Preserved granulocyte formation and function, as well as bone marrow innervation, in subjects with complete spinal cord injury


Per Ole Iversen, Department of Nutrition, Institute of Basic Medical Sciences, PO Box 1046 Blindern, 0316 Oslo, Norway.


Patients with a spinal cord injury are at risk of infections and is partly attributed to immobilization. Their lymphocyte-mediated immunity is impaired and the growth of blood progenitor cells is reduced. An adequate immune response depends on granulocytes being mobilized rapidly and activated properly, at the inflammatory site. Possibly this requires a coordinated interaction between the autonomous nervous system and cells within the haematopoietic bone marrow. Granulocyte function in the spinal cord injured has not been evaluated. Although there is evidence that the bone marrow in rodents is innervated, it is uncertain whether human bone marrow is similarly affected. Microscopy and immunolabelling followed by flow cytometry, showed that blood and bone marrow counts of leucocyte subsets were similar in paraplegic, tetraplegic and control subjects (P > 0·05). Neutrophilic migration and oxygen consumption, as well as eosinophil activation, assayed as release of eosinophilic cationic protein or CD69 expression, were not altered after spinal cord injury (P > 0·05). Cryostat sections of human bone marrow biopsies stained positive with glyoxylic acid, indicating the presence of catecholamine-containing nerves in both the patients and the controls. We conclude that terminal differentiation and formation of granulocytes, as well as their functional capacity, do not depend appreciably on supraspinal nervous regulation.

Being encased in the bone tissue, the bone marrow is not easy to study (Iversen, 1997). In rodents, both efferent and afferent nerve fibres have been detected. These bone marrow nerves contain catecholamines or various peptides, and they terminate in close proximity to the blood vessels, the stroma and the haematopoietic cells, indicating a regulatory role regarding bone marrow perfusion and/or haematopoiesis (Calvo, 1968; DePace & Webber, 1975; Yamazaki & Allen, 1990; Felten et al, 1992; Tabarowski et al, 1996). The presence of cognate receptors for neuronal transmitters on bone marrow cells supports this notion (Maestroni & Conti, 1994; Rameshwar & Gascon, 1995; Broome et al, 2000). Surprisingly therefore, we could not demonstrate any function of innervation on mouse tibial bone marrow function, by denervation, electrical nerve stimulation or neonatal chemical sympathectomy (Benestad et al, 1998a). However, these findings have been challenged on methodological grounds (Maestroni, 1998; Miyan et al, 1998a). Our response to the criticism was that previous findings did not necessarily invoke nerves, that the responses of intact animals may have been indirect (caused by, for example, released cytokines or endotoxin), and that published results were inconsistent and even contradictory (Benestad et al, 1998b).

It is well established that cytokines regulate haematopoiesis (Metcalf & Nicola, 1995). A rapid and adequate inflammatory response requires an early mobilization of granulocytes from the bone marrow storage pool to the inflammatory site as well as a properly timed activation of their microbe killing ability, processes that are possibly dependent on a functional autonomous nervous system, in addition to the cytokines. Patients with spinal cord injuries are prone to infections (Levi et al, 1995; Hartkopp et al, 1997). This probably is partly the result of immobilization, but could also reflect a maladaptive interaction between the decentralized autonomic nervous system and the bone marrow. Kliesch et al (1996) noted a depressed lymphocyte function 2 weeks after spinal cord injury, which could be partly restored upon rehabilitation. However, the study population was stressed patients (high urine cortisol levels) who probably had unstable injuries.

To provide further information about the nervous system concerning human bone marrow function, we have studied haematopoietic cell formation and lymphocyte function in subjects with complete and stable injuries to their spinal cords. These patients lack supraspinal activity and develop a decentralized autonomous nervous system below the injury level. In these subjects we demonstrated impaired growth of haematopoietic progenitor cells, and reduced lymphocyte-mediated immunity (Iversen et al, 2000), suggesting that the nervous system can modify blood cell formation and function.

To our knowledge, non-adaptive immunity has not been extensively studied in subjects with a spinal cord injury in their steady-state. Consequently, we examined the numbers and functions of blood neutrophilic and eosinophilic granulocytes in paraplegic and tetraplegic patients.

