expression of hif-1α by human macrophages: implications for the use of macrophages in...

Original Paper

Expression of HIF-1a by human macrophages:implications for the use of macrophages inhypoxia-regulated cancer gene therapy

Bernard Burke, Ngai Tang, Kevin P. Corke, Dean Tazzyman, Kurosh Ameri, Michael Wells and Claire E. Lewis*Tumour Targeting Group, Academic Unit of Pathology, Section of Oncology and Pathology, Division of Genomic Medicine, University of Sheffield MedicalSchool, Beech Hill Road, Sheffield S10 2RX, UK

*Correspondence to:Professor C. E. Lewis, TumourTargeting Group, Academic Unitof Pathology, Section of Oncologyand Pathology, Division ofGenomic Medicine, University ofSheffield Medical School, BeechHill Road, Sheffield S10 2RX, UK.E-mail: [email protected]

Received: 7 June 2001

Accepted: 20 September 2001

Published online:

11 December 2001

Abstract

Large numbers of monocytes extravasate from the blood into human tumours, where they

differentiate into macrophages. In both breast and prostate carcinomas, these cells accumulate in

areas of low oxygen tension (hypoxia), where they respond to hypoxia with the up-regulation of

one or more hypoxia-inducible factors (HIFs). These then accumulate in the nucleus and bind to

short DNA sequences called hypoxia-response elements (HREs) near or in such oxygen-sensitive

genes as that encoding the pro-angiogenic factor vascular endothelial growth factor (VEGF). This

stimulates gene expression and could explain why, in part, macrophages express abundant VEGF

only in avascular, hypoxic areas of breast carcinomas. It also suggests that macrophages could be

used to deliver HRE-regulated therapeutic genes specifically to hypoxic tumour areas. A recent

study suggested that hypoxic macrophages accumulate HIF-2 rather than HIF-1, prompting the

search for HRE constructs that optimally bind HIF-2 for use in macrophage-based gene therapy

protocols. However, the present study shows that human macrophages accumulate higher levels of

HIF-1 than HIF-2 when exposed to tumour-specific levels of hypoxia in vitro; that macrophages

in human tumours express abundant HIF-1; and that expression from HRE-driven reporter

constructs in the human macrophage-like cell line MonoMac 6 correlates more closely with HIF-1

than with HIF-2 up-regulation under hypoxia. Taken together, these findings suggest that

HIF-1 may be the major hypoxia-inducible transcription factor in macrophages and that HIF-1-

regulated constructs are likely to be effective in macrophage delivery of hypoxia-regulated gene

therapy to human tumours. Copyright # 2001 John Wiley & Sons, Ltd.

Keywords: HIF-1; HIF-2; hypoxia; macrophage; gene therapy; HRE; tumour targeting

Introduction

New blood vessels are usually disorganized in humantumours, often appearing incomplete and lacking instructural integrity. They are prone to collapse andoften become blocked by circulating leukocytes. Thisresults in the formation of areas of inadequateperfusion and transient hypoxia (low oxygen tension)in which the pO2 drops to 0–15 mmHg (<2% O2) [1].Rapid tumour cell proliferation in some areas mayoutstrip the rate of new blood vessel growth and thismay also cause hypoxic areas to form [2].

Tumour-associated macrophages (TAMs) are wide-spread in human breast carcinomas [3,4] and play animportant role in promoting tumour angiogenesis [4,5].TAMs accumulate in avascular areas of breast [4]and prostate [6] carcinomas, possibly as a result offactors released by hypoxic tumour cells in these areas[7]. TAMs then respond to the hypoxia present byexpressing pro-angiogenic cytokines such as vascularendothelial growth factor (VEGF) [8]. By contrast,TAMs in highly vascularized, well-oxygenated areas ofthe same tumours fail to express this cytokine [8].Moreover, macrophages exposed to hypoxia in vitroup-regulate their release of VEGF [9]. Thus, once

macrophages reach hypoxic regions in a given tumour,they promote tumour progression by releasing factorswhich stimulate tumour angiogenesis [10].

Only a scanty picture of the molecular mechanismsby which macrophages sense hypoxia and then switchon gene expression has emerged to date. In tumour celllines, hypoxia-regulated gene expression has beenshown to involve the stabilization, nuclear accumula-tion, and DNA binding of the transcription factorshypoxia inducible factor-1 (HIF-1) [11] and -2 (HIF-2,often referred to as EPAS-1) [12]. These proteins areheterodimers consisting of two different, hypoxia-inducible alpha subunits (HIF-1a and HIF-2a) and acommon, constitutive beta subunit. These HIFs thenaccumulate in the nucleus and bind to short DNAsequences called hypoxia-response elements (HREs)near or in oxygen-sensitive genes, stimulating geneexpression.

The relative contribution of HIF-1 and HIF-2 to theregulation of gene expression in hypoxic macrophagesis currently under debate. Two previous studies haveindicated that the main HIF up-regulated in hypoxicmacrophages is HIF-2 rather than HIF-1 [13,14].While human monocyte-derived macrophages (MDMs)express mRNA for the alpha subunits of both HIF-1 and

Journal of PathologyJ Pathol 2002; 196: 204–212.DOI: 10.1002 /path.1029

Copyright # 2001 John Wiley & Sons, Ltd.

