454
`
`DOI 10.1002/prca.201300095
`
`Proteomics Clin. Appl. 2014, 8, 454–463
`
`REVIEW
`Understanding fibroblast activation protein (FAP):
`Substrates, activities, expression and targeting for
`cancer therapy
`Elizabeth J. Hamson1, Fiona M. Keane1, Stefan Tholen2,3, Oliver Schilling3,4
`and Mark D. Gorrell1
`
`1 Molecular Hepatology, Centenary Institute and Sydney Medical School, University of Sydney, Sydney, Australia
`2 Institute of Molecular Medicine and Cell Research, University of Freiburg, Freiburg, Germany
`3 Faculty of Biology, University of Freiburg, Freiburg, Germany
`4 BIOSS Centre for Biological Signaling Studies, University of Freiburg, Freiburg, Germany
`
`Fibroblast activation protein (FAP) is best known for its heightened expression in tumour
`stroma. This atypical serine protease has both dipeptidyl peptidase and endopeptidase activ-
`ities, cleaving substrates at a post-proline bond. FAP expression is difficult to detect in non-
`diseased adult organs, but is greatly upregulated in sites of tissue remodelling, which include
`liver fibrosis, lung fibrosis, atherosclerosis, arthritis, tumours and embryonic tissues. Due to
`its restricted expression pattern and dual enzymatic activities, FAP is emerging as a unique
`therapeutic target. However, methods to exploit and target this protease are advancing more
`rapidly than knowledge of the fundamental biology of FAP. This review highlights this imbal-
`ance, emphasising the need to better define the substrate repertoire and expression patterns of
`FAP to elucidate its role in biological and pathological processes.
`
`Keywords:
`Collagen / Degradomics / Dipeptidyl peptidase / Endopeptidase / Exosite / Protease
`
`Received: September 27, 2013
`Accepted: October 16, 2013
`
`1
`
`Introduction
`
`Proteolysis is an irreversible and essential post-translational
`modification that controls the composition and activity of
`proteins in a biological system. Bioactive peptides influence
`a range of physiological processes, including glucose homeo-
`stasis, the immune response and cellular signalling, and
`modification of these peptides can have significant impacts
`on such processes. Identifying proteases and how they modify
`
`Correspondence: Associate Professor Mark D. Gorrell, Molecu-
`lar Hepatology, Centenary Institute and Sydney Medical School,
`University of Sydney, Locked Bag No. 6, Newtown, NSW 2042,
`Australia
`E-mail: m.gorrell@centenary.usyd.edu.au
`Fax: +61-2-95656101
`
`␣2-AP,
`amy-
`APP,
`alpha-2-antiplasmin;
`Abbreviations:
`chimeric antigen receptor;
`loid precursor protein; CAR,
`CN,
`collagen; DPP, dipeptidyl peptidase; ECM, extracel-
`lular matrix; FAP, fibroblast activation protein; NPY, neu-
`ropeptide Y; PREP, prolyl endopeptidase; SRM, selected
`reaction monitoring; TAILS, terminal amine isotopic labelling of
`substrates; TG, thapsigargin
`
`C(cid:2) 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
`
`their substrates is imperative in understanding cell biology,
`physiology and pathogenesis and in determining the poten-
`tial of these proteases as drug targets.
`The dipeptidyl peptidase (DPP) 4 protein family has
`four enzymatic members; DPP4, fibroblast activation protein
`(FAP), DPP8 and DPP9 (Fig. 1), which are able to hydrolyse
`a prolyl bond that is two amino acids from the N-terminus
`of a protein. Due to the cyclic nature of the proline (Pro)
`residue, this is a rare catalytic ability that makes this group
`of proteases interesting in many aspects of biology and as
`therapeutic targets [1, 2].
`DPP4 is the most profoundly characterised member and
`prototype of this family; it is expressed ubiquitously and
`has a role in a wide range of physiological and pathological
`processes via its cleavage of bioactive peptides [1]. DPP4 was
`identified as a drug target for type 2 diabetes after it was
`observed to cleave and inactivate glucagon-like peptide-1 in
`vivo, thereby muting insulin secretion by the pancreas. This
`is just one example of how an understanding of the sub-
`strates of a protease can provide crucial information on its
`physiological role and thus lead to a new therapy.
