexities in personalized medicine: harmonizing com-panion diagnostics across a class of targeted therapies[Internet]. [cited 2015 Nov 10]. Available from: http://www.aacr.org/AdvocacyPolicy/GovernmentAffairs/Documents/Rimm%20Comment%20to%20the%20FDA%20PD-L1%20v2.pdf

12. Barrett C, Magee H, O’Toole D, et al. Amplification of theHER2 gene in breast cancers testing 2+ weak positive byHercepTest immunohistochemistry: false-positive orfalse-negative immunohistochemistry? J Clin Pathol.2007;60(6):690–693. doi:10.1136/jcp.2006.039602.

13. Patel SP, Kurzrock R. PD-L1 expression as a predictivebiomarker in cancer immunotherapy. Mol Cancer Ther.2015;14(4):847–856. doi:10.1158/1535-7163.MCT-14-0983.

• Review of the early clinical experience and develop-ment of PD-L1 immunohistochemistry (IHC) as a pre-dictive biomarker for anti-PD-1 directed therapy inhuman cancer.

14. Schats K, Van Vré E, De Schepper S, et al. PD-L1 IHCvalidation and comparison of E1L3N & SP142 antibodiesin melanoma. Poster session at the meeting of theEuropean Society for Medical Oncology; 2015;Lausanne, Switzerland.

15. Anders RA, Atkins MB, Rizvi N, et al. Biomarkers to guideimmunotherapy treatment of solid tumors: strategies topromote best practices for quality care [Internet]. 2015.[cited 2015 Nov 9]. Available from: http://www.clinicaloptions.com

16. Gettinger SN, Hellman MD, Shepherd FA, et al. First-linemonotherapy with nivolumab (NIVO; anti-programmeddeath-1 [PD-1]) in advanced non-small cell lung cancer(NSCLC): safety, efficacy and correlation of outcomeswith PD-1 ligand (PD-L1) expression. J Clin Oncol.2015;33(suppl):8025. ASCO Annual Meeting.

17. Rizvi NA, Garon EB, Leighl N, et al. Optimizing PD-L1 as abiomarker of response with pembrolizumab (pembro;MK-3475) as first-line therapy for PD-L1–positive meta-static non-small cell lung cancer (NSCLC): updated datafrom KEYNOTE-001. J Clin Oncol. 2015;33(suppl):8026.ASCO Annual Meeting.

18. Horn L, Spigel DR, Gettinger SN, et al. Clinical activity,safety and predictive biomarkers of the engineeredantibody MPDL3280A (anti-PDL1) in non-small celllung cancer (NSCLC): update from a phase Ia study. JClin Oncol. 2015;33(suppl):8029. ASCO Annual Meeting.

19. Rebelatto M, Mistry A, Sabalos C, et al. Development of aPD-L1 companion diagnostic assay for treatment withMEDI4736 in NSCLC and SCCHN patients. J Clin Oncol.2015;33(suppl):8033. ASCO Annual Meeting.

20. Engel KB, Moore HM. Effects of preanalytical variables onthe detection of proteins by immunohistochemistry informalin-fixed, paraffin-embedded tissue. Arch PatholLab Med. 2011;135(5):537–543. doi:10.1043/2010-0702-RAIR.1.

21. Tumeh PC, Harview CL, Yearley JH, et al. PD-1 blockadeinduces responses by inhibiting adaptive immune resis-tance. Nature. 2014;515(7528):568–571. doi:10.1038/nature13954.

•• Pre-existing CD8 T cells distinctly located at the inva-sive tumor margin (melanoma) are associated withexpression of the PD-1/PD-L1 immune inhibitoryaxis and may predict response to anti-PD1 therapy.

22. Sharpe AH, Wherry EJ, Ahmed R, et al. The function ofprogrammed cell death 1 and its ligands in regulatingautoimmunity and infection. Nat Immunol. 2007;8(3):239–245.

