Differentiation of soft-tissue lymphoma from undifferentiated sarcoma: apparent diffusion coefficient histogram analysis

Abstract

Background

Making the preoperative diagnosis of soft-tissue lymphoma is important because the treatments for lymphoma and sarcoma are different.

Purpose

To determine the reliability and accuracy of single-slice and whole-tumor apparent diffusion coefficient (ADC) histogram analysis when differentiating soft-tissue lymphoma from undifferentiated sarcoma.

Material and Methods

Patients with confirmed soft-tissue lymphoma or undifferentiated sarcoma who underwent 3-T magnetic resonance imaging (MRI), including diffusion-weighted imaging, were included. Single-slice and whole-tumor ADC histogram analyses were performed using software. Mean, standard deviation (SD), 5th and 95th percentiles, skewness, and kurtosis were compared between groups, and a receiver operating characteristic curve with area under the curve (AUC) was obtained.

Results

Thirteen patients with soft-tissue lymphoma and 12 patients with undifferentiated sarcoma were included. ADC histogram analysis of single-slice and whole-tumor, mean, SD, and 5th and 95th percentiles was significantly lower in lymphoma than in undifferentiated sarcoma. Whole-tumor analysis kurtosis was significantly higher in lymphoma than in undifferentiated sarcoma. All AUCs were high in single-slice and whole-tumor analysis: 0.987 vs. 1.000 in mean; 0.821 vs. 0.782 in SD; 0.949 vs. 0.949 in 5th percentile; and 1.000 vs. 1.000 in 95th percentile without significant difference. AUC of kurtosis in whole-tumor ADC histogram analysis was 0.750.

Conclusion

Single-slice and whole-tumor ADC histogram analysis seems to be reliable and accurate for differentiating soft-tissue lymphoma from undifferentiated sarcoma.

Keywords

Diffusion magnetic resonance imaging lymphoma sarcoma

Introduction

Primary soft-tissue lymphoma is very rare, accounting for 0.01% of all soft-tissue tumors (1). Making the correct diagnosis of soft-tissue lymphoma is essential because the treatment strategies for soft-tissue lymphoma and soft-tissue sarcoma are different. The treatment of choice for lymphoma is chemotherapy with or without radiation therapy. On the other hand, surgery is the mainstream treatment for soft-tissue sarcoma.

A preoperative biopsy is vital to differentiate lymphoma from sarcoma and establish a treatment plan. According to a previous study, 17% of biopsy results may be interpreted as non-diagnostic in soft-tissue lymphoma (2). It was thought that this was due to discohesive and collapsed cells (3). In those cases, if a radiologist alerts oncologists to the possibility of lymphoma, a repeat biopsy or appropriate immunohistochemical evaluation would be performed (4). The possibility of making a correct preoperative diagnosis of soft-tissue lymphoma might be increased if the radiologist suspects lymphoma.

Like other soft-tissue tumors, magnetic resonance imaging (MRI) is the best modality for detecting and evaluating soft-tissue lymphoma. Lymphoma can occur anywhere and involve both cutaneous, subcutaneous, and deep soft tissues. Previous studies (4 –6) reported MRI findings of soft-tissue lymphoma. Unfortunately, MRI features of soft-tissue lymphoma overlap with those of soft-tissue sarcoma. Conventional MRI is often not reliable for differentiating lymphoma from sarcoma due to its non-specific MRI findings.

There are previous reports that diffusion-weighted imaging (DWI) improves the ability to differentiate between lymphoma and other tumors in many organs (7 –10). In addition, there are studies that report that malignant soft-tissue tumors had different apparent diffusion coefficient (ADC) values depending on the histology (11,12).

Undifferentiated sarcoma is the most common type of soft-tissue sarcoma in the extremities, accounting for 34% of all soft-tissue sarcomas in one sarcoma registry (13 –15). The MRI features of undifferentiated sarcoma are various and overlap with those of soft-tissue lymphoma on conventional MR images. Malignant soft-tissue tumors with myxoid and chondroid components have a high ADC value (11,16), and these can be differentiated from lymphoma, which has a low ADC value. In clinical practice, it is challenging to differentiate lymphoma from undifferentiated sarcoma, which is the most common type of soft-tissue sarcoma and can have similar imaging features as lymphoma.