`
`Colour Online: See the article online to view Figs. 1–4 in colour.
`
`www.clinical.proteomics-journal.com
`
`Petitioner GE Healthcare – Ex. 1024, p. 454
`
`

`

`Proteomics Clin. Appl. 2014, 8, 454–463
`
`455
`
`Figure 1. Enzymatic members of the DPP4 family. FAP and DPP4
`are type II integral membrane proteins that have intracellular and
`extracellular soluble, truncated forms. DPP8 and DPP9 are intra-
`cellular proteins.
`
`FAP is closely related to DPP4, sharing 52% amino acid
`sequence identity [3]. However, unlike DPP4, the substrate
`repertoire and the patterns of in vivo FAP expression are
`poorly defined. Despite this, FAP is attracting attention in
`cancer, cardiology and fibrosis research because its expres-
`sion is greatly upregulated in disease. A number of tools
`have thus been developed to target and exploit this protease
`for therapeutic intervention. The aim of this review is to
`emphasise the need to elucidate the roles and functions of
`this protease through substrate discovery in order to funda-
`mentally progress the field.
`
`2
`
`Fibroblast activation protein
`
`2.1 Enzyme activity and substrate specificity
`
`Unique from other members of the DPP4 family, FAP has
`both DPP and endopeptidase activities (Fig. 2). However,
`despite exhibiting both activities, FAP has been shown to be
`a more efficient endopeptidase than DPP [4, 5]. The activities
`and specificities of FAP have been investigated using artificial
`substrates and synthetic peptide libraries [4, 6–9], which re-
`vealed a strong preference for FAP cleavage of endopeptidase
`substrates after glycine-proline (Gly-Pro) motifs [4, 8, 9]. This
`restricted cleavage site specificity was supported by results
`obtained using a fluorescence resonance energy transfer sub-
`strate library, which also identified endopeptidase cleavage
`events only after Gly-Pro motifs [8]. However, a MS-based
`analysis of FAP endopeptidase specificity using gelatin as a
`substrate proposed that cleavage can also occur C-terminal
`to Ala-Pro, Arg-Pro, Lys-Pro and Ser-Pro motifs as well as
`the reported Gly-Pro motif [10]. The disparate results from
`these cleavage site specificity profiling studies highlight the
`
`C(cid:2) 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
`
`Figure 2. Dual-enzyme activity of FAP. (A) Dipeptidyl peptidase
`activity of FAP allows it to cleave two amino acids off the N-
`terminus of a protein. This cleavage occurs after a proline (Pro)
`residue. (B) Endopeptidase activity of FAP enables cleavage that
`is more than two amino acids from the N-terminus of a protein.
`Cleavage is restricted to the post-Pro bond after glycine-Pro (Gly-
`Pro) (adapted from [68] with permission).
`
`limitations of such biochemical approaches. Synthetic pep-
`tide libraries do not take into account native protein or peptide
`structure or the contribution of exosites in protease–substrate
`interactions. Thus, identification of novel physiological FAP
`substrates will assist in defining the cleavage site specificity
`of FAP and determining the potential contribution of exosite
`interactions of this protease during cleavage events.
`
`2.2 Substrates in vitro and in vivo
`
`Although FAP is involved in numerous (patho-) physiological
`processes, its substrate repertoire remains mostly unknown.
`However, screening known DPP4 substrates against a recom-
`binant human soluble form of FAP has identified a number of
`natural FAP substrates [11]. In that study, full-length neuro-
`logical peptides were cleaved by FAP and analysed by MALDI-
`TOF-MS. Four neuropeptides were efficiently cleaved by FAP,
`namely neuropeptide Y (NPY), peptide YY, B-type natriuretic
`peptide and substance P, with in vitro half-lives comparable to
`incubation with DPP4 [11]. Those results clearly show that the
`DPP specificity of FAP is much less restricted than previously
`thought, with cleavage occurring after Lys-Pro, Tyr-Pro, Arg-
`Pro and Ser-Pro dipeptides [11]. Interestingly, DPP cleavage
`was not observed for some other potential peptide substrates
`that were also screened and had the same N-terminal dipep-
`tide sequence as the four cleaved neuropeptide substrates.