HIF-2 in hypoxia, only up-regulation of HIF-2a pro-tein was detectable in these cells [13]. This findingwas supported by a recent report by Talks et al. [14],demonstrating detectable levels of HIF-2a, but notHIF-1a, in the human pro-monocytic cell line U937following hypoxic induction in vitro and in TAMsin various types of human tumour. However, immuno-blotting studies have shown that the murine alveolarmacrophage-like cell line MH-S up-regulates HIF-1aprotein under hypoxia [15] and immunoreactive HIF-1ahas been detected in human macrophages in the hypoxicsynovia of arthritic human joints [16].

Several studies have also addressed the question ofwhether HIFs expressed by macrophages are biologi-cally active. This has been assessed using reportergenes under the control of a minimal promoter down-stream of one or more HREs from the oxygen-sensitivegene for the enzyme murine phosphoglycerate kinase 1(PGK 1). Moreover, HRE-regulated reporter geneexpression can be induced by hypoxia in the murinemacrophage-like cell line ANA-1 [17] and the humanpro-monocytic cell line U937 [18]. When primaryhuman macrophages were infected with an adenoviruscontaining the gene for the pro-drug activating enzymecytochrome P450 under the control of a HRE and thenco-cultured with tumour spheroids (small three-dimensional tumour cell masses grown in vitro), theymigrated into the hypoxic centre of these structuresand expressed the enzyme. This produced markedtumour cell killing when spheroids were exposed tothe pro-drug, cyclophosphamide, as this was convertedto a cytotoxic metabolite [13]. This study demonstratesthat macrophages could have utility as cellular vehiclesto deliver hypoxia-regulated gene therapy to solidtumours.

In order to obtain maximal therapeutic expressionfrom transfected macrophages in hypoxic areas ofsuch tissues, it is necessary to optimize the type ofHRE used for the specific hypoxia-inducible transcrip-tion factors expressed by these cells. Here, we havecharacterized the expression of both HIF-1a and HIF-2a by human MDMs at various oxygen tensionsin vitro. We then investigated whether macrophages inhuman tumours express HIF-1a. Finally, we tested thehypoxia inducibility of a reporter gene (luciferase)

placed under the control of either a concatamerizedmurine PGK 1 HRE trimer, or a slightly amendedversion of this, called the ‘OB-HRE’ trimer [19] (seeTable 1) in the human macrophage-like cell line MonoMac 6 (MM6 [20]) in hypoxia. To elucidate the role ofHIF-1 and HIF-2 in the regulation of these constructs,the pattern of reporter gene inducibility was comparedwith that of HIF-1a or HIF-2a protein levels in MM6cells at the same oxygen tensions.

Materials and methods

Cells and tissues

All culture media and supplements used were sup-plied by Gibco BRL, UK, unless otherwise specified.Peripheral blood mononuclear cells were isolated fromplatelet-depleted buffy coats kindly provided by theBlood Transfusion Service, Sheffield, using Ficoll-Paque Plus (Pharmacia) according to the manu-facturer’s instructions. Mononuclear cells were culturedin Iscoves modified Dulbecco’s medium supplementedwith 2.5% human AB serum (Sigma), penicillin(100 IU/ml), streptomycin (100 mg/ml), and Fungizone(1.25 mg/ml). Flasks (75 cm2) were seeded with between4r107 and 1r108 cells, and glass coverslips (1.5 cmdiameter) in 12-well plates were seeded with 4r106

cells. After 2 h, the medium containing non-adherentcells was removed and replaced with fresh medium.Adherent cells (monocytes) were used in experimentsafter a further 6 days in culture, during which timemonocytes differentiated into MDMs. Immunohisto-chemical staining of these cells using a monoclonalantibody to the pan macrophage marker CD68 (seemethod below) indicated that more than 95% of thesecells were macrophages.

MM6 cells were a gift from Dr H. W. Ziegler-Heitbrock, Munich, Germany, and were cultured aspreviously described [20]. Fifteen formalin-fixed, para-ffin wax-embedded tumours were obtained from theHistopathology archives of the Royal HallamshireHospital, Sheffield, UK. These included five ductalinvasive carcinomas of the breast, five serous carcino-mas of the ovary, and five late-stage prostate carcino-mas which were randomly selected from the archive.

Table 1. DNA sequences of HRE trimers used in LUC induction studies in MM6 cells: OB-HRE [19], a trimer con-sisting of the 18 bp HRE sequence derived from the 5k flanking region of the murine PGK 1 gene (sequenceco-ordinates x307/x290) in natural orientation (separated by 6 bp spacer sequences, highlighted in bold); PGK 1, atrimer consisting of three contiguous copies (i.e. a concatamer) of the same 18 bp HRE sequence in the reverseorientation. The restriction sites used to clone the HREs into the minimal SV40 promoter of the pGL3 Promoterluciferase reporter construct, 5k NheI, 3k NheI/XbaI for OB-HRE, and 5k KpnI, 3k SacI for PGK 1, are underlined. TheEcoRI site added for screening purposes at the 3k end of the PGK 1 trimer is italicized. The sequences shown repre-sent the plus-sense (coding) strand in relation to the downstream LUC gene

HRE

construct Sequence of HREs (as inserted into pGL3 promoter)

OB-HRE GCTAGCGTCGTGCAGGACGTGACATCTAGTGTCGTGCAGGACGTGACATCTAGTGTCGTGCAGGACGTGACATCTAGC

PGK 1 GGTACCTGTCACGTCCTGCACGACTGTCACGTCCTGCACGACTGTCACGTCCTGCACGACGAATTCGAGCTC

HIF-1 expression by hypoxic macrophages 205

Copyright # 2001 John Wiley & Sons, Ltd. J Pathol 2002; 196: 204–212.