22 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

23. Spira AI, Park K, Mazières J, et al. Efficacy, safety andpredictive biomarker results from a randomized phase IIstudy comparing MPDL3280A vs docetaxel in 2L/3LNSCLC (POPLAR). J Clin Oncol. 2015;33(suppl):8010.ASCO Annual Meeting.

24. Herbst RS, Soria JC, Kowanetz M, et al. Predictive corre-lates of response to the anti-PD-L1 antibody MPDL3280Ain cancer patients. Nature. 2014;515(7528):563–567.doi:10.1038/nature14011.

• These data suggest that azetolizumab is most effec-tive in patients with high PD-L1 expression PD-L1 ontumor-infiltrating immune cells.

25. Schalper KA, Carvajal-Hausdorf DE, McLaughlin JF, et al.Significance of PD-L1, IDO-1, and B7-H4 expression inlung cancer. J Clin Oncol. 2015;33(suppl):3067. ASCOAnnual Meeting.

26. Pardoll DM. The blockade of immune checkpoints incancer immunotherapy. Nat Rev Cancer. 2012;12(4):252–264. doi:10.1038/nrc3239.

27. Taube JM, Klein A, Brahmer JR, et al. Association of PD-1,PD-1 ligands, and other features of the tumor immunemicroenvironment with response to anti-PD-1 therapy.Clin Cancer Res. 2014;20(19):5064–5074. doi:10.1158/1078-0432.CCR-13-3271.

28. Schmid P, Kowanetz M, Koeppen H, et al. NSCLC withhigh PD-L1 expression on tumor cells or tumor-infiltrat-ing immune cells represents distinct cancer subtypes.Poster session presented at the meetingof the European Cancer Congress; 2015; Vienna, Austria.

29. Le DT, Uram JN, Wang H, et al. PD-1 blockade in tumorswith mismatch-repair deficiency. N Engl J Med.2015;372:2509–2520. doi:10.1056/NEJMoa1500596.

•• Demonstrates that the mismatch-repair efficiency ispredictive for clinical benefit of immune checkpointblockade with pembrolizumab in colorectal cancerpatients.

30. Llosa NJ, Cruise M, Tam A, et al. The vigorous immunemicroenvironment of microsatellite instable colon canceris balanced by multiple counter-inhibitory checkpoints.Cancer Discov. 2015;5(1):43–51. doi:10.1158/2159-8290.CD-14-0863.

31. Rizvi NA, Hellmann MD, Snyder A, et al. Cancer immunol-ogy. Mutational landscape determines sensitivity to PD-1blockade in non-small cell lung cancer. Science. 2015;348(6230):124–128. doi:10.1126/science.aaa1348.

• Excellent paper in which the authors performed whole-exome sequencing of non-small cell lung cancers trea-ted with pembrolizumab, and showed that higher non-synonymousmutation burden in tumors was associatedwith improved objective response, durable clinical ben-efit, and progression-free survival.

32. Peng D, Kryczek I, Nagarsheth N, et al. Epigenetic silen-cing of TH1-type chemokines shapes tumour immunityand immunotherapy. Nature. 2015. Epub 2015 Oct 26.doi:10.1038/nature15520.

• These data show that hypermethylation of distinctgenes represses the tumor production of Th1-typechemokines and T-cell trafficking to the tumor micro-environment. Treatment with epigenetic modulatorsremoves this repression and increases effector T-celltumor infiltration, slows down tumor progression,

and improves the therapeutic efficacy of PD-L1inhibitors.

33. Gajewski TF. The next hurdle in cancer immunotherapy:overcoming the non-T-cell-inflamed tumor microenvir-onment. Semin Oncol. 2015;42(4):663–671. doi:10.1053/j.seminoncol.2015.05.011.

•• Describes one of the next significant hurdles todevelop new therapeutic interventions to be effec-tive in patients with the non-T-cell-inflamed pheno-type. Proposed is a detailed molecularunderstanding of the mechanisms that explain thepresence or absence of the T-cell-inflamed tumormicroenvironment, which in turn will benefit fromfocused interrogation of patient samples. It is envi-sioned that the end result of these investigations willbe an expanded array of interventions that willbroaden the fraction of patients benefitting fromimmunotherapies in the clinic.