`Furthermore, in a previous study using a synthetic dipeptide
`fluorescent substrate library, limited levels of cleavage by FAP
`
`www.clinical.proteomics-journal.com
`
`Petitioner GE Healthcare – Ex. 1024, p. 455
`
`

`

`456
`
`E. J. Hamson et al.
`
`Proteomics Clin. Appl. 2014, 8, 454–463
`
`were reported for these particular dipeptide sequences [6]. To-
`gether, these data strengthen the notion that exosite binding
`has a role in FAP cleavage events.
`As the endopeptidase activity of FAP distinguishes this
`protease from other members of the DPP4 family, it could
`be speculated that this may be its predominant enzymatic
`role as well as its unique role. Towards this, two physiological
`endopeptidase substrates have so far been discovered for FAP;
`denatured type I collagen (CN-I) [12, 13] and ␣2-antiplasmin
`(␣2-AP) [14,15]. In both cases, cleavage occurs after a Gly-Pro
`motif [10, 14].
`CN fibres are major components of the extracellular
`matrix (ECM), providing structural support for cells and tis-
`sues. ECM proteins also bind and sequester growth factors
`and bioactive peptides, thus regulating and influencing im-
`portant biochemical and biomechanical processes [16]. As
`CN is rich in Gly-Pro residues, it is unsurprising that FAP
`is able to digest this ECM protein. While FAP is unable to
`cleave CN-I in its native form, partial digestion, for example,
`by matrix metalloproteinase 1, results in unwinding of the
`CN fibre, and this facilitates FAP cleavage [12,17]. FAP is also
`able to cleave denatured CN-III in vitro [17]. CN deposition
`and degradation in ECM remodelling are common processes
`in tumourigenesis and fibrosis. The ability to recognise and
`digest denatured CN-I implicate FAP in ECM remodelling
`and the pathological processes where this is abnormal.
`As well as being a cell surface protease, FAP also exists
`in a truncated, soluble form in human plasma, lacking the
`transmembrane domain [15, 18]. Soluble FAP cleaves ␣2-AP
`at prolyl bonds Pro3-leucine4 and Pro12-Asn13 [18]. Dur-
`ing tissue repair, fibrin is deposited to form a fibrin clot.
`Fibrinolysis is the natural process in which a fibrin clot is
`dissolved by plasmin leading to scar resolution. ␣2-AP is an
`inhibitor of plasmin and therefore reduces the rate of lysis of
`the fibrin clot (Fig. 3). However, cleavage of ␣2-AP by FAP
`converts ␣2-AP into a more potent inhibitor of plasmin [18].
`We have therefore hypothesised that this activity of FAP leads
`to a reduction in fibrinolysis and promotion of scar formation
`in tissues [1].
`
`2.3 Expression pattern
`
`FAP is a difficult protein to study due to the limited avail-
`ability of specific or selective tools to detect and target it [19].
`However, recent developments have enabled more accurate
`and thorough characterisation of the expression pattern of
`this protease.
`Initially it was widely accepted that FAP was either not
`expressed or present at insignificant levels in normal adult
`tissue. However, under certain biological circumstances,
`FAP expression has been observed, such as during mouse
`embryogenesis [20] and in the resorbing tadpole tail [21].
`These observations suggest a role for FAP in normal devel-
`opmental processes and tissue remodelling. Despite the ob-
`served expression of FAP in healthy tissue and its presence
`
`C(cid:2) 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
`
`Figure 3. FAP modulates fibrinolysis. ␣2-Antiplasmin (␣2-AP)
`becomes a more potent inhibitor of plasmin following FAP-
`mediated cleavage between Pro12 and Asn13. This results in
`a shift from increased fibrinolysis and scar resolution in the
`absence of FAP (A), to decreased fibrinolysis and increased scar
`formation when FAP is present (B) (adapted from [1] with permis-
`sion).
`
`in plasma, the FAP knockout mouse has a healthy pheno-
`type [22, 23], suggesting that the role of this protease under
`normal circumstances is redundant or non-essential.
