High Levels of “Complexed” Interleukin-6 in Human Blood*

The biochemical nature of endogenous interleukin-6 (IL-6) as it exists in human serum or plasma was in-vestigated. Serum from a patient following bone marrow (BM) transplantation and fresh samples (interleukin)-6 content using The in of

The biochemical nature of endogenous interleukin-6 (IL-6) as it exists in human serum or plasma was investigated. Serum from a patient following bone marrow (BM) transplantation and fresh plasma samples from patients with epidermolysis bullosa or psoriasis, as well as from normal volunteers, were fractionated through G-200 columns and each of the eluted fractions assayed for IL (interleukin)-6 content using enzymelinked immunosorbent assays (ELISAs) based on the monoclonal antibody (mAb) pairs IG61/5IL6 or 4IL6/ 5IL6 and in the B9 hybridoma growth factor bioassay. The IG61/5IL6 ELISA and the B9 assay detected IL-6 in BM serum almost exclusively of molecular mass -20 kDa. In contrast, the 4IL6/5IL6 ELISA detected strong IL-6 immunoreactivity in complexes of size 100-150 and 400-500 kDa. IL-6 present in the 100-150-and 400-500-kDa complexes was purified by immunoaffinity chromatography through a 5IL6 mAb column. The 5IL6 mAb immunoaffinity column eluate of the respective pools from BM serum contained IL-6 at concentrations approaching 1 pg/ml as characterized by Western blotting. Sufficient IL-6 and associated proteins were purified by 5IL6 mAb immunoaffinity column chromatography of the 100-150-kDa complex from 0.8 ml of BM serum to allow (i) verification of three of the polypeptides as IL-6 by amino-terminal sequencing (estimate of IL-6 in original serum sample: 5-10 pg/ml), (ii) identification by amino acid sequencing of the "associated" proteins as complement factor C3b (carboxyl-terminal of the a-chain), complement factor C4b (y-chain), C-reactive protein, and albumin, and (iii) detection of an "associated" polypeptide consistent with the soluble IL-6 receptor. Taken together, these data establish that IL-6 is present at unexpectedly high concentrations in human blood in novel biochemical complexes that include other plasma proteins, which in turn, can camouflage IL-6 immunoreactivity and bioactivity as measured in conventional assays.
The cytokine interleukin-6 (IL-6)' has been increasingly studied as a protein with diverse and profound effects on a wide array of different cell types (for reviews, see Refs. [1][2][3][4]. IL-6, primarily Escherichia coli-derived recombinant human IL-6 (rIL-6), when used in cell culture experiments and in animal models, has been observed to mediate many of the acute phase alterations in plasma protein synthesis, enhance the proliferation and differentiation of B-cells and B-cell lines, activate T-cell function, increase the motility of some breast carcinoma cell lines, enhance differentiation of monocytes, and stimulate or inhibit the proliferation of epithelial cells depending on the cell type examined (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11). The importance of IL-6 in man in uiuo has been inferred from its elevated levels in the bloodstream and/or in compartmentalized body fluids in situations that range from sunburn to a variety of different bacterial and viral infection, trauma, and neoplasia (1)(2)(3)(4)(12)(13)(14)(15)(16). The human IL-6 gene promoter is readily induced by a wide variety of "noxious" stimuli that include most of the inflammation-associated cytokines (17)(18)(19)(20)(21)(22). What is the biochemical nature of IL-6 as it actually exists in the human body? What are the biological effects of endogenous human IL-6 compared with those of recombinant IL-6 produced in E. coli-or other expression systems?