Although the presence of nerves within the bone marrow parenchyma of rodents is fairly well established, firm evidence for their presence is apparently lacking for humans (Pelletier et al, 2002). Therefore, the second intention of the present study was to clarify the existence of human bone marrow innervation. To this end we examined the bone marrow from both healthy humans and subjects with complete and stable spinal cord injury, using histochemistry and antibody labelling. These data were compared with those obtained in mice.

Materials and methods


The murine study was approved by the animal experimentation committee. We used adult, male Balb/c mice (body weight about 25 g; Møllegaard, Denmark). The animals were kept in cages with a 12-h light/dark cycle and provided food and water ad libitum. They were put to death with an overdose of barbiturate (i.p.). Irides and tibial bones were carefully removed. The irides were mounted on coated glass slides. Smears were made from marrow plugs (Benestad et al, 1998a). These preparations were immediately processed for glyoxylic acid-induced fluorescence or immunohistochemistry with antibodies.

Human subjects

The study was conducted according to the Helsinki declaration. It was approved by the regional committee on medical research ethics, and written consent was obtained. Males (age 24–42 years) whose complete injury of their spinal cords, either in the cervical (C5–C7, tetraplegics, n = 6) or thoracic (Th5–Th10, paraplegics, n = 6) region, occurred more than 5 years previously, were included (Iversen et al, 2000). These spinal cord injured subjects met the internationally accepted criteria for level and completeness of their injuries (Dunno et al, 1994), and they were otherwise healthy. As controls, six age-matched, healthy male subjects were included.

The subjects fasted overnight before blood and bone marrow specimens were collected between 09.00 and 10.00 am to avoid possible circadian variations (Maestroni et al, 1998; Iversen et al, 2002). A venous blood sample (5 ml) was collected, and after providing local anaesthesia with lidocain (10 mg/ml; Xylocain, AstraZeneca, Sweden), bone marrow aspirates (2 ml) were collected from the sternum and the iliac crest, and bone marrow biopsies from the iliac crest only. We obtained human normal gut preparations from patients undergoing bowel resection as a result of trauma (Department of Gastrosurgery, Ullevaal University Hospital, Oslo, Norway). This tissue and the bone marrow biopsies were quick-frozen in a mixture of isopenthane and dry ice and stored at −80°C until further analyses.

Haematological variables, neutrophil migration and oxygen consumption assays

Haematological values were determined with standard methods. A 200-cell differential count was performed on a May–Grünwald/Giemsa stained blood smear.

Neutrophils were isolated from the blood sample using density-gradient centrifugation (Lymphoprep; Nygaard, Oslo, Norway) after erythrocyte aggregation. Membrane exclusion of trypan blue in these neutrophils exceeded 98%. We determined their migration rate across a 140 μm thick porous membrane (pore diameter 5 μm) as described (Grimstad & Benestad, 1982). Briefly, the cell suspension was positioned in a test tube containing interferon-γ (specific activity 6·34 × 107 U/mg protein; Boehringer Mannheim, Mannheim, Germany). The test tube was then incubated for 2 h at 37°C before the migrated neutrophils were washed free and counted electronically with a Coulter counter (Coulter Electronics Ltd, Harpenden, UK). The migration rate of these primed neutrophils was determined with or without the test-stimulus N-formyl-methionyl-leucyl-phenylalanine (FMLP; 10−8 mol/l).

Oxygen consumption in neutrophils was measured as a change in oxygen tension in an incubation chamber, as described (Opdahl et al, 1987). Briefly, the neutrophils were kept in a chamber where the oxygen tension was continuously measured with a polarographic electrode (MSE, UK). Measurements were conducted both at baseline and upon stimulation with FMLP (10−8 mol/l).

Assessment of eosinophil cationic protein release and eosinophil CD69 expression

The purification of eosinophils and assessment of eosinophilic cationic protein (ECP) were performed as outlined by Iversen et al (1997). Briefly, the eosinophils were isolated with a discontinuous metrizamide gradient separation (Metrizamide; Nycomed, Oslo, Norway), yielding a cell purity above 97% and with a Trypan blue membrane exclusion exceeding 97%. The eosinophils were stimulated with granulocyte-macrophage colony-stimulating factor (GM-CSF; 1 ng/ml, 1 h), while Hanks’ balanced salt solution was used as the negative control. The eosinophils were incubated with serum-coated beads before ECP was determined with a radioimmunoassay (Pharmacia, Uppsala, Sweden).