The aim of the present study was to determine the reliability and accuracy of single-slice and whole tumor ADC histogram analysis when differentiating soft-tissue lymphoma from undifferentiated sarcoma using 3-T MRI.

Material and Methods

Patient population

The institutional review board with jurisdiction approved this retrospective study and waived the requirement of informed consent. From August 2009 to March 2018, patients with pathologically confirmed soft-tissue lymphomas and undifferentiated sarcomas were selected. Among these, patients who had undergone 3-T MRI, including DWI, were included. Exclusion criteria were poor image quality, previous treatment, and small (<1 cm) lesions (Fig. 1). Finally, 25 patients (11 men, 14 women; mean age = 64 years; age range = 37–98 years) were included in the study. Among a total of 25 lesions, 13 lesions were soft-tissue lymphomas and 12 lesions were undifferentiated sarcomas.

Fig 1.

Flow diagram of the study.

MRI protocols

MRI examinations were performed using a 3-T MRI unit (MAGNETOM Verio; Siemens Healthineers). The conventional MRI protocols included longitudinal fat-suppressed T2-weighted (T2W) turbo spin-echo (TSE) imaging, axial T1-weighted (T1W) TSE imaging, and axial T2W TSE imaging with and without fat suppression. The acquisition parameters were as follows: field of view (FOV) = 100–280 mm²; matrix size = 512 × 216; slice thickness = 3–10 mm; intersection gap = 0 mm; repetition time (TR) = 680–870/4000–5600 ms; echo time (TE) = 11–21/63–83 ms; turbo factor = 3/13; and number of excitations = 1. A single-shot spin-echo echoplanar DWI sequence was obtained in the axial plane, before contrast enhancement. A parallel imaging technique using GeneRalized Autocalibrating Partially Parallel Acquisitions (GRAPPA) was used with an acceleration factor of 2. Sensitizing diffusion gradients were applied in the x, y, and z directions. Early in the study period, DWI was obtained with four b-values: 0, 300, 800, and 1400 s/mm² (16,17). The acquisition parameters were as follows: FOV = 100–400 mm²; matrix size = 70 × 42–128 × 128; slice thickness = 3–10 mm; intersection gap = 0 mm; TR = 3400–10,800 ms; TE = 49–90 ms; turbo factor = 56; and number of excitations = 3–5. Pixel-based ADC maps were constructed based on a mono-exponential calculation from DWI. Subsequently, a DWI sequence was changed to intravoxel incoherent motion (IVIM) DWI, which used nine b-values: 0, 25, 50, 75, 100, 200, 300, 500, and 800 s/mm². The acquired DWI data were postprocessed to obtain an ADC map. Because DWI sequences changed during the study period, ADC maps were obtained from common b-values of two DWI techniques: 0 and 800 s/mm².

ADC histogram analysis of single slice and whole tumor

ADC histogram analysis was performed using prototype software (Multiparametric Analysis, Siemens Healthineers). A musculoskeletal radiologist (with three years of experience in musculoskeletal radiology), who was blinded to the final pathology report of the tumor, reviewed the MR images on a picture archiving and communication system (PACS). ADC maps were imported into the prototype software. The regions of interest (ROI) of the soft-tissue tumor were drawn on a representative single slice of the ADC map using the postprocessing workstation. Histogram analysis of the single slice was then performed.

For analysis of the whole tumor, a ROI of the soft-tissue tumor was drawn to include the entire tumor on multiple slices. After drawing multiple ROIs, the volume of interest (VOI) was automatically generated on the workstation; then, histogram analysis of the whole tumor was performed. Finally, mean, SD, 5th and 95th percentiles, skewness, and kurtosis were calculated.

Statistical analysis

The pathologic confirmation was used as the standard of reference. We used the Mann–Whitney U test to compare differences in the parameters between lymphoma and undifferentiated sarcoma. The areas under the receiver operating characteristic (ROC) curves (AUCs) were calculated to evaluate diagnostic accuracy. The diagnostic performance of the cut-off value was obtained. The Youden Index, defined as sensitivity + specificity – 1, was used to obtain the optimal cut-off value. We compared the AUCs between single-slice and whole-tumor analysis using the empirical method by Delong et al. (18). Statistical analysis was performed with commercial software (SPSS, version 19; IBM Corp., Armonk, NY, USA and MedCalc, version 11.3.0.0; MedCalc Software, Ostend, Belgium). P < 0.05 was considered statistically significant.