Natural human IL-6 prepared in cell culture exhibits differential post-translational modifications that, in turn, determine its higher order structure and biological properties (23). Human fibroblasts secrete at least six distinct IL-6 forms of molecular mass in the range 23-30 kDa that are resolvable in SDS-PAGE under completely denaturing and reducing conditions (24,25): at least three O-glycosylated species of mass 23-25 kDa and three Nand O-glycosylated species of mass 27-30 kDa. All of the fibroblast-derived IL-6 species are phosphorylated on serine residues (26). Particular human and New World monkey (tamarin) B-cell lines or T-cell lines can secrete IL-6 species that are unglycosylated (21-kDa) and/or exclusively N-glycosylated (25-kDa) or even exclusively Nand O-glycosylated (30-kDa) (27,28). Under partially denaturing conditions, IL-6 can appear as 43-45-kDa species by Western blotting (15,25,29). The amino terminus of the 23-25-kDa exclusively O-glycosylated fibroblast-derived IL-6 is at AlazR (IL-6 polypeptide of 185 residues), whereas that of the 27-30-kDa Nand O-glycosylated fibroblast-derived IL-6 is a t Val3' (IL-6 polypeptide of 183 residues) (27). The major amino terminus of the nonglycosylated 19-21-kDa IL-6 secreted by the human T-cell line TCL-Nal has been identified to be at Proz9 (28). 19698 Under nondenaturing conditions, pure fibroblast-derived IL-6 (prepared under serum-free conditions) behaves as a high molecular weight multimeric aggregate (23). When electrophoresed through nondenaturing Nonidet P-40-polyacrylamide gels, fibroblast-derived IL-6 appears to be largely a trimeric 85-kDa complex containing both the 23-25-kDa 0glycosylated and the 28-30-kDa Nand 0-glycosylated species. Sephadex G-200 gel filtration chromatography of fibroblast-derived IL-6 reveals that this cytokine elutes in a broad range of sizes from 20 to 70 kDa. The high molecular mass fractions (>60 kDa by gel filtration) are enriched in the 27-30-kDa monomeric species. Although all of the IL-6-containing fractions exhibit hepatocyte-stimulating factor activity as monitored by the stimulation of cul-antichymotrypsin secretion by Hep3B cells, the high molecular mass fractions (50-70-kDa) are approximately 10-fold less active than the lower molecular mass fractions  in the B9 hybridoma growth factor assay (23). A distinction between the high and low molecular mass fractions of fibroblast IL-6 eluted off a G-200 column was also observed in immunoassays.* Using various pairs of anti-IL-6 monoclonal antibodies (mAb) as probes for IL-6 structure (30),' we have observed that the 4IL6/5IL6 ELISA preferentially detected the high molecular mass IL-6, whereas the IG61/5IL6 ELISA preferentially detected the lower molecular mass IL-6. Although the latter ELISA was highly sensitive to rIL-6 (sensitivity, 1-5 pg/ml), it failed to adequately react with fibroblast-derived >6O-kDa IL-6 eluted off gel filtration columns. It is noteworthy that only the trapping mAb ("4IL6," as compared with "IG61") was different in these two ELISAs; the biotinylated reporter mAb ("5IL6") was the same. ' We have characterized IL-6 as it is present in human serum and in fresh plasma using a variety of biochemical, immunological, and bioassay procedures. We have made the unexpected discovery that the bulk of endogenous IL-6 present at high concentrations in human serum or plasma occurs in two high molecular mass complexes of size 100-150 and 400-500 kDa, as characterized by Sephadex G-200 gel filtration. IL-6 in these circulating complexes lacks the customary B9 growth factor activity and the epitopes for IL-6-antigen recognition by ELISAs that otherwise recognize E. coli-derived rIL-6 with great sensitivity. Unlike IL-6 multimers observed in pure fibroblast-derived IL-6 preparations, the serum/plasma-derived IL-6 complexes appear to contain other plasma proteins in "association" with IL-6. The new data point to novel biochemical mechanisms that may regulate the transport and bioavailability of IL-6 as it is present in the peripheral circulation.

EXPERIMENTAL PROCEDURES
Materials-Acrylamide used for all gels was purchased from National Diagnostics (Manville, NJ). The Vectastain ABC Elite kit purchased from Vector Laboratories (Burlingame, CA) was used for all Western blot analyses. The affinity resin Hydrazide Avidgel AX was purchased from BioProbe International, Inc. (Tustin, CA). Polyvinylidene difluoride (PVDF) and nitrocellulose membranes were purchased from Schleicher and Schuell (Keene, NH). All other reagent grade chemicals were purchased from Sigma.
