Oncogenic Osteomalacia due to FGF23-Expressing Colon Adenocarcinoma

Abstract

Context:

Oncogenic osteomalacia, a paraneoplastic syndrome associated with hypophosphatemia due to increased urinary phosphate excretion, is caused by excessive synthesis and secretion of fibroblast growth factor 23 (FGF23), a phosphaturic hormone that is normally produced by osteocytes. Most cases of oncogenic osteomalacia have been associated with benign tumors of bone or soft tissue; however, whether malignant neoplasms can also produce and secrete FGF23 is currently unknown.

Objective:

The aim was to determine whether a malignant neoplasm could cause oncogenic osteomalacia through excessive production and secretion of FGF23.

Setting:

We describe an 80-year-old woman with stage IV colon adenocarcinoma who presented with severe hypophosphatemia (0.4 mg/dL; reference, 2.6–4.5 mg/dL).

Results:

Fractional excretion of phosphate was 34% (reference, <5% in the setting of hypophosphatemia), and plasma levels of FGF23 were highly elevated at 674 RU/mL (reference, <180 RU/mL). Immunohistochemical analysis of the patient's tumor showed strong staining for FGF23. Genetic analyses revealed a point mutation in the KRAS gene.

Conclusions:

We present the first case in which a malignant neoplasm is documented to produce and secrete FGF23, leading to renal phosphate-wasting. Oncogenic osteomalacia should be considered in the differential diagnosis for patients with a malignant tumor who present with hypophosphatemia.

Oncogenic osteomalacia, also known as tumor-induced osteomalacia (TIO), is a paraneoplastic syndrome characterized by hypophosphatemia due to increased urinary phosphate excretion and low levels of 1,25-dihydroxyvitamin D [1,25(OH)₂D]. Patients with TIO produce an excess amount of fibroblast growth factor 23 (FGF23), a phosphaturic hormone that is normally expressed in osteocytes, often leading to major elevations in plasma FGF23 levels (1). The vast majority of TIO cases have been associated with benign tumors of bone or soft tissue, such as hemangiopericytomas or other neoplasms of mesenchymal origin (2). Although malignant neoplasms can be associated with severe urinary phosphate-wasting and thus hypophosphatemia, it is unknown whether such tumors produce and secrete FGF23.

Several cases of TIO associated with malignant neoplasms, particularly small cell lung carcinoma (3–7) and prostate adenocarcinoma (8–12), have been reported; however, circulating FGF23 levels were not measured in most of these cases and evaluation of FGF23 expression by these neoplasms has never been reported. We performed serological, immunohistochemical, and genetic investigations in a woman with severe hypophosphatemia to establish that her metastatic colon adenocarcinoma produced FGF23.

Case Report

An 80-year-old woman with recently diagnosed stage IV adenocarcinoma of the cecum with liver metastases was referred to the inpatient nephrology consultation service for evaluation of severe symptomatic hypophosphatemia. Her comorbidities included hypertension, hypothyroidism, chronic obstructive pulmonary disease, and osteoporosis. Her medications were atenolol, levothyroxine, inhaled fluticasone, alendronate, and calcium supplements. There was no family history of hypophosphatemia and/or childhood rickets. She is married, has 3 healthy daughters, and is a former smoker.

Two days before admission she had undergone a liver biopsy, which showed metastatic adenocarcinoma of colorectal origin. On the day of admission she presented to clinic with several days of anorexia, fatigue, and muscle weakness, and laboratory studies revealed a serum phosphate level of 0.4 mg/dL (reference range, 2.6–4.5 mg/dL). Her last serum phosphate level had been 2.6 mg/dL when it was checked 10 years earlier. She had no known history of hypophosphatemia, alcohol intake, parathyroid disease, or gastrointestinal disease.

She was admitted to the hospital for further evaluation. Additional laboratory data are shown in Table 1. She was given oral sodium phosphate 750 mg (24 mmol) 3 times daily, in addition to multiple doses of iv sodium phosphate (Figure 1). Her symptoms improved; however, serum phosphate levels remained low despite frequent repletion. Fractional excretion of phosphate was inappropriately elevated, suggesting renal phosphate-wasting as the etiology of her hypophosphatemia. She did not have hypouricemia, metabolic acidosis, or glucosuria, thus excluding generalized proximal tubular dysfunction. Plasma levels of FGF23, measured with an assay that detects the intact hormone as well as C-terminal fragments, were significantly elevated, whereas the 1,25(OH)₂D levels were decreased despite severe hypophosphatemia and normal 25-hydroxyvitamin D levels (Table 1); these laboratory findings were consistent with TIO.