Results

ADC histogram analysis of a single slice

The mean, SD, and 5th and 95th percentiles of lymphoma were significantly lower than those of undifferentiated sarcoma. Skewness and kurtosis were higher in lymphoma than in undifferentiated sarcoma, but these differences were not statistically significant. Table 1 shows parameters from the ADC histogram analysis of the single slice. We obtained the ROC curves of parameters that showed a significant difference between lymphoma and undifferentiated sarcoma (Fig. 2). The AUC of each parameter is summarized in Table 2. The AUC of the 95th percentile was the highest (1.000, 95% confidence interval [CI] = 0.863–1.000). Mean (0.987, 95% CI = 0.840–1.000), 5th percentile (0.949, 95% CI = 0.780–0.997), and SD (0.821, 95% CI = 0.616–0.944) came next. With a cut-off value of 921.00 µm²/s in the 95th percentile, sensitivity and specificity were 100%. With a cut-off value of 535.00 µm²/s in the 5th percentile and a mean of 687.672 µm²/s, sensitivity was 100% and specificity was 85%. With a cut-off value of 37.08 in the SD, sensitivity was 100% and specificity was 62%.

Table 1.

ADC histogram analysis of a single slice in lymphoma and undifferentiated sarcoma.

	Lymphoma	Undifferentiated sarcoma	P value
Mean	545.2 (519.6–629.0)	995.4 (849.7–1143.1)	<0.001
SD	32.4 (28.8–56.1)	87.9 (47.2–140.6)	0.009
5th percentile	493.5 (474.5–512.5)	905.0 (640.5–991.3)	<0.001
95th percentile	584.5 (567.5–778.5)	1163.0 (1037.5–1348.0)	<0.001
Skewness	0.1 (–0.1–0.6)	0.3 (0.2–1.0)	0.477
Kurtosis	–0.2 (–0.6– –0.01)	0.2 (–0.6–1.4)	0.069

Values are given as median (interquartile range). The ADC value is in µm²/s.

ADC, apparent diffusion coefficient; SD, standard deviation.

Fig 2.

ROC curves for ADC histogram analysis of a single slice. The corresponding AUC values are listed in Table 2. The 95th percentile showed the highest AUC (1.000, 95% CI = 0.863–1.000) among the parameters. ADC, apparent diffusion coefficient; AUC, area under the ROC curve; CI, confidence interval; ROC, receiver operating characteristic.

Table 2.

Diagnostic performance for ADC histogram analysis of a single slice.

	Cut-off value	Sensitivity (%)	Specificity (%)	AUC (95% CI)
Mean	687.672	100	85	0.987 (0.840–1.000)
SD	37.08	100	62	0.821 (0.616–0.944)
5th percentile	535.00	100	85	0.949 (0.780–0.997)
95th percentile	921.00	100	100	1 (0.863–1.000)

The ADC value is in µm²/s.

ADC, apparent diffusion coefficient; AUC, area under the receiver operating characteristic curve; CI, confidence interval; SD, standard deviation.

ADC histogram analysis of the whole tumor

The mean, SD, and 5th and 95th percentiles of lymphoma were significantly lower than those of undifferentiated sarcoma. The kurtosis was significantly higher in lymphoma than undifferentiated sarcoma. Skewness was also higher in lymphoma than for undifferentiated sarcoma, but it was not significantly different (Table 3). We obtained the ROC curves of parameters that showed a significant difference between lymphoma and undifferentiated sarcoma (Fig. 3, Table 4). The AUC of the mean (1.000, 95% CI, 0.863–1.000) and 95th percentile (1.000, 0.863–1.000) were the highest and the 5th percentile (0.949, 0.780–0.997), SD (0.782, 0.573–0.921) and kurtosis (0.750, 0.538–0.900) came next. With a cut-off value of 935.51µm²/sec for the mean and 1276.25 µm²/sec in the 95th percentile, sensitivity and specificity were 100%. With a cut-off value of 615.50 µm²/sec in the 5th percentile, sensitivity was 100%, and specificity was 85%. Sensitivity was 75%, and specificity was 77%, with a cut-off value of 146.62 in SD. With a cut-off value of 0.71, sensitivity was 77%, and specificity was 75% in kurtosis.