Hypoxic induction studies

Cells were incubated for 16 h at 37uC in humidifiedHeto multigas incubators containing either air (20.9%oxygen; normoxia), or air sufficient to give 1%, 0.5% or0.1% oxygen (i.e. various levels of hypoxia). Five percent CO2 was used in all normoxic and hypoxicincubators, with the balance being nitrogen in hypoxicincubations. Oxygen levels were confirmed during allexperiments, using Analox oxygen meters placed inincubators. It should be noted that due to the poordiffusibility of oxygen through liquids, the oxygentensions experienced by cells at the base of culturewells are not necessarily identical to those in theambient gas in the incubator. To minimize thedifference between the two, medium depths of lessthan 2 mm were used throughout this study.

Antibodies

The following antibodies (and dilutions) were used toimmunolocalize HIF-1a in cells/sections and demon-strate HIF-1a and HIF-2a protein in cell extracts byimmunoblotting (see section below). Monoclonal anti-human HIF-1a (clone 54, Transduction Labs. Inc.) wasused at 1 : 3000 for immunoblotting and 1 : 100 forimmunostaining. A second HIF-1a monoclonal (cloneH1a67, purchased from Novus Biologicals Inc.,Littleton, CO, USA) was used at a dilution of 1 : 2000for immunoblots and 1 : 100 for immunocytochemistry.The two monoclonal HIF-1a antibodies were raised topeptides derived from two different regions of theprotein: amino acids 610–727 for the antibody suppliedby Transduction Laboratories (TL) and 432–528for the antibody from Novus Biologicals Inc. (NB).Neither shows any homology with any region of HIF-2a. The longest contiguous stretch of amino acidhomology between HIF-1a and HIF-2a in the regionsused to make these antibodies is two amino acids forthe TL antibody and four for the NB one. Thus,neither antibody could cross-react with HIF-2a.

The HIF-1a immunostaining technique was vali-dated in two ways: (i) by the use of a positive controlcell preparation of HT1080 cells (a human fibrosar-coma cell line known to produce abundant HIF-1a inhypoxia); these cells were exposed to normoxia (20.9%O2) or hypoxia (0.5% O2) for 16 h and then fixed,processed into paraffin wax, sectioned, and stained forHIF-1a (as below); and (ii) using a second mono-clonal anti-HIF-1a (Novus Biologicals Inc., USA) asdescribed previously [21]. As a negative control, theHIF-1a antibody was replaced in the immunostainingprocedure by an inappropriate monoclonal of thesame IgG isotype (IgG1) and final working concen-tration (2.5 mg/ml) as the HIF-1a antibody.

A rabbit polyclonal antibody raised to HIF-2a (‘NB100–122’; Novus Biologicals Inc., Littleton, CO, USA)was used at 1 : 1000 in immunoblotting studies. Theimmunogen used to raise this antiserum was a peptideequivalent to amino acids 633–647 of human HIF-2a.This sequence bears no homology with any region of

HIF-1a, so this antibody does not detect HIF-1a. Thespecificity of the HIF-2a immunoblotting was con-firmed using a second, monoclonal HIF-2a antibody(190b, a gift from Professor A. Harris, Institute forMolecular Medicine, Oxford, UK) used at 1 : 3000.This was raised to amino acids 535–631 of HIF-2a andhas been shown previously not to cross-react withHIF-1a [14].

A monoclonal anti-human CD68 antibody (KP-1,Dako Ltd., UK) was used at 1 : 50 to identify macro-phages grown in culture vessels or present in tumoursections, as described by us in a previous publication [4].

Immunoblots

Total cell extracts were prepared by heating cells to95uC for 5 min in 2% SDS, 100 mM DTT, 60 mM

Tris–HCl (pH 6.8); passage through a 25-gauge needleten times; and centrifugation at 12 000 g for 10 min atroom temperature to remove cell debris. Proteins werequantified using the Bradford assay (Bio-Rad). SDSPAGE gels were run and electro-blotted onto PVDFmembranes using standard methods. Biotinylated mole-cular weight markers (Bio-Rad) were used to enableaccurate estimation of band molecular weights. Themembranes were blocked for 16 h at 4uC with 5%skimmed milk powder, 0.5% Tween 20 in PBS, andincubated in antibody diluted in 5% skimmed milkpowder in PBS. The membranes were then exposed toa mixture of the appropriate HRP-conjugated second-ary antibodies (Dako Ltd., UK) at a dilution of1 : 2000, and a streptavidin–HRP conjugate to enabledetection of the biotinylated markers (Bio-Rad, diluted1 : 3000), in Tris-buffered saline (20 mM Tris base,500 mM NaCl, pH 7.5) containing 0.05% Tween 20.Antibody-bound, blotted proteins and biotinylatedmarkers bound to the streptavidin–HRP conjugatewere visualized using enhanced chemiluminescencereagents (Nycomed-Amersham).