`FAP expression is, however, implicated in a number of
`pathological processes, resulting in the emergence of this
`protease as a potential therapeutic target. FAP expression
`has been observed during wound healing [24], at sites of
`inflammation including arthritis [25, 26] and in atheroscle-
`rotic plaques [27]. Greatly increased intrahepatic FAP
`expression occurs in cirrhosis and correlates with liver fi-
`brosis severity, where it is expressed by activated hepatic
`stellate cells [13, 28–30]. FAP is also strongly expressed by
`stromal fibroblasts in over 90% of epithelial carcinomas [31].
`However, despite efforts to study this protease in various can-
`cer models [32–35], little is known about its enzymatic role in
`the tumour microenvironment. It is often speculated that the
`collagenolytic role of FAP promotes invasion of tumour cells.
`Supporting this idea, FAP-positive fibroblasts isolated from
`human breast tumour stroma co-cultured with breast cancer
`cells lead to increased cancer cell migration and induction
`of epithelial–mesenchymal transition [36]. FAP-expressing
`cells have been shown to have an immunosuppressive role
`in mice in an immunogenic cancer model [37]. In that study,
`depletion of FAP-expressing cells resulted in rapid hypoxic
`necrosis of tumour and stromal cells. The enzymatic or extra-
`enzymatic mechanism by which FAP is involved in these
`processes requires further investigation.
`Several studies have linked FAP to promoting angiogen-
`esis in the tumour microenvironment [34, 38, 39]. One such
`study demonstrated that inhibition of both DPP4 and FAP led
`
`www.clinical.proteomics-journal.com
`
`Petitioner GE Healthcare – Ex. 1024, p. 456
`
`

`

`Proteomics Clin. Appl. 2014, 8, 454–463
`
`457
`
`to decreased tumour vascularisation, but such a decrease was
`not observed when DPP4 alone was inhibited [34], thus impli-
`cating the enzyme activity of FAP in promoting angiogenesis.
`FAP is able to cleave NPY to generate a truncated form [11]
`that is proangiogenic [40], which indicates that increased vas-
`cularisation may be mediated partly by FAP cleavage of NPY.
`The persistent expression of FAP in tumour stroma clearly
`implicates this protease in tumourigenesis [31]. However,
`there is controversy over this role for FAP, with some
`studies correlating elevated FAP expression with tumour
`suppression, while others show a correlation with tumour
`progression [33]. An important consideration when interpret-
`ing results from such studies is the source of FAP expression.
`Although it is well established that FAP is expressed by stro-
`mal cells [31], particularly at the invasive front [36, 41], many
`studies that aim to investigate this protease in the tumour
`microenvironment have focussed on cancer models using
`tumour cells that artificially express FAP [42], which is not
`an accurate reflection of the in vivo tumour microenviron-
`ment and results from such studies should be interpreted
`cautiously. In spite of this, the role of FAP in tumour stroma
`is still highly contextual and further investigation is required
`to elucidate its function in tumourigenesis.
`In addition to the strong focus on FAP as a tumour stroma
`marker, FAP is also considered to be a protease of interest in
`several other pathological conditions, including atherosclero-
`sis [27, 33], where FAP enzyme activity has been implicated
`via cleavage of CN-I, promoting plaque instability [27]. In
`this condition, FAP is expressed in human aortic smooth
`muscle cells and FAP expression correlates with progression
`of atherosclerotic plaques. Moreover, in this study, higher
`FAP levels were detected in thin fibrous caps compared to
`thick caps and treating human sclerotic plaques in vitro with
`FAP enzyme activity neutralising antibodies decreased FAP-
`associated CN degradation [27]. Thus, the collagenolytic activ-
`ity of FAP may be an important driver of fibrous cap rupture.
`In vivo studies are needed to further evaluate FAP as a can-
`didate therapeutic target for the treatment of patients with
`high-risk atherosclerotic plaques.