Table 1.

Laboratory Data

Parameter	Reference Range	On Admission	5 wk After Admission	12 wk After Admission
Phosphate, mg/dL	2.6–4.5	0.4	2.6	3.8
Calcium, mg/dL	8.5–10.5	9.2	9.0	9.1
Creatinine, mg/dL	0.6–1.5	0.7	0.7	0.7
FePO4, %	<5a	34	41	23
Alkaline phosphatase, U/L	30–100	757	363	313
25-hydroxyvitamin D, ng/mL	33–100	33	22	22
1,25(OH)2D, pg/mL	18–78	13	93	53
PTH, pg/mL	10–60	36	132	94
PTHrP, pmol/L	<2.0	1.1	—	—
FGF23, RU/mL	≤180	674	174	167

Abbreviation: FePO₄, fractional excretion of phosphate.

Dashes indicate that PTHrP levels were not obtained at those time points.

^aIn the context of hypophosphatemia; otherwise, reference range is 5–20%.

Figure 1.

Serum phosphate levels over time. Closed arrows indicate administration of sodium phosphate (15 mm iv), and the open arrow indicates administration of oxaliplatin/fluorouracil. Shaded box indicates the reference range for serum phosphate (PO₄).

Subject and Methods

Immunohistochemical analyses

To evaluate whether elevated levels of plasma FGF23 were originating from the patient's malignancy, we examined tissue obtained from the liver biopsy that had demonstrated metastatic colon adenocarcinoma. Immunohistochemical detection of FGF23 was performed using techniques adapted from studies on bone, which we have reported previously (13).

Sections were stained separately using antibodies directed against epitopes comprising amino acid residues 186–206 and 225–244 located within the C-terminal portion of FGF23 (Immutopics, San Clemente, California). Immunohistochemistry was performed simultaneously on several tumor sections and on control sections from bone biopsy specimens obtained from patients with chronic kidney disease. For negative controls, the anti-FGF23 antibody was omitted. Furthermore, sections of colon adenocarcinoma from 2 individuals who did not have clinical or laboratory findings to suggest TIO were stained for FGF23.

Molecular genetic analyses

Genomic DNA was extracted from tissue using the QIAamp Blood Mini Kit (QIAGEN, Valencia, California) using a modified protocol (14). To evaluate for the presence of copy number alterations, array comparative genomic hybridization was employed on tumor DNA. Furthermore, a multiplexed allele-specific assay, SNaPshot version 3 (Applied Biosystems, Carlsbad, California), was used to detect somatic mutations in tumor DNA extracted from the patient's liver biopsy. This assay is designed to search for 68 frequent, previously identified mutations in 14 cancer genes. Multiplex PCR using tumor DNA was followed by mutation detection using single base extension sequencing technology using an ABI 3730 DNA analyzer (Applied Biosystems, Carlsbad, California). In addition, separate PCRs using primers flanking EGFR exon 19, EGFR exon 20, and ERBB2 exon 20 were used to detect in-frame activating insertions or deletions in EGFR and ERBB2 that may not be detected using SNaPshot.

Results

Immunohistochemical features

When probed with the anti-FGF23(225–244) antibody, the patient's adenocarcinoma (but not the surrounding normal liver tissue) was strongly positive for FGF23 (Figure 2, B and D); the intensity of the staining was comparable to that of osteocytes in bone biopsy sections of a chronic kidney disease patient, which served as a positive control (Figure 2F). In the absence of the anti-FGF23 antibody, no staining was observed in tumor or bone sections (Figure 2, A, C, and E). Indistinguishable results were obtained in the presence of the anti-FGF23 (186–206) antibody, whereas sections of colon adenocarcinoma from 2 individuals without clinical or laboratory evidence of TIO revealed no staining for FGF23 (Supplemental Figure 1, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org).

Figure 2.

Staining of colon adenocarcinoma for FGF23. A–D, Incubation of sections of the colon adenocarcinoma, including adjacent normal liver tissue, without (A and C) and with (B and D) an affinity-purified anti-FGF23(225–244) antibody. Bone biopsy sections from a patient with chronic kidney disease served as a positive control (E and F, without and with primary antibody). Red indicates staining for FGF23. Magnification, ×100 for A and B; ×200 for C–F.

Molecular genetic features

A previously described point mutation in the proto-oncogene KRAS was detected, substituting thymine for adenine at codon 12 (35G>T, Gly12Val) (15, 16). Comparative genomic hybridization showed no FGF23 copy number alterations.