Immunolocalization of HIF-1a and CD68

Cells grown on coverslips and exposed to normoxia(20.9%O2) or hypoxia (0.1%O2) were removed from theincubator and fixed immediately in acetone or 10%neutral buffered formalin (depending on which pri-mary antibody was to be used) for 10 min and storedin PBS at 4uC until required. MM6 or HT1080 cellsfrom hypoxic induction experiments were fixed imme-diately in 10% formalin in PBS for 30 min. Cells werethen pelleted by centrifugation at 350 g for 5 min,mixed with 100 ml of rabbit thromboplastin (11.6 mM;Sigma) and human plasma (Sigma, UK), left to clot atroom temperature for 20 min, and then processed toparaffin wax. Five-micrometre sections of MM6 cellpellets or paraffin wax-embedded human tumours weresectioned onto APES (3,3-aminopropyltriethoxysilane)-coated slides and dried overnight at 4uC. For thehuman tumour samples, two sequential 3 mm sectionswere cut from each block and stained for either CD68or HIF-1a, using the method described below.

206 B. Burke et al.


After deparaffinization and rehydration (whereappropriate), endogenous peroxidase was blockedusing 3% H2O2 in methanol for 20 min, followed bythe appropriate antigen retrieval method: (i) for HIF-1a: microwave treatment in 0.1 M tri-sodium citrate(pH 7.4) for 10 min; (ii) for CD68: 0.01% protease typeXXIV (Sigma, UK) in PBS at 37uC for 20 min. Afterwashing in PBS and incubation with normal horseserum (Vector Laboratories), slides were incubatedwith the appropriate primary antibody overnight at4uC. The reaction was then detected using a stand-ard three-stage immunoperoxidase technique (VectorLaboratories Elite ABC kit) using the red chromo-gen, AEC, or the brown chromogen, DAB. Sectionswere then dehydrated, cleared, and mounted usingaquamount (BDH).

Hypoxia response element (HRE)–luciferasereporter constructs

A concatamerized PGK 1 trimer was made by syn-thesizing an oligonucleotide containing three copies(in reverse orientation, as this has been shown toexhibit good hypoxic inducibility in human cells [22])of the 18 bp HRE from the mouse phosphoglyceratekinase 1 gene (PGK 1; see Table 1), plus additionalbases incorporating SacI and KpnI restriction sites forcloning and an EcoRI site for screening of clones.These oligonucleotides were annealed so as to giveKpnI and SacI compatible overhangs at the 5k and 3kends, respectively. These were then cloned into the 5kKpnI and 3k SacI sites of the pGL3 Promoter luciferase(LUC) reporter construct (Promega Corp., UK),upstream of the minimal SV40 promoter.

The OB-HRE construct was made as previouslydescribed [19]. Briefly, oligonucleotides containingthree repeats of the murine PGK1 HRE separated byspacer sequences were synthesized, including additionalbases at the termini incorporating NheI and XbaI sitesat the 5k and 3k ends, respectively. These oligos wereannealed and cloned into the NheI site in pGL3Promoter (see Table 1). All DNA used in transfectionswas prepared using Qiagen endotoxin-free maxi prepkits (Qiagen Ltd., Crawley).

Transient transfection of MM6 cells

MM6 cells were split a day before the transfection togive a density of 5r105 cells /ml. For each DNAconstruct to be transfected, 107 cells were harvested bycentrifugation at 350 g for 5 min at room temperatureand resuspended in 1 ml of cold serum-free andsupplement-free RPMI 1640 medium (Gibco BRL).Cell suspension, 0.5 ml, was pipetted into a disposable0.3 cm gap electroporation cuvette (Bio-Rad); the DNAto be transfected was added (20 mg in total); and thecells were electroporated using a GenePulser (Bio-Rad,Hemel Hempstead) at 300 V and 975 mFarads. Cellswere resuspended in 4 ml of PBS and layered onto 5 mlof Ficoll-Paque Plus (Amersham Pharmacia Biotech,St. Albans, UK), followed by centrifugation at 400 g

for 15 min at room temperature. The band of live cellsat the interface was removed, resuspended in 5 ml ofPBS, and recovered by centrifugation at 350 g for5 min at room temperature. Cells were resuspended incomplete RPMI 1640 culture medium, counted, andaliquoted into 24-well plates at the required density ina total volume of 250 ml per well. Triplicate wells wereset up for each oxygen tension used. Cells wereincubated in a 5% CO2–95% air incubator at 37uC for1 h before being moved to hypoxic incubators.

Following incubations for 16 h at the requiredoxygen tension, cells were transferred to 1.5 ml micro-centrifuge tubes and centrifuged at 2000 g for 30 s, andthe pellets resuspended in 100 ml of 1r passive lysisbuffer (Promega). Cells were incubated at room tem-perature for 30 min, freeze-thawed once at x20uC, andthen assayed for luciferase activity using the PromegaDual Luciferase kit.