Table 3.

ADC histogram analysis of the whole tumor in lymphoma and undifferentiated sarcoma.

	Lymphoma	Undifferentiated sarcoma	P value
Mean	646.2 (586.9−720.6)	1253.0 (1119.5−1455.5)	<0.001
SD	126.8 (80.7−143.0)	206.1 (142.2−257.7)	0.014
5th percentile	520.5 (441.5−559.5)	922.0 (651.5−1097.3)	<0.001
95th percentile	873.5 (743.5−994.5)	1554.5 (1416.8−1899.5)	<0.001
Skewness	0.9 (0.2−1.3)	0.1 (−0.2−0.2)	0.067
Kurtosis	1.9 (0.9−3.2)	0.1 (−0.2−0.8)	0.034

Values are given as median (interquartile range). The ADC value is in µm²/s.

ADC, apparent diffusion coefficient; SD, standard deviation.

Fig 3.

ROC curves for ADC histogram analysis of the whole tumor. The corresponding AUC values are listed in Table 4. Mean (1.000, 95% CI = 0.863–1.000) and 95th percentile (1.000, 95% CI = 0.863–1.000) showed the highest AUC among the parameters. ADC, apparent diffusion coefficient; AUC, area under the ROC curve; CI, confidence interval; ROC, receiver operating characteristic.

Table 4.

Diagnostic performance for ADC histogram analysis of the whole tumor.

	Cut-off value	Sensitivity (%)	Specificity (%)	AUC (95% CI)
Mean	935.51	100	100	1 (0.863–1.000)
SD	146.62	75	77	0.782 (0.573–0.921)
5th percentile	615.50	100	85	0.949 (0.780–0.997)
95th percentile	1276.25	100	100	1 (0.863–1.000)
Kurtosis	0.71	77	75	0.750 (0.538–0.900)

The ADC value is in µm²/s.

ADC, apparent diffusion coefficient; AUC, area under the receiver operating characteristic curve; CI, confidence interval; SD, standard deviation.

We compared the diagnostic performance of the ADC histogram analysis of a single slice and that of the whole tumor. There was no significant difference between the AUC of the mean (P = 0.408), SD (P = 0.654), and 5th (P = 0.999) and 95th percentiles (P = 0.211) in single-slice and whole-tumor analysis.

Figures. 4 and 5 show the representative cases of lymphoma and undifferentiated sarcoma, respectively.

Fig 4.

An 89-year-old woman with pathologically confirmed large B-cell lymphoma. Axial T2W image (a) shows hyperintense subcutaneous mass in the right anterior chest wall. This mass shows a hypointense signal on ADC map (b). ADC histogram analysis of a single slice revealed mean 607.38 µm²/s, SD 56, 5th percentile 503.5, and 95th percentile 690.5. ADC histogram analysis of the whole tumor revealed mean 646.15 µm²/s, SD 54.14, 5th percentile 559.5, 95th percentile 730.5, and kurtosis 1.24. Using the cut-off values, all parameters except SD in the single-slice analysis were compatible with lymphoma. ADC, apparent diffusion coefficient; SD, standard deviation; T2W, T2-weighted.

Fig 5.

A 65-year-old man with pathologically confirmed undifferentiated sarcoma. Heterogeneously hyperintense subcutaneous mass is seen in the right posterolateral chest wall on axial T2W image (a). This mass shows a hypointense signal on ADC map (b). ADC histogram analysis of a single slice revealed mean 783.62 µm²/s, SD 104, 5th percentile 640.5, and 95th percentile 959.5. ADC histogram analysis of the whole tumor revealed mean 971.52 µm²/s, SD 291.64, 5th percentile 654.5, 95th percentile 1595.5, and kurtosis 1.54. Using the cut-off values, all parameters except kurtosis in whole-tumor analysis were compatible with undifferentiated sarcoma. ADC, apparent diffusion coefficient; SD, standard deviation; T2W, T2-weighted.

Discussion

The present study demonstrated that ADC histogram analysis of both a single slice and the whole tumor is useful to differentiate soft-tissue lymphoma from undifferentiated sarcoma.