B9 Growth Factor Assay for IL-6-The IL-6 content of all chromatography eluates and the various IL-6 preparations were assayed by monitoring their ability to induce the proliferation of murine B9 hybridoma cells using standard procedures (31,32).
Sephadex through a Sephadex G-200 column (2.5 X 60 cm; V,, = 125 ml; V. = 325 ml). Eluate fractions of volume 2.5 ml each were collected after the first V, and continued for an additional two V0s (approximately different ELISAs (4IL6/5IL6 and IG61/5IL6) for IL-6, as well as in 100 fractions). Each of the eluted fractions was then assayed in two the 8 9 growth factor assay, in each instance using the 88/514 reference rIL-6 preparation obtained from the National Institutes of Health, Bethesda, as the calibrating standard (30,32).' Fractions around and including each IL-6 peak (by appropriate ELISA) were pooled, and the IL-6 and "associated" plasma proteins were isolated by immunoaffinity chromatography using the mAb "5IL6-Hl7"' cross-linked to Hydrazide Avidgel AX (BioProbe International) as per manufacturer's instructions. This mAb recognizes native human IL-6 (natural or recombinant) with high affinity and is the biotinylated reporter mAb in both the 4IL6/5IL6 and IG61/5IL6 ELISAs. The 100-150-and 400-500-kDa G-200 fraction pools were passed through a 1-2-cm column of the 5IL6 mAb affinity resin, and the columns were washed extensively with 75-100 ml of 150 mM NaCl, 10 mM Tris-C1 (pH 7.5). The adsorbed IL-6 and other "associated proteins were then eluted with 0.1 M glycine (pH 2.4) buffer, and the eluted fractions were immediately neutralized with an equimolar amount of Tris-C1 (pH 8.7). Aliquots of the immunoaffinity columnpurified fractions were then tested for IL-6 content in the B9 growth assay and by Western blot analyses using a rabbit anti-rIL-6 antiserum (23). Purified fractions containing significant quantities of IL-6 were then concentrated by freeze-drying without allowing the sample to dry completely. The concentrated sample was then electrophoresed through an SDS-PAGE gel and electroblotted to a PVDF membrane that was immediately stained with Coomassie Brilliant Blue for approximately 5 min. The stained protein bands were then excised and destained as previously described (23,27). Amino-terminal sequencing was performed directly on the excised protein band(s) bound to PVDF membrane by automated Edman degradation using an Applied Biosystems gas phase sequenator as described by Moos et al. (33). The FASTP algorithm was used to search the GenBank protein sequence database.
SDS-PAGE and Western Blot Analysis-Aliquots of various IL-6 preparations were suspended in Laemmli buffer (34), heated for 1 min at 100 "C, electrophoresed through an SDS-polyacrylamide gel (15%) under reducing conditions (0.7 M p-mercaptoethanol), and transferred to BA-85 nitrocellulose paper according to the method of Towbin et al. (35). IL-6 on the electroblotted nitrocellulose was detected using a rabbit polyclonal antibody to rIL-6 (23) and an immunoperoxidase procedure (Vectastain, ABC Elite kit).
ELISAs for Human IL-6"ELISAs were formatted and performed essentially as described by Kenney et al. (361.' Briefly, the 4IL6/5IL6 and IG61/5IL6 ELISAs used either 4IL6-Hll or IG61 as the trapping mAb. In both instances, biotinylated 5IL6-Hl7 was used as the reporter mAb. Each sample, usually undiluted, was assayed in triplicate using 50-p1 aliquots. The human E. coli-derived rIL-6 reference preparation 88/514 and a fibroblast-derived natural IL-6 preparation (the "laboratory standard LRS-1) were used for calibration of every assay. The concentration of the 88/514 standard was taken to be the designated concentration (1.25 pg/vial resuspended in 1 ml) (verified by us by Western blotting), whereas that of the fibroblast preparation the IG61/5IL6 ELISA (IL-6 in LRS at 0.56 pg/ml, corresponding to was derived by comparing it to the 88/514 rIL-6 preparation using 37,000 units/ml of the B9 assay) (32). In light of the particular data presented in this article, we have verified that the 4IL6/5IL6 ELISA does not react with any other human cytokines (including IL-la, IL-I@, basic fibroblast growth factor, IL-4, tumor necrosis factor a (TNF), granulocyte-macrophage colony-stimulating factor, and nerve growth factor) or serum components.