The Renilla LUC-expressing plasmid pRL SV40(Promega) was included in each transfection, as aninternal control to normalize LUC expression; 0.5 mgof pRL SV40 was used per transfection, with 19.5 mgof the firefly LUC-expressing HRE reporter plasmids.The pGL3 Promoter plasmid, which expresses fireflyLUC under the control of a minimal, enhancerlessSV40 promoter, was used as a negative control.

Results

Immunoblots for detection of hypoxia-inducibletranscription factors in MDMs and MM6 cells

Both HIF-1a and HIF-2a proteins were increased inMDMs exposed to hypoxia in vitro (Figure 1). Neg-ligible levels of HIF-1a were present following exposureto normoxia (20.9% O2). A slight hypoxia-inducibleband (molecular weight 120 kD) appeared at 1% O2,but was more marked at 0.5% and 0.1% O2. In MM6cell extracts, HIF-1a protein was up-regulated at 1%and 0.5% O2, but maximal at 0.1% O2 (Figure 2A).

Two bands of similar molecular weight appearedfor HIF-2a in MDM extracts. The upper band wasconstitutively expressed at all oxygen tensions, while alower, hypoxia-inducible band (which was absent at20.9% and 1% O2) appeared at 0.5% and 0.1% O2. InMM6 extracts, the lower band was sometimes present

Figure 1. Hypoxic induction of HIF-1a and HIF-2a in humanmonocyte-derived macrophages. A representative immunoblot(of three separate blots) showing bands (at approximately120 kD) in total cell extracts following exposure to normoxia(20.9% O2) or hypoxia (1%, 0.5% or 0.1% O2) for 16 h in vitro.C=constitutive band detected by the HIF-2a antibody



under normoxic conditions, albeit to a lesser extentthan under hypoxic conditions. This was only slightlyup-regulated at 1%, 0.5%, and 0.1% O2 (Figure 2A).Similar band intensities and expression profiles wereobtained for both MDM and MM6 extracts using adifferent HIF-2a antibody, the monoclonal 190b, aspreviously described [14] (data not shown).

MM6 cell transfections and HRE-driven luciferaseexpression

Both HRE-driven LUC reporter constructs were up-regulated in MM6 cells exposed to hypoxia, but onlythe concatamerized PGK 1 trimer was inducible at

both 1% and 0.1% O2 (Figure 2B). Up-regulation ofboth HRE-driven constructs was maximal at 0.1% O2.The absolute levels of firefly LUC expression weresignificantly higher for the PGK 1 HRE-LUC con-struct than for the OB-HRE-LUC construct at 1% and0.1% O2, but because the PGK 1 HRE-LUC alsoresulted in higher LUC expression in normoxia, thefold induction values (hypoxic induction/normoxicinduction) were similar. For example, at 0.1% O2, thePGK 1 HRE gave 19.4-fold induction over normoxiclevels and the OB-HRE gave 17.8-fold induction.

HIF-1a immunostaining of macrophages in vitroand in human tumour sections

Immunoreactive HIF-1a protein was evident in boththe cytoplasm and the nuclei of the positive controlpreparation (HT1080 cells grown on coverslips) onlyafter the cells had been exposed to hypoxia, either at0.5% oxygen (Figure 3B) or at 0.1% oxygen (data notshown). No HIF-1a staining was seen in normoxicHT1080 cells (Figures 3A). A similar pattern of stain-ing was seen using a second monoclonal anti-HIF-1aantibody (not shown). Human MDMs also expressedimmunoreactive HIF-1a protein (mainly cytoplasmic,but occasionally also nuclear) following exposure to0.5% oxygen for 16 h (Figures 3C and 3D). Essentiallysimilar staining was seen for HIF-1a in hypoxic MM6cells (not shown).

Immunostaining of sequential sections of malignantovarian (Figures 3E and 3F), breast (Figures 3G and3H) and prostate (not shown) tumours showed thepresence of clusters of CD68-positive macrophagesexpressing HIF-1a in all tumours examined. Stainingof TAMs was mainly cytoplasmic, with occasionalnuclei also showing HIF-1a positivity. Comparison ofCD68 and HIF-1 stained breast and ovarian sequentialsections (using the method described by us in ref. 8)indicated that between 5% and 50% of CD68-positiveTAMs expressed detectable levels of HIF-1a, with thisvarying markedly between different areas of eachtumour and between different tumours. Expression ofHIF-1a by TAMs was closely associated with that byneighbouring tumour cells in the same areas oftumours. TAMs in or abutting areas of necrosis wereHIF-1a-positive. Tumours containing high numbers ofTAMs tended to exhibit a markedly lower proportionof HIF-1a-positive TAMs (<10%) than those with lownumbers of TAMs (20–50%).

A similar pattern of staining was seen with the twoHIF-1a antibodies. No immunostaining was evident incell preparations or on tissue sections when the HIF-1aantibody was replaced by an IgG isotype-matchedinappropriate antibody (not shown).