The eosinophil expression of the activation marker CD69 was measured according to Hartnell et al (1993). Freshly prepared eosinophils were incubated with a monoclonal antibody raised against the CD69 antigen or with an irrelevant isotype antibody. After one wash, a secondary antibody was added before the samples were analysed by flow cytometry (FACScan; Beckton-Dickinson, Mountain View, CA, USA). CD69 expression was measured both at baseline and after stimulation with GM-CSF (10 ng/ml, 1 h).

Human bone marrow examination

Marrow smears were stained with May–Grünwald/Giemsa and 200 cells were morphologically classified by microscopy.

Mononuclear cells were isolated from the aspirates after haemolysis of erythrocytes and centrifugation. These cells were further phenotyped with fluorescently labelled monoclonal antibodies (Beckton-Dickinson) raised against the antigens: CD3 (T lymphocytes), CD19 (B lymphocytes), CD34 (immature progenitors), CD45 (pan leucocyte marker), and CD16/CD56 [natural killer (NK) cells]. The fractions of cells expressing these antigens were determined with flow cytometry (FACScan; Beckton-Dickinson).

Histochemistry for visualization of nerve fibres

The glyoxylic acid condensation reaction was used to visualize sympathetic nerve fibres (Furness & Costa, 1975; Voyvodic, 1989; Warburton & Santer, 1994). Cryostat sections of frozen tissue (14 μm) were melted on to uncoated glass slides. The sections were immediately covered with 2% glyoxylic acid (Sigma-Aldrich, St Louis, MO, USA) in 0·1 mol/l phosphate buffer, for 30 s at room temperature. The liquid was removed with lens tissue paper. The sections were dried under a stream of filtered compressed air for 1 min and then placed above Kieselgel in a closed container for 10 min. Finally, sections were mounted in liquid paraffin and incubated for 2 min at 100°C. The unfrozen tissue of irides and bone marrows from mice were treated identically. Nerve fibres stained with glyoxylic acid were viewed in a fluorescence microscope (excitation 390–440 nm, beam splitter 510 nm, emission barrier 520 nm).

To identify catecholaminergic nerve endings, we used anti-tyrosine hydroxylase (TH) monoclonal antibodies (clones 2 and 16; Sigma-Aldrich) and immunohistochemistry. As additional neuronal markers, we used antibodies raised against either neurofilament 200 (NF200; monoclonal antibody, clone N52, Sigma-Aldrich), or neuronal nitric oxide synthase (nNOS; polyclonal antibody, Chemicon International Inc., Temecula, CA, USA). As secondary antibodies we used rhodamine-conjugated goat antimouse IgG purchased from Sigma-Aldrich or Molecular Probes Inc. (Eugene, OR, USA) and goat antirabbit IgG (Chemicon). Cryostat sections (14 μm) melted on to coated glass slides were pretreated with acetone or 4% paraformaldehyde at room temperature (5 min) and methanol at 4°C (10 min), rinsed in 0·1 mol/l phosphate-buffered saline (PBS) and incubated in 0·1 mol/l PBS containing either 0·2% bovine serum albumin or 5% goat serum, and 0·3% Triton X-100. Then the sections were incubated with the primary antibody diluted in PBS containing serum and Triton X-100 (dilutions: anti-TH 1:1000, anti-NF200 1:50, anti-nNOS 1:1000) in a humid chamber for either 1 h at room temperature or overnight at 4°C. After rinsing with PBS, the sections were incubated with the secondary antibody diluted in PBS containing serum and Triton X-100 (dilutions: antimouse IgG 1:200, antirabbit IgG 1:100) in a dark humid chamber for 1 h at room temperature, rinsed with PBS and finally cover-slipped in mounting medium (Citifluor Ltd, London, UK). Sections without the primary antibody were used as negative controls. The sections were examined with the microscope equipped with appropriate filter cubes to detect rhodamine-stained structures. Unfrozen specimens were treated identically. The micrographs in Figs 1 and 2 were obtained with Fujichrome (Fuji Photo Film Co. Ltd, Tokyo, Japan) 400 ASA film pushed to 800 ASA.

Figure 1.

Sympathetic nerve fibres stained with glyoxylic acid. Nerves were detected in human bone marrow specimens from a control subject (A), a paraplegic patient (B), and a tetraplegic patient (C). Positive controls were nerves in human gut wall (D), and in mouse bone marrow (E) and iris (F). The arrowheads indicate the nerve fibres. Each of the illustrations (A–C) is from one subject and is representative of the other five subjects in each experimental group. Scale bar represents 50 μm.