Sephadex G-200 Minicolumns-In order to rapidly fractionate plasma samples to obtain the high molecular mass protein fraction (>60 kDa), Sephadex G-200 minicolumns were used. A 1-ml Rainin blue pipette tip was packed with 300 p1 of prehydrated Sephadex G-200 gel. Eighty p1 of fresh plasma was then placed on top of this minicolumn and centrifuged for 15-20 s in a 1.5-ml Eppendorf tube in a microcentrifuge, and the resulting eluate was discarded. 100 pl of buffer (150 mM NaCl and 5 mM Tris-C1, pH 7.5) was then added to the minicolumn, which was centrifuged as before. This eluate, which represented the void volume ( VJ, was discarded. Another 100 pl of the same buffer was then added to the minicolumn and centrifuged as before. This eluate, which was calculated to contain proteins and protein complexes of size >60 kDa, was saved and assayed for IL-6 content in different ELISAs and in the B9 bioassay.
Human Serum and Plasma Samples-Serial serum samples from

IL-6 Comple
patients who had undergone bone-marrow transplantation were provided by Dr. F. Symington (37). These were kept frozen at -20 "C until used. Blood was drawn from patients with epidermolysis bullosa or extensive psoriasis (through the kind help of Dr. A. Gottlieb, Rockefeller University) or from normal volunteers into plain Vacutainer tubes or into tubes containing heparin or EDTA and centrifuged immediately for 15 min a t 3000 rpm and the plasma was separated and loaded immediately onto Sephadex G-200 columns. The maximum interval between venisection and loading of the plasma sample on to the G-200 column was 30 min.

RESULTS
Sephadex G-200 Gel Filtration of Bioactive and Immunoreactive IL-6 in Serum and Plasma-Previously published descriptions of the apparent size of serum-derived IL-6 eluted off a gel filtration column (38,39) were confirmed using the B9 growth factor assay or the IG61/5IL6 ELISA. Fig. 1 illustrates B9 bioassay and IG61/5IL6 ELISA data obtained when a 0.8-ml sample of serum from a bone marrow transplant patient with severe intercurrent infection (patient succumbed the next day) was fractionated through a Sephadex G-200 column. The highest level of B9 growth activity in the unfractionated serum from this patient was approximately 300 units/ml on the day just before death (37). Fig. 1 shows that all of the B9-active IL-6 was found in a low molecular mass (13-20 kDa) peak that coincided with the bulk of the IG61/ 5IL6-detectable IL-6 antigen. (Monomeric rIL-6 of 21-kDa mass by SDS-PAGE elutes from a Sephadex gel filtration column with a Stokes radius consistent with that of a 13.5-kDa protein; see Ref. 40). The highest concentration of IL-6 antigen detected using the IG61/5IL6 ELISA in any fraction was approximately 200 pg/ml.
In contrast, use of the 4IL6/5IL6 ELISA to assay for IL-6 in these serum-derived fractions led to an unexpected result (Fig. 1). Although this ELISA was able to detect the low molecular mass B9-active IL-6, the bulk of the IL-6 antigen was in two peaks of 100-150-and 400-500-kDa mass at concentrations in the ng/ml range. It is noteworthy that the eluted fractions across these two peaks were completely devoid of B9 HGF activity either when assayed as such or following heat treatment (56 "C for 0.5 h). Furthermore, the addition of exogenous fibroblast-derived or E. coli-derived IL-6 to these fractions did not significantly inhibit their B9 activity when compared with the addition of exogenous IL-6 to the column buffer as a control (data not shown).