Primary soft-tissue lymphoma is distinctly uncommon. One previous report revealed that only eight soft-tissue lymphomas were found in 7000 lymphomas (19); another report showed that only 472 soft-tissue lymphomas were found in a review of 39,179 soft-tissue tumors over 10 years (1). Although lymphoma has a low incidence, differentiation of lymphoma from sarcoma is vital in order to make the correct treatment decision. Previous studies have tried to determine the MRI features of soft-tissue lymphoma (4 –6). Lymphoma usually shows iso-signal intensity on T1W images, intermediate to high-signal intensity on T2W images, and mild to diffuse homogeneous enhancement. However, these findings are not specific for soft-tissue lymphoma.

DWI reflects cellularity by evaluating the free motility of water-bound protons within tissues (16,17). DWI has been used to detect, characterize, and determine the extent of disease, and assess treatment response in soft-tissue tumors. There have been reports that DWI showed added value when differentiating benign and malignant soft-tissue tumors (11,20,21). In addition, there are previous investigations that compared ADC values among different histologic types of malignant soft-tissue tumors (11,12). According to previous reports, ADC values of lymphoma were lower than those of other malignant tumors. Surov et al. (12) revealed that mean ADC values of muscle lymphomas (0.76 ± 0.14 × 10⁻³ mm²s⁻¹) were significantly lower than those of muscle metastases (1.28 ± 0.24 × 10⁻³ mm² s⁻¹, P = 0.01) and sarcomas (1.82 ± 0.63 × 10⁻³ mm²s⁻¹, P = 0.001). Lower ADC values in lymphoma are thought to be due to high cellularity leading to more restricted water diffusibility (22). The ADC values used in the previous reports were mean, minimum, maximum, and range. Among these, the mean of ADC was most widely used. We also found that the means of ADC were significantly lower in lymphoma than undifferentiated sarcoma using histogram analysis of both a single slice and the whole tumor. These results concur with results from previous studies (11,12). As well as the mean of ADC, we also obtained the SD, 5th and 95th percentiles, skewness, and kurtosis of the ADC map by performing a histogram analysis.

Histogram analysis assesses the distributions of signal intensity in a tumor and reflects tumoral heterogeneity. Previous research examined whether histogram analysis could be a quantitative biomarker for various cancers (23). In the present study, we obtained more quantitative parameters by histogram analysis than by ADC measurement. In addition to the mean, the SD and 5th and 95th percentiles showed a significant difference between lymphoma and undifferentiated sarcoma in both single-slice and whole-tumor analysis; kurtosis also showed a significant difference between the two groups in whole-tumor analysis. The parameters obtained from the histogram analysis showed good diagnostic performance in this study. In particular, the 95th percentile in single-slice ADC histogram analysis and mean and 95th percentile in whole-tumor ADC histogram analysis showed perfect prediction of soft-tissue lymphoma. We compared the diagnostic performance of ADC histogram analysis of the single slice and that of the whole tumor. Both showed high diagnostic performance without a significant difference.

The present study has several limitations. First, the study population was small. Second, selection bias was possible due to its retrospective nature, although we recruited consecutive patients who satisfied the inclusion criteria. Third, we used b-values of 0 and 800 s/mm². Thus, we could not eliminate perfusion-related diffusion effects in the ADC.

In conclusion, ADC histogram analysis of both a single slice and the whole tumor seems to be reliable and accurate for the differentiation of soft-tissue lymphoma from undifferentiated sarcoma. The mean and 95th percentile from histogram analysis of the whole tumor and 95th percentile from histogram analysis of the single slice could be used as quantitative predictive markers to differentiate soft-tissue lymphoma from undifferentiated sarcoma.

Footnotes

Acknowledgements

The author(s) thank Yohan Son and Mun Young Paek (Siemens Healthineers Ltd., Seoul, Republic of Korea) and Robert Grimm (Siemens Healthcare, Erlangen, Germany) for providing the Multiparametric Analysis prototype software.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Won-Hee Jee

References

Kransdorf

MJ.

Malignant soft-tissue tumors in a large referral population: distribution of diagnoses by age, sex, and location. AJR Am J Roentgenol 1995; 164:129–134.