`
`2.4 FAP as a biomarker
`
`A major hurdle in the study of FAP as a potential disease
`biomarker is the inability to specifically measure FAP enzyme
`activity. However, an FAP-specific substrate has recently
`become available, enabling the measurement of FAP activity
`in fluids and organs from diseased and non-diseased samples
`[43]. There is a clear need to closely examine whether FAP
`enzyme activity levels associate with disease progression
`and pathogenesis to better understand roles of FAP, and to
`determine whether an FAP-specific enzyme assay could be
`an informative diagnostic tool in the clinical setting. For ex-
`ample, such an assay may be useful for selecting patients who
`may be more responsive to a FAP-targeted therapy.
`C(cid:2) 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
`
`However, regardless of whether FAP itself correlates with
`severity of disease or tumour progression, the ratio of full-
`length to cleaved FAP substrates may be found to change
`in relation to pathogenesis. If so, FAP substrates might
`become disease biomarkers. Consider the case of FAP-
`mediated cleavage of NPY, where the resulting cleaved form
`of NPY is proangiogenic [40]. In the normal healthy adult,
`FAP digestion of NPY may be negligible and potentially re-
`dundant as DPP4 is ubiquitous and cleaves NPY efficiently.
`However, in a patient with an aggressive epithelial tumour,
`the stromal fibroblasts surrounding the tumour can express
`high levels of cell surface FAP and thus the contribution
`by FAP to the N-terminal cleavage of NPY in that microen-
`vironment may supersede the contribution of DPP4. Thus,
`levels of cleaved versus full-length FAP substrates may have
`potential as biomarkers of disease progression. Proteomics-
`based assays utilising selected reaction monitoring (SRM)
`can be designed and optimised to detect and quantify lev-
`els of peptides in complex biological samples. SRM analysis
`has previously been applied to not only determine levels of
`cleaved and full-length caspase substrates in vivo, but also
`the rate of cleavage [44]. Such techniques may be applicable
`to FAP substrates.
`
`3
`
`Targeting FAP
`
`As the role of FAP seems to be highly contextual, it is
`important to consider the pathological situation in order to
`determine the best method to target FAP-expressing cells. A
`number of tools have therefore been designed to exploit the
`enzyme activity of FAP or to target this protein on cells.
`
`3.1 Inhibitors
`
`A major hurdle in the study of FAP enzyme activity has been
`the lack of selective inhibitors against this protease. FAP
`shares DPP specificity with the enzyme members of the DPP4
`family, DPP4, DPP8 and DPP9, as well as endopeptidase
`specificity with prolyl endopeptidase (PREP). Thus, design-
`ing inhibitors that are selective for FAP over other DPPs and
`PREP is challenging. Recently, however, two independent
`groups have designed potent FAP-selective inhibitors [45, 46]
`(Table 1).
`Poplawski et al. have recently developed potent FAP-
`selective and PREP-selective inhibitors using boronic acid-
`based compounds [46]. One compound, N-(pyridine-4-
`carbonyl)-D-Ala-boroPro, has a more than 350-fold selectively
`for FAP over PREP, and also has very strong selectivity over
`DPP4, DPP8 and DPP9. The development of such a potent
`and selective inhibitor will no doubt prove extremely useful
`in the investigation of the biological role of FAP.
`An independent study has developed a new class
`of FAP inhibitors based on an N-(4-quinolinoyl)-Gly-
`(2-cyanopyrolidine) scaffold [45]. Of
`the 34 compounds
`
`www.clinical.proteomics-journal.com
`
`Petitioner GE Healthcare – Ex. 1024, p. 457
`
`

`

`458
`
`E. J. Hamson et al.
`
`Proteomics Clin. Appl. 2014, 8, 454–463
`
`Table 1. Properties of potent FAP-selective inhibitors
`
`Compound
`
`IC50 (nM)
`
`DPP9
`3400 ± 800
`
`PREP
`13 000 ± 4300
`
`Ki (nM)
`
`FAP
`9 ± 0.9
`
`PREP
`3100 ± 260
`
`>100 000
`
`860 ± 70
`
`3 ± 0.4 N/A
`
`References
`
`[46]
`
`[45]
`
`DPP4
`
`>100 000
`
`DPP8
`5600 ± 1300
`
`FAP
`36 ± 4.8
`
`N-(Pyridine-4-
`carbonyl)-D-
`Ala-boroPro,
`(compound 6)
`10.3 ± 0.4 >100 000 N/A
`Compound 7
`Values are expressed as ±SEM. N/A, not available.