G 'xes in Blood out the addition of heparin or of EDTA) from patients with active cutaneous disease (epidermolysis bullosa or psoriasis) or even from disease-free volunteers revealed that virtually all of the IL-6 antigen reactive in the 4IL6/51L6 ELISA was reproducibly present in the two 100-150-and 400-500-kDa molecular mass peaks at concentrations in the ng/ml range (Fig. 2). This high molecular mass IL-6 antigen, or any IL-6 antigen for that matter, was minimally detectable in the IG61/ 5IL6 ELISA (Fig. 2). None of the unfractionated or plasmaderived fractions had any detectable B9 growth activity even after "heat treatment." Only low levels of B9 activity were detected in unfractionated serum from patients illustrated in Fig. 2 antigen were reproducibly detected using the 4IL6/5IL6 ELISA in elution fractions derived from plasma from normal volunteers (Fig. 2 0 is an example).
The 100-150-and 400-500-kDa high molecular mass complexes are exclusive to endogenous IL-6 because mixing fibroblast-derived "'S-labeled IL-6 with serum followed by gel filtration does not generate these complexes (data not shown). Even following addition of undiluted serum to the IL-6 sample (to approximately 80% v/v), the labeled fibroblast-derived IL-6 elutes in a broad distribution from 20 to 70 kDa (23).
Characterization of IL-6 in the High Molecular Mass Complexes-Did the two high molecular mass peaks reactive in the 4IL6/5IL6 ELISA contain the IL-6 protein or did they represent cross-reaction of the mAbs with other antigens? IL-6 present in pools of fractions corresponding to the 100-150and 400-500-kDa regions in the elutions illustrated in Figs. 1 and 2 were purified by immunoaffinity chromatography through a 5IL6 mAb column. Western blot analyses of the immunoaffinity column eluate under completely reducing and denaturing conditions using a polyclonal anti-rIL-6 rabbit antiserum revealed that the 100-150-kDa complexes contained very high concentrations of the 23-25-and 27-30-kDa monomeric IL-6 species, as well as a 15-17-kDa IL-6 species (Fig. 3A). In contrast, the 400-500-kDa complexes contained exclusively the 27-30-kDa IL-6 monomer (Fig. 3B). Thus, there is an intrinsic biochemical distinction to be made between the 100-150-and 400-500-kDa IL-6 complexes. It is noteworthy that IL-6 was present a t unexpectedly high concentrations in immunoaffinity column eluates from pools of the serum elution in Fig. 1 (approaching 1 pg/ml). For comparison in the Western blot, the lanes in Fig. 3 marked "IL-6" each contain 25 ng of fibroblast-derived IL-6 as a standard. Strikingly, the immunoaffinity column eluates, including those from normal volunteers, were active in the B9 growth assay to an extent approximately commensurate with the presence of the Western-blottable IL-6 antigen (Table I).
Amino-terminal Sequencing of Serum-derived IL-6 and Its "Associated"Proteins-In order to unequivocally demonstrate that the strong IL-6 antigenicity recognized in the 4IL6/5IL6 ELISA (Figs. 1 and 2) and the IL-6 proteins detected (Fig. 3) were in fact IL-6 molecules, the immunoaffinity-purified preparation from the 100-150-kDa pool illustrated in Fig. 1 and in Fig. 3 was concentrated and electrophoresed in a preparative SDS-PAGE, electroblotted onto PVDF membrane, and stained briefly with Coomassie Brilliant Blue (Fig. 4). Despite very extensive washing of the 5IL6-mAb immunoaffinity column prior to protein elution, a t least nine Coomassie-stained protein bands ranging in molecular mass from 15 to 70 kDa could be clearly distinguished on the electroblot. Each band was excised and destained, and the amino terminus was

. SDS-PAGE and Coomassie Brilliant Blue staining of proteins transferred to PVDF
membrane from the 5IL6 mAb immunoaffinity-purified IL-6 preparation (Fig. 3) derived from the 100-150-kDa serum-derived IL-6 pool illustrated in Fig. 1. Each of the hands listed on the right (peptides 1-9) was cut and sequenced using the automated Edman degradation procedure. sequenced using the automated Edman degradation procedure according to the method of Moos et al. (33). Table I1 lists the amino-terminal sequences obtained and the identity of each of the proteins sequenced.