Figure 2.

Neurofilaments were stained by anti-NF200 monoclonal antibodies in mouse bone marrow but not in human bone marrow. Nerve fibres (arrowheads) in close contact with a blood vessel in the human gut wall (A). Nerve fibre (arrowhead) in mouse bone marrow (B), and a bundle of nerve fibres (arrowhead) in mouse iris (C). Scale bar represents 50 μm.


We performed triplicate measurements and calculated mean and standard error of the mean (SEM) for the six human subjects in each of the three study groups. Differences were evaluated with the Kruskall–Wallis test with the Bonferroni correction, or Wilcoxon sum rank test for paired samples, as appropriate. Significance was assumed for P < 0·05.


Unchanged blood and bone marrow cell levels in spinal cord injured subjects

It is evident from Table I that routine haematological variables, the numbers of leucocytes and their subsets, did not deviate in any direction (P > 0·05) when the paraplegic and tetraplegic patients were compared with controls. The numbers of NK cells, T and B lymphocytes were all within the normal range (data not shown).

Table I.  Haematological characteristics of the three study groups.
Blood variableControlsTetraplegiaParaplegia
  1. The neutrophil counts include segmented and band granulocytes, the latter never accounted for more than 2% of the total number of granulocytes. Values are given as mean ± SEM; n = 6, P > 0·05 for each parameter when tested among the three study groups.

Haemoglobin (g/dl)14·3 ± 1·814·1 ± 1·013·7 ± 1·5
Erythrocytes (×1012/l)5·1 ± 0·65·0 ± 0·75·4 ± 0·7
Haematocrit0·41 ± 0·050·38 ± 0·080·40 ± 0·07
Thrombocytes (×109/l)389 ± 23418 ± 19401 ± 21
Leucocytes (×109/l)8·9 ± 1·29·7 ± 1·79·4 ± 2·1
Leucocyte subsets (×109/l)
 Lymphocytes2·6 ± 0·073·0 ± 0·132·9 ± 0·2
 Neutrophils5·5 ± 0·116·1 ± 0·115·7 ± 0·16
 Eosinophils0·4 ± 0·010·3 ± 0·010·4 ± 0·02
 Monocytes0·3 ± 0·010·3 ± 0·010·4 ± 0·02

Table II shows that there were no apparent differences between the percentages of any bone marrow cell type among the paraplegic, tetraplegic and control subjects. Moreover, we could not find any differences when we compared smears from the sternal and iliac crest aspirates within each study group (Table II).

Table II.  Differential count of nucleated bone marrow cells.
SternumIliac crestSternumIliac crestSternumIliac crest
  1. The values are given as ranges. Megakaryocytes were observed in all samples. No cells of non-marrow origin were identified in any of the samples (n = 6).

Erythropoiesis (%)26–3929–3824–3624–3927–3325–38
Myelopoiesis (%)51–6958–6653–6755–6950–6351–61
Blasts/promyelocytes (%)1–32–42–31–31–42–3
Myelo-/metamyelocytes (%)21–3126–3325–3027–3327–3426–31
Granulocytes (%)25–3829–3326–3824–3627–3922–36
Lymphocytes (%)4–83–85–93–84–73–8
Plasma cells (%)1–31–31–41–31–41–3

In support of the morphological findings, we could not detect any significant differences in the cell concentrations either among the three study groups, or between samples from the sternal and iliac crest aspirates, after phenotyping and subsequent flow cytometry (Table III).

Table III.  Phenotypes of bone marrow cells.
CD markerControlsTetraplegiaParaplegia
SternumIliac crestSternumIliac crestSternumIliac crest
  1. Values are expressed as a percentage of 10 000 scored cells and presented as mean ± SEM; n = 6, P > 0·05 for each variable and sample type when tested among the three study groups. The sternum-to-iliac crest ratios are given in parentheses, which were never significantly different.