`
`developed in that study, compound 7 has particular selec-
`tivity towards FAP over the related proteases DPP4, DPP8
`and DPP9, and is also selective over PREP. Modifications
`to the quinolinoyl ring improved the selectivity for FAP
`over PREP [45], indicating that further developments of this
`scaffold can yield compounds with more selectivity for FAP.
`As combined inhibition of FAP and DPP4 enzyme
`activities attenuates tumour growth more than DPP4 inhi-
`bition alone [34, 47], a potent FAP-selective inhibitor may
`have potential for treating certain epithelial tumours and
`may be useful in a number of other clinical settings, such
`as atherosclerosis and arthritis [33]. The development of po-
`tent selective FAP inhibitors will be very useful in developing
`an understanding of the biological roles of FAP.
`
`3.2 Prodrugs
`
`Prodrugs consist of a cytotoxic agent coupled to a peptide that
`encodes the consensus sequence for cleavage by a particular
`protease. The cytotoxic agent becomes active upon release
`from the peptide, which only occurs in the presence of the
`specific protease (Fig. 4). Due to the restricted expression
`pattern and unique enzyme activity of FAP, FAP-activated
`prodrugs have potential as a therapeutic intervention in a
`number of pathological processes where FAP-expressing cells
`are implicated. Several FAP-activated prodrugs and protoxins
`have been designed.
`The honeybee toxin melittin has been modified to
`produce an FAP-activated protoxin that causes lysis of FAP-
`positive tumour stromal cells [48]. Intra-tumoural injection
`of this protoxin into breast and prostate cancer xenographs
`in mice resulted in significant cell lysis and inhibition of
`tumour growth. However, intravenous administration of this
`protoxin caused haemolysis and death, indicating that use
`of this protoxin should be limited to situations where intra-
`tumoural injection is possible. Despite its limitations, this
`study demonstrates that FAP-activated protoxins can signifi-
`cantly impair tumour growth.
`Doxorubicin has had limited clinical success as a
`chemotherapeutic agent due to cytotoxic side effects [49].
`An FAP-activated doxorubicin prodrug has been shown to
`minimise these off-target effects [9]. This prodrug produced
`C(cid:2) 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
`
`Figure 4. FAP-activated prodrugs. Prodrugs consist of a cytotoxic
`agent conjugated to a peptide sequence specific for a particular
`protease, in this case FAP. Exposure of the prodrug to FAP (A)
`results in cleavage of the peptide sequence thereby releasing
`and activating the cytotoxic agent (B), which acts on the cell.
`
`similar anti-tumour effects to doxorubicin alone, without the
`severe cytotoxic side effects in normal tissue. However, low
`levels of non-specific activation of the prodrug were observed
`in normal tissue and plasma, which was attributed to the
`instability of the peptide bond linking the dipeptide to doxo-
`rubicin. However, the dipeptide sequence used in that study
`is known to be cleaved by other oligopeptidases, including
`PREP [8], which is ubiquitous in normal tissues. Such issues
`of linker instability, peptide specificity and protease distribu-
`tion are commonly encountered in the development and use
`of protease-activated prodrugs.
`Another FAP-activated prodrug contains the plant-derived
`toxin thapsigargin (TG) [50], which inhibits calcium pumps,
`leading to an intracellular build up of calcium, resulting in
`
`www.clinical.proteomics-journal.com
`
`Petitioner GE Healthcare – Ex. 1024, p. 458
`
`

`

`Proteomics Clin. Appl. 2014, 8, 454–463
`
`459
`
`cell death. The FAP-activated TG prodrug caused significant
`apoptosis of stromal cells, compared to untreated controls
`and mice treated with TG prodrugs conjugated to a scrambled
`peptide sequence. Although cell death was not observed in the
`cancer cells themselves, the stromal cell death was significant
`enough to inhibit growth of both human breast and human
`prostate cancer xenografts in that study [50].