The 15-17 and the 27-30 kDa bands contained IL-6 proteins starting at Val:"', which is the correct amino terminus of the Nand O-glycosylated forms of fibroblast-derived IL-6 (27), whereas the 23-25 kDa band was IL-6 protein with the amino terminus a t Ala'", which is the correct amino terminus of the O-glycosylated fibroblast-derived IL-6 (27). From the concentration of each of the IL-6-derived amino acid residues (signals in the 5-10 PM range) in the Edman degradation, we estimate that the concentration of IL-6 protein in the original  Fig. 1 is likely to have been at least 5 pg/ml. At the very least, these data bring into serious question the numerical validity of the ever-burgeoning IL-6 clinical literature.
In addition to unambiguously identifying serum-derived IL-6 by its amino-terminal sequence, the data summarized in Table I1 also identify several of the "associated" proteins. Peptides 1 and 2 included the correct amino terminus of Creactive protein. Peptides 4 and 8 have sequences identical with the amino terminus of serum albumin. The amino terminus of peptide 5 matched that of the carboxyl-terminal cleavage product of complement factor C4b (the "7-chain" of C4). Peptide 6 is the protease factor I-cleavage product of the a-chain of complement factor C3. The molecular mass and IL-6 association of peptide 7 is consistent with the inference that it is likely to be the soluble IL-6 receptor (a truncated 50-kDa version of IL-6R p80) (41). It is noteworthy that thus far we have not detected a2-macroglobulin in "association" with IL-6 (42).
Fractionation of Fresh Plasma through Sephadex G-200 Minicolumns-We have compared the ability of the 4IL6/ 5IL6 and IG61/51L6 ELISAs to estimate IL-6 concentrations in fresh plasma and have attempted to relate these data to the B9 growth factor activity observed (Figs. 5 and 6). For this purpose, blood was collected from two normal volunteers (samples marked "A" and "B") either with heparin or with EDTA as the anticoagulant, and the respective plasma samples were assayed in the two ELISAs either unfractionated or following Sephadex G-200 minicolumn fractionation. Additionally, samples of heparinized plasma from 14 different patients with varying degrees of psoriasis (samples marked 1 through 14) were also assayed in the two ELISAs either unfractionated or following G-200 minicolumn fractionation. Each of the ELISAs was calibrated using either the 88/514 E. coli-derived rIL-6 standard (Fig. 5, A and B ) or the fibroblastderived natural IL-6 laboratory standard (Fig. 5, C and D).
Aliquots from unfractionated heparinized plasma (samples A and B and 1-14) were also heat-treated at 56 "C for 30 min and then assayed in the B9 growth factor assay (Fig. 6). Fig. 5 shows that the 4IL6/5IL6 ELISA provides estimates of IL-6 levels in plasma that are generally in the 1-10 ng/ml range, whereas the IG61/5IL6 ELISA yields estimates that are more than an order of magnitude lower. The use of fibroblast-derived IL-6 (LRS-1) as standard provides a higher estimate of plasma IL-6 levels than does the rIL-6 standard (IRS), even though the concentration of IL-6 in LRS-1 was assigned using the designated value of IRS in the IG61/5IL6 ELISA; this result is consistent with the observation that the biochemical nature of rIL-6 and fibroblast-derived IL-6 is  Fig. 6. The bioassay was also calibrated using the 88/514 rIL-6 interim reference standard. different from that of IL-6 in human plasma. The use of heparin as anticoagulant increases the IL-6 immunoreactivity detected in unfractionated plasma when compared with EDTA as anticoagulant. The G-200 minicolumn fractionations that yield plasma eluates containing proteins >60 kDa lead to increased detection of the IL-6 antigen in some instances, as compared with unfractionated heparinized plasma. The data obtained in the B9 assay not only do not correlate with any of the ELISA data but also lead to very low estimates of IL-6 concentrations (one B9 assay unit corresponds to approximately 15 pg/ml of fibroblast-derived natural IL-6 in our hands) (compare Figs. 5 and 6). Taken together, the salient feature that emerges from the data in Figs. 5 and 6 is that the plasma samples evaluated using the 4IL6/5IL6 ELISA, even those from normal volunteers, contain IL-6 at concentrations in the range of 1-10 ng/ml. This is an estimate that is far higher than that obtained using either the IG61/ 5IL6 ELISA or the B9 bioassay.