Discussion

The data presented here show for the first time thatprimary human MDMs accumulate HIF-1a whenexposed in vitro to the severe hypoxia present in

Figure 2. (A) Levels of HIF-1a and HIF-2a present in the humanmacrophage-like cell line MM6 at various oxygen tensions. Arepresentative immunoblot (of three separate blots) showingbands (at approximately 120 kD) in total cell extracts of MM6cells following exposure to normoxia (20.9% O2) or hypoxia(1%, 0.5%, or 0.1% O2) for 16 h. C=constitutive band detectedby the HIF-2a antibody. (B) Hypoxic induction of HRE-drivenconstructs in MM6 cells in vitro. Firefly LUC expression(normalized against Renilla LUC from the co-transfected pRLSV40 plasmid and expressed as a ratio of the two; relative lightunits, RLU) by MM6 cells following transient transfection withpGL3 Promoter-based LUC plasmids containing no enhancer(pGL3), a trimer of the HRE from the murine PGK 1 gene inreverse orientation (PGK 1), or a trimer of the HRE from themurine PGK 1 in natural orientation plus spacer sequences (OB-HRE). Transfected cells were exposed to normoxia (20.9%oxygen) or hypoxia (1% or 0.1% oxygen) for 16 h before beinglysed and assayed for LUC. Means (tSEM) of triplicate wells foreach condition are given. Essentially similar results wereobtained in three separate experiments

208 B. Burke et al.


Figure 3. Immunolocalization of HIF-1a in HT1080 cells (A, B) and human monocyte-derived macrophages (C, D) followingexposure to either normoxia (20.9% O2; A, C) or hypoxia (0.5% O2; B, D) for 16 h in vitro. Immunoreactive HIF-1a is seen in boththe cytoplasm and the nuclei (arrows) of both cell types following hypoxic induction. Expression of immunoreactive HIF-1a byCD68-positive macrophages in a human ovarian (E, F) and breast (G, H) carcinoma. Sequential 3 mm sections of wax-embeddedtumours were immunostained for the pan-macrophage marker CD68 (E, G) and for HIF-1a (F, H). HIF-1a is present in both thecytoplasm and the nuclei of macrophages in these tissues (scale bar=50 mm, relates to panels E–H)



many forms of human tumour (0.1–1% oxygen). Wethen extended these studies to show that (i) hypoxichuman MDMs in vitro express HIF-1a as detected byimmunostaining; (ii) TAMs also express this transcrip-tion factor in focal, presumably hypoxic, areas ofvarious types of human tumour (although other factorscould also influence HIF levels in vivo); and (iii) ahuman macrophage-like cell line also shows hypoxia-inducible expression of HIF-1a and the level of thisprotein correlates with HRE-driven reporter geneexpression in hypoxia.

In a previous study we detected HIF-1a mRNA inhypoxic MDMs, but not HIF-1a protein [13]. Theamended Immunoblotting protocol used in the immuno-blotting present study facilitated detection of HIF-1a bytwo different antibodies. Moreover, the two HIF-1aantibodies used were raised to immunogens derived fromcompletely separate parts of the protein, thus confirmingthe identity of the hypoxia-inducible protein detected.

These results contrast with those of Talks et al. [14],who showed that the human pro-monocytic cell lineU937 up-regulates HIF-2a (not HIF-1a) followingexposure to hypoxia in vitro. Also, their immunostain-ing studies detected only HIF-2a in TAMs in humantumours, whereas we have shown here that with theappropriate antigen retrieval methods and HIF-1aantibodies, HIF-1a is also readily detectable inTAMs, although the exact proportion of TAMsexpressing detectable HIF-1a varied considerablybetween tumours. The possibility that both HIF-1amonoclonals used in the present study may have cross-reacted with HIF-2a can be excluded by comparing theimmunogens used to make these antibodies and theamino acid sequence of HIF-2a [23]. These show thatthe longest contiguous stretch of full amino acidhomology between HIF-1a and HIF-2a in the regionsused to make these antibodies is two or four aminoacids respectively, so neither is likely to bind HIF-2a.

As HIF-1 is a transcription factor known toaccumulate in the nuclei of hypoxic cells, it wasinteresting to note that HIF-1a was detected predomi-nantly in the cytoplasm of hypoxic macrophages andin TAMs in the present study. This parallels the findingthat HIF-2a was also predominantly cytoplasmic inTAMs [14]. One possible explanation for this might bethat HIFs bind to other factors and/or undergoconformational change in the nuclei of macrophages,thereby reducing their immunoreactivity. Alternatively,ubiquitination, an important prerequisite for proteo-somal degradation of HIF-1a, may occur, causingthem to be rapidly shuttled out of the nucleus andto accumulate mainly in the cytoplasm.

When HIF-2a (EPAS-1) levels were monitored atdifferent oxygen tensions in human MDMs and MM6cells by immunoblotting, it was routinely found to beless inducible than HIF-1a. Low levels were oftendetected in MM6 cells under normoxic conditions, afinding that accords well with a previous studyshowing that HIF-2a was present at higher levels thanHIF-1a under normoxia in various cell lines [12].

Furthermore, at 0.1% and 0.5% oxygen (i.e. severehypoxia), both the hypoxia-induced increases and theultimate levels of HIF-2a were minor compared withthose for HIF-1a (using either set of HIF antibodies),although it should be noted that immunoblottingprovides only approximate estimates of protein levels.These findings are in keeping with a previous reportshowing that the levels of HIF-1a and HIF-2a mRNAexpressed under hypoxia in a particular cell type areoften markedly different. Certain cell lines express upto 21 times as much HIF-2a mRNA as HIF-1a mRNAunder hypoxia, while others have more HIF-1a thanHIF-2a mRNA [12].