`
`3.3 T-cell immunotherapy
`
`Based on its consistent expression in tumour stroma, FAP
`has become the focus of two recent studies utilising T-cell
`immunotherapy [51, 52]. In both of these studies, chimeric
`antigen receptors (CARs) reactive against FAP were geneti-
`cally engineered and expressed by T cells.
`Tran et al.
`looked at the effect of FAP-CAR T cells
`in mouse models of cancer as well as human pancreatic
`cancer xenografts [52]. Despite the in vitro efficacy of this
`FAP-CAR T-cell against recombinant FAP as well as HEK293
`cells stably expressing FAP, a limited anti-tumour effect was
`observed in mouse cancer models. Unexpected off-target ef-
`fects in bone marrow and skeletal muscle, including cachexia,
`were also observed. To further investigate this result, non-
`tumour-bearing mice were treated with FAP-CAR T cells but
`bone toxicities and cachexia were still observed. That study
`identified FAP-positive mesenchymal stromal cells in bone,
`adipose and muscle tissues in both non-tumour-bearing and
`tumour-bearing mice. Interestingly, similar side effects have
`been observed in a separate study in which FAP gene tran-
`scribing cells were targeted and killed with diphtheria toxin
`[53].
`Kakarla et al. studied the effects of FAP-CAR T cells in a
`tumour initiation model as well as an established lung cancer
`xenograft model [51]. In both of these models, a significant
`decrease in FAP-positive stromal cells was observed as well
`as suppression of tumour growth. Although this anti-tumour
`effect was short lived, combined administration of FAP-CAR
`T cells and cancer cell targeted T cells led to a significant
`and prolonged inhibition of tumour growth. Furthermore,
`no off-target effects, including bone marrow abnormalities,
`were observed.
`immunotherapy relies on T-cell receptor
`CAR T-cell
`specificity and avidity, so these disparate results may be ex-
`plained following close examination of the CARs that were
`used. In addition, cytotoxic T cells are associated with cytokine
`release that may damage non-target cells. Parallel studies in
`FAP gene knockout mice would greatly assist in investigating
`the specificity of these FAP-CAR T cells and their effects in
`vivo.
`
`3.4 Considerations when targeting FAP
`
`Clearly there are significant differences between targeting
`FAP gene transcribing cells versus targeting FAP protein
`C(cid:2) 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
`
`or enzyme activity. Prodrugs exploit FAP enzyme activity to
`target and kill FAP-positive cells, whereas T-cell immunother-
`apy approaches target antigenic FAP protein to achieve this.
`Inhibitors, on the other hand, only block enzyme activity,
`leaving the cell intact. These are important considerations
`when determining which approach to use when targeting
`FAP, and also when interpreting the results from applying
`such techniques.
`The non-concordant data obtained when targeting FAP-
`positive cells also raise questions about whether
`the
`expression patterns of FAP mRNA, protein and enzyme ac-
`tivity differ in normal adult tissues. The data might also be
`influenced by a natural FAP inhibitor, if it exists. Such an
`inhibitor may limit prodrug activation or compete with FAP
`inhibitory compounds, but investigation on this topic is
`needed.
`A major limitation of antibody-based therapies is the
`limited availability of FAP-specific antibodies, particularly
`antibodies that cross-react with FAP derived from different
`species. This issue has been discussed thoroughly [19], and
`has hindered the implementation of antibody-based thera-
`pies in the clinical setting [54, 55]. This issue also highlights
`the advantages of targeting FAP enzyme activity with pro-
`drugs and inhibitors, as the enzyme activity of this protease
`is conserved across species.
`There clearly are large gaps in knowledge of the funda-
`mental biology of this protease. The expression pattern and
`substrate repertoire of this protease require greater defini-
`tion in order to elucidate the physiological roles of FAP to
`better target this protease in pathogenic settings. Improved
`knowledge of the basic biology of FAP will better predict
`potential effects of targeting this protease and also pre-
`dict potential impacts on the proteome when FAP enzyme
`activity is inhibited or ablated.