CD 4593·8 ± 3·890·4 ± 7·689·8 ± 2·192·7 ± 3·889·2 ± 1·992·6 ± 3·6
(1·02 ± 0·23) (0·97 ± 0·32) (0·98 ± 0·21) 
CD 315·6 ± 4·914·1 ± 6·914·1 ± 5·014·6 ± 6·112·5 ± 4·211·8 ± 3·1
(1·01 ± 0·21) (0·99 ± 0·14) (1·03 ± 0·18) 
CD 193·5 ± 1·13·9 ± 1·83·2 ± 1·43·5 ± 2·74·0 ± 1·83·2 ± 1·0
(0·96 ± 0·19) (0·94 ± 0·24) (1·04 ± 0·15) 
CD 16/564·6 ± 0·44·3 ± 0·84·2 ± 0·84·6 ± 0·44·2 ± 1·83·4 ± 1·1
(1·02 ± 0·11) (0·93 ± 0·23) (1·04 ± 0·16) 
CD 341·8 ± 0·71·6 ± 0·82·0 ± 1·42·1 ± 1·11·9 ± 0·51·6 ± 0·7
(1·01 ± 0·19) (0·99 ± 0·06) (1·03 ± 0·21) 

No change in granulocyte function in spinal cord injured subjects

Expressed as fraction of baseline values (random migration) the blood neutrophil migration rates were 194 ± 24%, 203 ± 30% and 185 ± 25% (n = 6, P > 0·05) in the controls, in the tetraplegic subjects and in the paraplegic subjects, respectively. No significant changes in neutrophil oxygen consumption, expressed as a fraction of baseline values, were detected among the three study groups: 175 ± 11 (controls), 183 ± 15 (tetraplegia), 180 ± 13% (paraplegia).

While the eosinophil stimulator, GM-CSF, markedly enhanced the ECP release from blood eosinophils, it was similar (P > 0·05) in the three study groups, expressed as a fraction of baseline values: 157 ± 11% (controls), 149 ± 14% (tetraplegia), 161 ± 12% (paraplegia), as was the spontaneous release, i.e. without stimulation with GM-CSF (data not shown). The eosinophil expression of the activation marker CD69 was greatly enhanced by GM-CSF, but to similar (P > 0·05) levels in all three groups: 345 ± 23% (controls), 314 ± 26% (tetraplegia), 357 ± 30% (paraplegia), all expressed as a fraction of baseline values.

Human bone marrow is supplied with sympathetic nerves

We applied glyoxylic acid to identify sympathetic nerves. It is clear from Fig 1(A–C), that human bone marrow biopsies from both controls and from spinal cord injured subjects were innervated with catecholaminergic nerve terminals. In both patients and normal subjects, about five nerve fibre segments were seen in each cryostat section (about 15 × 2 mm in size). The length of each fibre segment varied from 10 to 125 μm. These segments were either single fibres in close contact with parenchymal cells or bundles of fibres in close proximity to blood vessels. As positive controls we could demonstrate varicose nerve fibres in the human gut wall (Fig 1D). Glyoxylic acid also stained nerve terminals in bone marrows and irides from mice (Fig 1E and F). In smears from mouse marrow, the nerve fibres were seen as a dense plexus around blood vessels.

In contrast, using the anti-TH monoclonal antibodies, we were unable to identify nerves either in human or mouse bone marrow, but varicose fibres could be visualized both in the human gut wall as well as in mouse irides (data not shown).

Neither could axons be detected in the human bone marrow with the anti-NF200 monoclonal antibody raised against neuronal filaments, whereas axons were easily recognized as single fibres or bundles of fibres in the specimen from a human gut (Fig 2A). Similarly, axons were labelled with the anti-NF200 monoclonal antibody in mouse bone marrows and irides (Fig 2B and C).

Whereas the anti-nNOS polyclonal antibody stained ganglion cells in the human gut wall, no stained neuronal structures were found in human bone marrow biopsies (data not shown).


This study did not find any impairment of granulocyte formation or function in subjects with complete spinal cord injury. We did not detect any change in the migration rate of blood neutrophils or in their consumption of oxygen when the three study groups were compared. Moreover, we did not demonstrate any difference in the eosinophil surface expression of the activation marker CD69 or the release of eosinophilic cation protein between the three study groups. Although it is known that granulocytes can be primed and activated ex vivo (MacNee & Selby, 1990), our handling of granulocytes did not seem to change their function so much as to invalidate conclusions drawn from their behaviour in vitro (Farstad et al, 1991; Løvås et al, 1996). Both the neutrophils and the eosinophils were strongly activated upon appropriate stimulation with either FMLP (neutrophils) or GM-CSF (eosinophils), both in controls and in spinal cord injured subjects. Hence it appears that the ability to mount a non-adaptive granulocytic response remains intact in the patients. This contrasts somewhat with our previous observation that the lymphocyte-mediated non-specific (NK cell) and adaptive (B and T lymphocyte) immunity were markedly reduced in otherwise healthy subjects with spinal cord injury (Iversen et al, 2000).