Chang

Huang

Bredella

, et al. Percutaneous CT-guided needle biopsies of musculoskeletal tumors: a 5-year analysis of non-diagnostic biopsies. Skeletal Radiol 2015; 44:1795–1803.

Peng

Cheng

Chang

KC.

Burkitt lymphoma with Azzopardi phenomenon.

Arch Pathol Lab Med 2007; 131:682–683.

Murphey

Kransdorf

MJ.

Primary Musculoskeletal Lymphoma. Radiol Clin North Am 2016; 54:785–795.

Chun

Jee

Park

, et al. MRI features of skeletal muscle lymphoma. AJR Am J Roentgenol 2010; 195:1355–1360.

Suresh

Saifuddin

O’Donnell

Lymphoma presenting as a musculoskeletal soft tissue mass: MRI findings in 24 cases. Eur Radiol 2008; 18:2628–2634.

Kickingereder

Wiestler

Sahm

, et al. Primary central nervous system lymphoma and atypical glioblastoma: multiparametric differentiation by using diffusion-, perfusion-, and susceptibility-weighted MR imaging. Radiology 2014; 272:843–850.

Maeda

Kato

Sakuma

, et al. Usefulness of the apparent diffusion coefficient in line scan diffusion-weighted imaging for distinguishing between squamous cell carcinomas and malignant lymphomas of the head and neck. AJNR Am J Neuroradiol 2005; 26:1186–1192.

Politi

Forghani

Godi

, et al. Ocular adnexal lymphoma: diffusion-weighted mr imaging for differential diagnosis and therapeutic monitoring. Radiology 2010; 256:565–574.

10.

Sugimoto

Koyama

Ichimura

, et al. Comparison of MR imaging features of uterine neuroendocrine carcinoma and uterine malignant lymphoma. Abdom Radiol (NY) 2019; 44:3377–3387.

11.

Nagata

Nishimura

Uchida

, et al. Diffusion-weighted imaging of soft tissue tumors: usefulness of the apparent diffusion coefficient for differential diagnosis. Radiat Med 2008; 26:287–295.

12.

Surov

Nagata

Razek

, et al. Comparison of ADC values in different malignancies of the skeletal musculature: a multicentric analysis. Skeletal Radiol 2015; 44:995–1000.

13.

Fletcher

Gustafson

Rydholm

, et al. Clinicopathologic re-evaluation of 100 malignant fibrous histiocytomas: prognostic relevance of subclassification. J Clin Oncol 2001; 19:3045–3050.

14.

Gibbs

Huang

Lee

, et al. Malignant fibrous histiocytoma: an institutional review. Cancer Invest 2001; 19:23–27.

15.

Le Doussal

Coindre

Leroux

, et al. Prognostic factors for patients with localized primary malignant fibrous histiocytoma: a multicenter study of 216 patients with multivariate analysis. Cancer 1996; 77:1823–1830.

16.

Suzuki

Maeda

Matsumine

, et al. Apparent diffusion coefficient of subcutaneous epidermal cysts in the head and neck comparison with intracranial epidermoid cysts. Acad Radiol 2007; 14:1020–1028.

17.

Khoo

Tyler

Saifuddin

, et al. Diffusion-weighted imaging (DWI) in musculoskeletal MRI: a critical review. Skeletal Radiol 2011; 40:665–681.

18.

DeLong

Clarke-Pearson

DL.

Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44:837–845.

19.

Travis

Banks

Reiman

HM.

Primary extranodal soft tissue lymphoma of the extremities.

Am J Surg Pathol 1987; 11:359–366.

20.

Lee

Jee

Jung

, et al. Differentiation of malignant from benign soft tissue tumours: use of additive qualitative and quantitative diffusion-weighted MR imaging to standard MR imaging at 3.0 T. Eur Radiol 2016; 26:743–754.

21.

Razek

Nada

Ghaniem

, et al. Assessment of soft tissue tumours of the extremities with diffusion echoplanar MR imaging. Radiol Med 2012; 117:96–101.

22.

Guo

Cummings

Dash

, et al. Lymphomas and high-grade astrocytomas: comparison of water diffusibility and histologic characteristics. Radiology 2002; 224:177–183.

23.

Just

Improving tumour heterogeneity MRI assessment with histograms. Br J Cancer 2014; 111:2205–2213.