`
`4
`
`Proteomic and degradomic applications
`
`To date, most efforts towards understanding the biologi-
`cal functions of FAP have focussed on defining its cleav-
`age site specificity [4, 7, 10]. These results are invaluable and
`have no doubt enabled the design and development of FAP-
`selective inhibitors and FAP-activated prodrugs. However,
`research into the degradome of this protease now needs to be
`undertaken.
`Very few physiological substrates of FAP have been
`identified. Recently, new natural substrates were identified
`by screening known DPP4 substrates against FAP [11]. The
`four neuropeptides identified require in vivo investigation to
`determine the contribution of FAP to their physiological pro-
`cessing. This limited knowledge of the substrate repertoire
`of this protease emphasises the need to move away from
`traditional biochemical approaches and towards more un-
`biased, global approaches to identify novel physiological
`substrates of FAP.
`
`www.clinical.proteomics-journal.com
`
`Petitioner GE Healthcare – Ex. 1024, p. 459
`
`

`

`460
`
`E. J. Hamson et al.
`
`Proteomics Clin. Appl. 2014, 8, 454–463
`
`A number of degradomic techniques have been devel-
`oped for natural substrate discovery [56, 57]. Terminal amine
`isotopic labelling of substrates (TAILS) is a proven degrad-
`omic technique that identifies N-terminal cleavage products
`generated by a protease of interest [58]. The advantage of such
`global degradomic approaches, compared to traditional bio-
`chemical methods for substrate discovery, is that substrates
`exist in their native forms and locations [56]. The physio-
`logical setting of whether the substrate and protease actu-
`ally come into contact is thereby taken into account. TAILS
`has been applied to a number of intracellular and extracellu-
`lar proteases to successfully identify their substrates [59–62],
`including other members of the DPP4 family [63].
`
`4.1 Proteomic and degradomic analyses of related
`proteases
`
`A number of physiological DPP4, DPP8 and DPP9 substrates
`have been identified using proteomic- and degradomic-based
`techniques.
`Tinoco et al. utilised a label-free LC-MS-based peptidomics
`platform to investigate the contribution of DPP4 enzyme
`activity to peptide metabolism in the kidney [64]. Global
`comparison of peptide levels in kidneys isolated from wild
`type and DPP4 gene knockout mice identified 70 renal sub-
`strates of DPP4. These substrates were validated in vitro using
`recombinant DPP4. It was also shown that aminopeptidases
`work in concert with DPP4 to catabolise Pro-containing
`peptides in the kidney. This method proved highly success-
`ful in implicating DPP4 in catabolic pathways in the kidney.
`However, a major contributing factor to this success was the
`ability to optimise the technique based on detection of known
`DPP4 substrates [65].
`Unlike DPP4, the substrate repertoires of DPP8, DPP9
`and FAP are poorly defined and therefore require more global,
`unbiased degradomic techniques.
`Wilson et al. applied TAILS to identify 14 potential natural
`substrates of DPP8 and DPP9 in cytoplasmic extracts from
`SKOV3 cells overexpressing catalytically active or inactive
`human DPP8 or DPP9 [63]. Of these 14 natural substrates,
`nine were validated using peptide fragments in vitro. Further
`investigations will include whether naturally occurring forms
`of the putative substrates are hydrolysed.
`
`4.2 Validation of substrates
`
`A major hurdle in the application of degradomic techniques
`is refining the methods to validate and explore the biological
`significance of candidate substrates both in vitro and in vivo.
`There are multiple levels of validation, which all contribute
`to assigning a potential molecule as a physiological substrate
`of a protease.
`Positional degradomic approaches, such as TAILS, are
`advantageous as they enable both identification of the sub-
`C(cid:2) 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
`
`strate as well as the site of cleavage. Thus, a lot of information
`can be drawn from such experiments, including cleavage site
`specificity. This provides an extra level of confidence if the
`cleavage site specificity of a protease is known and can there-
`fore be matched to the observed cleavage site in the candidate
`substrate. In the case of FAP, cleavage occurs predominantly
`after a Pro residue, usually preceded by a Gly residue. TAILS
`data can therefore be manually parsed to identify cleavage
`sites occurring after Gly-Pro motifs.
`Biochemical validation of candidate substrates can be
`performed through a variety of methods usually involving
`incubation