Clinical follow-up

The patient was treated with chemotherapy consisting of oxaliplatin, fluorouracil, and bevacizumab. She was discharged home on oral sodium phosphate supplements, 750 mg (24 mmol) 3 times daily, which were eventually decreased to 250 mg (8 mmol) 3 times daily. A computerized tomography scan of her abdomen and pelvis obtained 8 weeks after admission revealed a decrease in the number and size of liver metastases. Repeat FGF23 levels obtained 5 and 12 wk after admission were decreased to the upper end of the normal range, along with marked increases in PTH and 1,25(OH)₂D levels (Table 1). By week 12, serum phosphate levels had normalized, and fractional excretion of phosphate was markedly improved. By week 16, PTH levels had completely normalized to 52 pg/mL.

Discussion

We describe a patient with severe urinary phosphate-wasting that is most likely due to FGF23 production by a metastatic colon adenocarcinoma, as indicated by our immunohistochemical findings and marked reduction in plasma FGF23 levels after chemotherapy. To our knowledge, this represents the first case in which a malignant neoplasm is documented to produce FGF23 leading to renal phosphate-wasting.

Previous reports have suggested an association between TIO and various malignancies, in particular small cell lung carcinoma (3–7) and prostate adenocarcinoma (8–12). However, in the vast majority of cases, plasma FGF23 levels were not measured, and in none of the reported cases were the tumors evaluated immunohistochemically for FGF23 expression. Additionally, in some of these reports hyperphosphaturia was most likely due to an acquired Fanconi syndrome (17), as evidenced by generalized proximal tubular dysfunction with glucosuria, aminoaciduria, and uricosuria. In other reports, Fanconi syndrome was due to multiple myeloma and the established direct nephrotoxicity of Bence Jones protein secreted by these malignant plasma cells.

The current patient showed biochemical abnormalities consistent with TIO, and her liver metastases stained strongly for FGF23. Furthermore, chemotherapy not only reduced her tumor burden but also improved her serum phosphate levels in association with reduced urinary phosphate excretion, reduced plasma FGF23, and markedly increased 1,25(OH)₂D levels. The therapy-associated initial increase in PTH levels was most likely related to secondary hyperparathyroidism due to increased calcium needs for bone remineralization. Reduction in FGF23 levels may have also contributed because FGF23 has been shown to directly reduce PTH secretion (18).

The current patient's initial FGF23 levels were markedly elevated for a patient with normal renal function and similar to those in other patients with TIO (1). The FGF23 secreted by the tumor was clearly biologically active, as demonstrated by the combination of hypophosphatemia and low 1,25(OH)₂D levels. Low 1,25(OH)₂D levels in this setting are caused by the inhibitory effect of FGF23 on the 25-hydroxyvitamin D-1α-hydroxylase and its stimulatory effect on the 25-hydroxyvitamin D-24-hydroxylase, thereby simultaneously inhibiting the production and enhancing the catabolism of 1,25(OH)₂D (19).

The mechanisms leading to FGF23 production by the adenocarcinoma remain to be defined. It is intriguing to speculate that FGF23, in addition to its paraneoplastic effects, may have contributed to the proliferation of primary or metastatic cancer cells. In fact, a recent preliminary study demonstrated elevated risk of recurrent colorectal adenomas among patients with higher levels of FGF23 (20). Furthermore, reduced levels of 1,25(OH)₂D caused by elevated FGF23 levels may have enhanced the development and/or proliferation of colon cancer cells because administration of 1,25(OH)₂D suppresses proliferation of colon cancer cells in vitro (21). Finally, epidemiological studies have consistently shown an inverse association between levels of 25-hydroxyvitamin D, the precursor to 1,25(OH)₂D, and the risk of colon cancer (22).

The finding of a KRAS mutation in the patient's tumor is intriguing. KRAS mutations are routinely assessed for among patients with colon adenocarcinoma, given their poor response to EGFR-inhibiting drugs such as cetuximab (23). Point mutations at the same codon in KRAS observed in the current patient's tumor were also recently reported in several patients with linear nevus sebaceous syndrome (Schimmelpenning syndrome) (24), a disease in which affected individuals can develop hypophosphatemia due to excessive FGF23 production (25). It remains to be determined, however, whether the identified KRAS mutation contributes directly to increased FGF23 production.

In summary, TIO should be considered in the differential diagnosis for any patient with a malignant tumor who presents with hypophosphatemia due to renal phosphate-wasting and low 1,25(OH)₂D levels. In such patients, FGF23 measurement would not only help establish the cause of the biochemical abnormalities, but might also serve as a biomarker of tumor burden and the efficacy of chemotherapeutic interventions.