Intriguingly, with both anti-HIF-2a antibodies usedhere (one polyclonal and the other monoclonal), weobserved a constitutive protein band, with a molecularweight of around 120 kD, just above the inducibleHIF-2a band in MDM extracts and in MM6 cells; andothers have reported it in the human pro-monocyticcell line U937 [14]. Since the immunogens used toproduce the two HIF-2a antibodies employed in thepresent study were from two different regions of theprotein, bearing no homology, our data point towardsthe possibility of a constitutive, slightly larger form ofHIF-2a (or at least a closely related protein whichshares at least two epitopes with HIF-2a). It remains tobe seen whether this protein is transcriptionally active,in which case it could be involved in maintainingexpression of some genes in both normoxic andhypoxic cells. Alternatively, it could be an inactiveform, requiring post-translational modifications tobecome functional. This could conceivably allow thecell to respond very rapidly to the onset of hypoxia,without the delay required to synthesize and accumu-late HIF subunits.

The data we report here, showing that HIF-1a ismore inducible and appears to be expressed to higherlevels than HIF-2a in hypoxic macrophages, suggestthat the former may play a dominant role in mediatingmany of the effects of hypoxia on macrophagefunction, such as increasing the expression of pro-angiogenic factors in human tumours [10]. To date, nogenes have been identified as being exclusively respon-sive to either HIF-1a or HIF-2a; but it is possible thatthese two related transcription factors serve distinctfunctions within the cell, perhaps having partiallyoverlapping groups of target genes, some of whichcould be up-regulated equally well by either HIF, andsome more strongly by one or the other. Interestingly,hypoxia-induced expression from the VEGF promoterin a HIF-1a/HIF-2a-negative tumour cell line wasstimulated to a greater extent when HIF-2a, ratherthan HIF-1a, was overexpressed in these cells [12].However, it remains to be seen whether HIF-2 is also astronger stimulus for VEGF in macrophages, especiallyas HIF-1 is present in abundance in these cells underhypoxia.

Establishing the ability of macrophages to up-regulate HIF-1a in hypoxia also has importantimplications for assessing the feasibility of using those

210 B. Burke et al.


cells to deliver hypoxically activated anti-tumour gene

therapy, for example anti-angiogenic factors, to poorly

vascularized areas of solid tumours [13]. This strategy

relies on the tendency of macrophages to accumulate in

these sites, either by being drawn in by chemoattrac-

tant factors released by hypoxic cells, or via the direct

inhibitory effect of hypoxia on macrophage migration

[24]. The latter might cause macrophages migrating

randomly within tumours simply to become immobi-

lized in hypoxic areas. Once there, macrophages would

produce HIFs capable of activating gene expression

from HRE-driven therapeutic constructs.With the aim of producing optimal HRE constructs

for possible macrophage-delivered therapeutic applica-

tions and of establishing the oxygen tensions required

for maximal HRE-mediated gene expression in macro-

phages, we tested the hypoxic inducibility of two

PGK 1 HRE trimer-LUC constructs. Both the con-

catamerised and the modified (OB-HRE [19]) forms

showed maximal inducibility in MM6 cells in 0.1%

oxygen. However, the levels of LUC produced by these

two constructs were markedly different at this low

oxygen tension, with the PGK 1 HRE trimer showing

a markedly higher output than the OB-HRE trimer.

This suggests that the former HRE configuration may

have greater utility in macrophage-based delivery of

HRE-regulated therapeutic genes in vivo. However, the

higher output of this HRE trimer in normoxia may be

a limiting factor, since it may trigger expression in

normoxic/mildly hypoxic sites (i.e. outside the tumour),

leading to undesirable side-effects of this form of gene

therapy.Both HRE-driven constructs exhibited maximal

inducibility when HIF-1a levels were at their peak in

MM6 cells (i.e. at 0.1% oxygen). By contrast, HIF-2alevels were similar at 0.1–1% oxygen. This, when taken

together with (i) our finding that HIF-2a, but not HIF-

1a, was often present under normoxic conditions when

reporter gene expression was negligible; and (ii) the

strong hypoxic induction of HIF-1a in hypoxic MM6

cells, which mirrors LUC induction, suggests that HIF-

1a may be mainly responsible for the observed hypoxic

induction of PGK 1 HRE-driven reporter gene expres-

sion in human macrophages.In sum, our data show that HIF-1a is markedly up-

regulated by hypoxia in human macrophages in vitro

and is expressed by these cells in human tumours. This

suggests that HIF-1 is likely to play an important part

in mediating the myriad effects of hypoxia on gene

expression in macrophages. They also show that gene

constructs placed under the control of HREs that are

HIF-1-inducible would have utility for macrophage-

based gene delivery to hypoxic sites in solid tumours.

Acknowledgements

We gratefully acknowledge the support of Yorkshire Cancer

Research for this project. We thank Professor Adrian Harris for

the gift of the monoclonal anti-HIF-2a 190b.

References

1. Vaupel P, Kallinowski F, Okunieff P. Blood flow, oxygen and

nutrient supply, and metabolic microenvironment of human

tumours: a review. Cancer Res 1989; 49: 6449–6465.

2. Brown JM, Giaccia AJ. The unique physiology of solid tumours:

opportunities (and problems) for cancer therapy. Cancer Res

1998; 58: 1408–1416.