DISCUSSION
The data in this article raise novel questions about the transport and bioavailability of human IL-6 in the peripheral circulation. As such, these questions are of broad relevance to discussions about the structure and functions of cytokines in vivo. IL-6 in plasma/serum exists primarily in the form of 100-150-and 400-500-kDa complexes that are not readily detectable in the conventional B9 bioassay and in ELISAs that are otherwise "highly sensitive" for rIL-6; however, using appropriate assays that include the 4IL6/5IL6 ELISA and even direct amino acid sequencing, the presence of IL-6 in such complexes at high concentration can be unequivocally established. The unexpected properties of the 4IL6/5IL6 ELISA include its ability to preferentially detect "complexed high molecular mass 1L-6 in human plasma/serum.
Early studies evaluating the concentration and nature of IL-6 in human body fluids, particularly serum, in health and disease using different bioassays presented a dilemma (1,4,14,15,29,38,43,44). When the "hybridoma" growth factor assay (carried out in murine cell lines such as B9 or MH60.BSF-2) was used, sera from normal individuals had little detectable IL-6 activity, whereas sera (or other body fluids) from individuals in rejection following renal transplantation, those with rheumatoid arthritis, or those with various bacterial infections, and even volunteers injected with TNF or endotoxin had detectable hybridoma growth activity in the range from barely detectable to a maximum of a few hundred biological units/ml of serum. Given the conversion factor that approximately 15 pg/ml of natural human IL-6 can correspond to 1 hybridoma growth factor unit (e.g. in B9 cells), these data suggested that the concentration of circulating IL-6 was approximately in the range from a few pg/ml to a maximum of approximately 1 ng/ml. Nevertheless, when human sera were assayed using the hepatocyte-stimulating factor assay in human cells (stimulation of al-antichymotrypsin secretion by human Hep3B cells, and its complete neutralization by polyclonal anti-rIL-6 antiserum), we estimated circulating IL-6 concentrations in the range of 2.5-120 ng/ml in patients with sepsis and in volunteers given endotoxin or T N F (14,15,29). In at least one instance (volunteers given TNF), the same serum samples that contained a few hundred units/ml of hybridoma growth activity were simultaneously estimated to contain >lo0 ng/ml hepatocyte-stimulating activity that was completely neutralizable by anti-IL-6 antibody (29). In a second study (patients with active psoriasis), plasma samples that appeared to contain IL-6 hepatocyte-stimulating activity (1-10 ng/ml) did not contain any detectable B9 growth factor activity (despite the use of the conventional heat-inactivation step)(lO).
Furthermore, two different laboratories reported that gel filtration of serum followed by assays of the eluted fractions using the murine B9 cells to detect human IL-6 showed that all of the detectable IL-6 was -20-25-kDa molecular mass (38,39). However, when we purified IL-6 from sera/plasma of volunteers administered endotoxin or TNF using a polyclonal anti-rIL-6 immunoaffinity column and characterized it by Western blots under "nominally denaturing and reducing conditions," all of the IL-6 appeared to be 43-45 kDa molecular mass (15,29). Under completely denaturing conditions, the monomeric IL-6 species purified from serum, cerebrospinal fluid, or amniotic fluid of patients with bacterial infection consisted of proteins of molecular mass in the range of 23-30 kDa, as evaluated by SDS-PAGE and Western blotting (15,16).
Since these early studies, numerous other investigators have used the hybridoma growth factor assay as carried out in murine cell lines and have reported elevations of human IL-6 levels in sera of patients with various diseases; invariably, the concentrations reported were in the range from barely detectable to a few hundred units/ml (44,45). Also, various investigators have since developed mAb-based ELISAs or radioimmunoassays for IL-6 using various recombinant or cell culture-derived IL-6 preparations for raising the mAbs and for optimizing and calibrating the assays (32,46). These immunoassays, which are generally highly sensitive when calibrated using rIL-6 or cell-culture-derived IL-6 (sensitivity down to 1-5 pg/ml) (46), have as a class, revealed that IL-6 is present in sera from normal individuals at a concentration in the range of 10-75 pg/ml, whereas individuals with various disease states (infections, neoplasia, autoimmune diseases) have elevations in IL-6 levels that are usually a few hundred pg/ml but can be a maximum of 1-2 ng/ml (see Ref. 46 for an example).