34. Gajewski T, Zha Y, Hernandez K, et al. Density of immuno-genic antigens and presence or absence of the T cell-inflamed tumor microenvironment in metastatic melanoma.J Clin Oncol. 2015;33(suppl):3002. ASCO Annual Meeting.

35. Ribas A, Robert C, Hodi S, et al. Association of responseto programmed death receptor 1 (PD-1) blockade withpembrolizumab (MK-3475) with an interferon-inflamma-tory immune gene signature. J Clin Oncol. 2015;33(suppl):3001. ASCO Annual Meeting.

36. Seiwert T, Burtness B, Weiss J, et al. Inflamed-phenotypegene expression signatures to predict benefit from theanti-PD-1 antibody pembrolizumab in PD-L1+ head andneck cancer patients. J Clin Oncol. 2015;33(suppl):6017.ASCO Annual Meeting.

37. Yearley J, Gibson C, Yu N, et al. PD-L2 expression inhuman tumors: relevance to anti-PD-1 therapy in cancer.Abstract 18LBA. Vienna: European Cancer Congress; 2015.

38. Complexities in personalized medicine: harmonizing com-panion diagnostics across a class of targeted therapies,public workshop 2015. A blueprint proposal for compa-nion diagnostic comparability [Internet]. [cited 2015 Nov9]. Available from: http://www.fda.gov/downloads/MedicalDevices/NewsEvents/WorkshopsConferences/UCM439440.pdf

• Initiation of activities to increase the understandingof the analytic performance of companion diagnosticassays to facilitate testing and decision-making in theclinic.

39. Helwick C. Expert point of view: Solange Peters [Internet].The ASCO Post. 2015;6(18). [cited 2016 Feb 1]. http://www.ascopost.com/issues/october-10,-2015/expert-point-of-view-solange-peters,-md,-phd.aspx

40. HistoGeneX [Internet]. [cited 2016 Jan 22]. Availablefrom: http://www.histogenex.com/next-generation-immunohistochemistry

41. Howat WJ, Lewis A, Jones P, et al. Antibody validation ofimmunohistochemistry for biomarker discovery: recom-mendations of a consortium of academic and pharma-ceutical based histopathology researchers. Methods.2014;70(1):34–38. doi:10.1016/j.ymeth.

42. Dennis L, Kaufmann S, Danaher P, et al. Multiplexedcancer immune response analysis nCounter® pancancerimmune profiling panel for gene expression. NanoString

EXPERT REVIEW OF PRECISION MEDICINE AND DRUG DEVELOPMENT 23

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016

Technologies [Internet]. 2014. [cited 2015 Nov 9].Available from: http://www.nanostring.com

43. Van den Bulcke M, San Miguel L, Salgado R, et al. Nextgeneration sequencing gene panels for targeted therapy inoncology and haemato-oncology. Health TechnologyAssessment (HTA) Brussels: Belgian Health Care KnowledgeCentre (KCE, 2015). KCE Reports 240. D/2015/10.273/26.

44. Multiplicom. Multiplex PCR technology [Internet]. 2015.[cited 2015 Nov 9]. Available from: http://www.multiplicom.com/onco-genetics

45. Barrett MT, Anderson KS, Lenkiewicz E, et al. Genomicamplification of 9p24.1 targeting JAK2, PD-L1 and PD-L2is enriched in high-risk triple negative breast cancer.Oncotarget. 2015;6(28):26483–26493.

46. Schreuer M, Meersseman G, Jansen Y, et al. Quantitativeassessment of BRAF V600 mutant cell-free tumor DNAfrom plasma as a diagnostic and therapeutic biomarkerin pts with BRAF V600 mutant melanoma. Poster sessionat the meeting of the Association of MolecularPathology; 2015; Austin, TX.

24 C. UNG AND M. M. KOCKX

Dow

nloa

ded

by [

73.1

63.1

85.9

5] a

t 07:

21 1

5 Fe

brua

ry 2

016