3. Kelly PM, Davison RS, Bliss E, McGee JO. Macrophages in

human breast disease: a quantitative immunohistochmeical

study. Br J Cancer 1988; 57: 174–177.

4. Leek RD, Lewis CE, Whitehouse R, Greenall M, Clarke J,

Harris AL. Association of macrophage infiltration with angio-

genesis and prognosis in invasive breast carcinoma. Cancer Res

1996; 56: 4625–4629.

5. Vacca A, Ribatti D, Ruco L, et al. Angiogenesis extent and

macrophage density increase simultaneously with pathological

progression in B-cell non-Hodgkin’s lymphomas. Br J Cancer

1999; 79: 965–970.

6. Burton JL, Wells JM, Corke KP, Maitland N, Hamdy FC, Lewis

CE. Macrophages accumulate in avascular, hypoxic areas of

prostate tumours: implications for the targeted therapeutic gene

delivery to such sites. J Pathol 2000; 192: 8A.

7. Collingridge DR, Hill SA, Chaplin DJ. Proportion of infiltrating

IgG-binding immune cells predict for tumour hypoxia. Br

J Cancer 2001; 84: 626–630.

8. Lewis JS, Landers R, Underwood JCE, Harris AL, Lewis CE.

Expression of VEGF by macrophages is upregulated in avascular

areas of breast carcinomas. J Pathol 2000; 192: 150–158.

9. Harmey JH, Dimitriadis E, Kay E, Redmond HP, Bouchier-

Hayes D. Regulation of macrophage production of VEGF by

hypoxia and TGF b-1. Ann Surg Oncol 1998; 5: 271–278.

10. Lewis JS, Lee J, Underwood JCE, Harris AL, Lewis CE.

Macrophage responses to hypoxia: implications for disease

mechanisms. J Leuk Biol 1999; 66: 889–900.

11. Wenger PH. Mammalian oxygen sensing, signalling and gene

regulation. J Exp Biol 2000; 203: 1253–1263.

12. Wiesener MS, Turley H, Allen WE, et al. Induction of

endothelial PAS protein-1 by hypoxia: characterization and

comparison with hypoxia-inducible factor-1a. Blood 1998;

92: 2260–2268.

13. Griffiths L, Binley K, Iqball S, et al. The macrophage: a novel

system to deliver gene therapy to pathological hypoxia. Gene

Ther 2000; 7: 255–262.

14. Talks KL, Turley H, Gatter KC, et al. The expression and

distribution of the hypoxia-inducible factors HIF-1alpha and

HIF-2alpha in normal human tissues, cancers, and tumour-

associated macrophages. Am J Pathol 2000; 157: 411–421.

15. Yu AY, Frid MG, Shimoda LA, Wiener CM, Stenmark K,

Semenza GL. Temporal, spatial, and oxygen-regulated expres-

sion of hypoxia-inducible factor-1 in the lung. Am J Physiol

1998; 275: L818–L826.

16. Hollander AP, Corke KP, Freemont AJ, Lewis CE. Expression

of HIF-1a by macrophages in the rheumatoid synovium:

implications for targeting hypoxia-regulated gene therapy to the

inflamed joint. Arthrit Rheumatol 2001; 44: 1540–1544.

17. Melillo G, Musso T, Sica A, Taylor LS, Cox GW, Varesio L. A

hypoxia-responsive element mediates a novel pathway of activa-

tion of the inducible nitric oxide synthase promoter. J Exp Med

1995; 182: 1683–1693.

18. Maxwell PH, Pugh CW, Ratcliffe PJ. Inducible operation of the

erythropoietin 3k enhancer in multiple cell lines: evidence for a

widespread oxygen-sensing mechanism. Proc Natl Acad Sci

U S A 1993; 90: 2423–2427.

19. Boast K, Binley K, Iqball S, et al. Characterization of

physiologically regulated vectors for the treatment of ischemic

disease. Hum Gene Ther 1999; 10: 2197–2208.

20. Ziegler-Heitbrock HW, Thiel E, Futterer A, Herzog V, Wirtz A,

Riethmuller G. Establishment of a human cell line (Mono Mac

6) with characteristics of mature monocytes. Int J Cancer 1998;

41: 456–461.



21. Zhong H, De Marzo AM, Laughner E, et al. Over-expression of

hypoxia-inducible factor 1alpha in common human cancers and

their metastases. Cancer Res 1999; 59: 5830–5835.

22. Firth JD, Ebert BL, Pugh CW, Ratcliffe PJ. Oxygen-regulated

control elements in the phosphoglycerate kinase 1 and lactate

dehydrogenase A genes: similarities with the erythropoietin 3kenhancer. Proc Natl Acad Sci U S A 1994; 91: 6496–6500.

23. Tian H, McKnight SL, Russell DW. Endothelial PAS domain

protein 1 (EPAS1), a transcription factor selectively expressed in

endothelial cells. Genes Dev 1997; 11: 72–82.

24. Turner L, Scotton C, Negus R, Balkwill F. Hypoxia inhibits

macrophage migration. Eur J Immunol 1999; 29: 2280–2287.

212 B. Burke et al.


expression of hif-1α by human macrophages: implications for the use of macrophages in...

Documents