The numbers of blood leucocyte subsets were apparently not different among the three study groups. Similarly, the phenotyping of bone marrow cells revealed no differences in the concentrations of total leucocytes (CD45), or the lineage-committed cells, such as T-lymphocytes (CD3), B-lymphocytes (CD19), or the NK cells (CD16/CD56) among the three study groups. In line with this, we did not detect any alterations in the number of CD34+ cells. The numbers of the very immature, uncommitted CD34+ CD38 cells never exceeded 2% of the total number of CD34+ cells, and were similar among the three study groups (data not shown). Although the numbers of these early progenitor cells are apparently unchanged in spinal cord injured subjects, their proliferative capacity was found to be substantially reduced when assessed in vitro as the colony growth of long-term colony initiating cells (Iversen et al, 2000).

Data on innervation of human haematopoietic bone marrow is sparse. Recently, Pelletier et al (2002) demonstrated synaptophysin-containing nerve fibres close to human bone marrow blood vessels, but no fibres were detected among marrow parenchymal cells, i.e. in the haematopoietic compartment. We found that staining nerve fibres in bone marrow, particularly in human specimens, was more difficult than in other tissues, such as iris and gut, which we used as positive controls. The reason for this is not known. Of the antibodies that we used (different anti-TH antibodies, anti-NF200 and anti-nNOS), only anti-NF200 stained nerve fibres in bone marrow only in mice. Glyoxylic acid, however, stained varicose nerve fibres brightly in all cases. To our knowledge, the data we report here provide the first firm evidence for bone marrow innervation by the sympathetic nervous system in humans. Varicose nerve fibres, stained by glyoxylic acid, occurred in association with blood vessels as well as near parenchymal cells. Although data on bone marrow innervation are limited compared with those on other lymphoid tissues (Elenkov et al, 2000), the innervation of rodent bone marrow appears to be firmly established, both with direct staining of nerve fibres, e.g. with glyoxylic acid (Benestad et al, 1998a; Artico et al, 2002) and demonstration of transmitters (Maestroni et al, 1998; Broome & Miyan, 2000).

Although the staining of nerve fibres was only examined visually, there were no apparent differences between the controls and the two patient groups. If anything, the nerve staining was more, rather than less, preponderant in the marrow of patients.

In paraplegic and tetraplegic patients, the decentralized parts of the spinal cord and the sympathetic ganglia still function as reflex centres for the neuronal activity that controls autonomic functions of the body, although often in maladaptive ways that involve hyperreflexia (Karlsson et al, 1998; Weaver et al, 2001). Our findings suggest that the terminal differentiation and function of granulocytes are not under supraspinal control. There is a possibility though, that such control has been disrupted in the decentralized bone marrow, leading to a uniform hypoplasia with retained size-relation between the cell subsets. If this were the case, the still centrally connected bone marrow must have hypertrophied to furnish a normal compartment of blood cells. As cell formation in the bone marrow can be accelerated only by one order of magnitude or so at the most, and the supraspinally connected marrow in tetraplegic patients contributes only a small fraction of the total mass of haematopoietic marrow, this possibility is less likely.

The concept of an integrated immune defence system, involving the haematopoietic, immune and the neuroendocrine organs, has gained increasing experimental support (Blalock, 1994; Miyan et al, 1998b; Downing & Miyan, 2000). In addition, recent data also suggest that the innervation can modify bone marrow angiogenesis (Pelletier et al, 2002). Nevertheless, to clarify the mechanism underlying the possible interaction between nerves and bone marrow, further and more detailed investigation into the properties of the innervated and denervated bone marrow microenvironments is needed.


We thank all the subjects who generously gave their consent to participate in this study. The technical assistance from T. Flatebø and I. Strøm-Gundersen is gratefully appreciated. E. Sclichting kindly provided the human gut specimens. Financial support was provided by the Norwegian Cancer Society, the Research Council of Norway, and Throne-Holst Foundation.