Using the murine B9 growth factor assay and the "highly sensitive" (for rIL-6) IG61/5IL6 ELISA, we can essentially confirm many of the prior observations in the 1L-6 literature summarized above. As examples, using these two assays, we found that plasma from normal volunteers had little detectable IL-6 and that plasma from patients with psoriasis had IL-6 concentrations that ranged from undetectable to a few hundred pg/ml at the most (Fig. 5). As additional confirmation of the previously published literature, Fig. 1 verifies that almost all of the 1L-6 detected in serum using the B9 bioassay or the IG61/5IL6 ELISA is of low molecular mass. However, the bulk of circulating IL-6, in complexes of 100-150-and 400-500-kDa mass, was not visible in the B9 bioassay or in the IG61/5IL6 ELISA but was detected using the 4IL6/5IL6 ELISA.
The 100-150-and 400-500-kDa IL-6 complexes contained a distinctive distribution of IL-6 species. The 100-150-kDa complexes contained the 15-17-, 23-25-, and 27-30-kDa IL-6 species, whereas the 400-500-kDa complexes contained exclusively the 27-30-kDa IL-6 species, suggesting that the presence of particular IL-6 species in these complexes was a specific biochemical event. Additionally, a number of other plasma proteins copurified during the IL-6 immunoaffinity column chromatography of the serum-derived 100-150-kDa pool from Fig. 1. Amino-terminal sequencing unequivocally identified not only the three IL-6 polypeptides isolated but also identified the y-chain of complement factor C4b, Creactive protein, serum albumin, and the carboxyl-terminal protease factor I-cleavage product of the a-chain of complement factor C3b (also called C3c in Ref. 47) as part of the "associated" proteins. Of the two other IL-6-associated protein bands that could not be sequenced by automated Edman degradation because of blocked amino termini, peptide 7 corresponds in molecular mass (50 kDa) to the soluble IL-6 receptor that has been observed in normal serum at high concentrations (up to 70 ng/ml) (41). Because of the report that rIL-6 binds to a,-macroglobulin that then is supposed to serve as a carrier protein for this cytokine (42), it is noteworthy that we have not yet detected a,-macroglobulin as part of the "IL-6-associated proteins. Although most of the plasma proteins enumerated in Table I1 do not appear to be bound to and eluted from immunoaffinity columns used as controls (perhaps with the exception of albumin), it remains to be critically evaluated whether each of these polypeptides is "specifically" bound to IL-6 and with what stoichiometry. Nevertheless, the major conclusion from our data is that the human IL-6 polypeptide is indeed present at high concentrations in the 100-150-kDa complexes in serum/plasma, as verifiable by direct amino acid sequencing.
The presence of large quantities of "complexed IL-6 in the peripheral circulation may constitute a reservoir or depot for this cytokine in which IL-6 awaits the appropriate local signal for dissociation of the complex and its consequent physiological action. The protein context from within which IL-6 is presented to different target tissues may determine the spectrum of its biological actions. The present data indicate that the presence of circulating IL-6 in high molecular mass complexes prevents the cytokine from displaying biological effects otherwise attributable to rIL-6. The markedly altered immunological reactivity of IL-6 in the 100-150-and 400-500-kDa complexes provides additional evidence of its altered biochemical and biological status.
The new data raise novel questions about IL-6 biology. What is the spectrum of biological activities truly attributable t o IL-6 as it exists in human blood? What is the nature of IL-6 in the extravascular tissue fluid? Are there biochemical mechanisms that control the bioavailability of circulating cytokines? In essence, we are beginning to ask a layer of questions that had hitherto not been asked about cytokine transport and function in vivo.