Article Text

Download PDFPDF

Cytohistological diagnosis of pancreatic serous cystadenoma: a multimodal approach
  1. Michael Steel1,
  2. Samarth Rao2,
  3. Julie Ho3,
  4. Fergal Donnellan2,
  5. Hui-Min Yang1,
  6. David F Schaeffer1
  1. 1 Department of Pathology and Laboratory Medicine, Vancouver General Hospital, Vancouver, British Columbia, Canada
  2. 2 Division of Gastroenterology, Vancouver General Hospital, Vancouver, British Columbia, Canada
  3. 3 Department of Molecular Oncology, British Columbia Cancer Agency, Vancouver, British Columbia, Canada
  1. Correspondence to Dr Michael Steel, Department of Pathology and Laboratory Medicine, Vancouver General Hospital, Vancouver, BC V5Z 1M9, Canada; smike{at}alumni.ubc.ca

Abstract

Aims Serous cystadenomata (SCAs) are benign pancreatic cystic neoplasms that present a diagnostic challenge despite many investigational approaches. Notwithstanding the promise of molecular diagnostics, these tests have limited accessibility in day-to-day surgical pathology practices. We aim to corroborate and build on recent evidence which suggests that positive α-inhibin immunohistochemistry (IHC) is a helpful adjunct in the biopsy confirmation of pancreatic SCA.

Methods We retrospectively reviewed 22 fine-needle aspirates/biopsies from 14 patients (mean age 65 years, 47–83 years) with pancreatic multicystic lesions radiologically suspicious for SCA (location: 6 body, 2 head, 4 tail, 1 neck, 1 uncinate; cyst size: mean 3.7 cm, 2.0–7.6 cm), as well as an additional 10 pancreatic resection specimens with confirmed SCA; α-inhibin IHC was performed on all cell blocks, biopsy slides and representative resection specimen sections. Where available, associated cyst fluid was analysed for correlative vascular endothelial growth factor A (VEGF-A) and carcinoembryonic antigen levels.

Results An α-inhibin IHC sensitivity of 80% was observed in the cases with resection confirmed SCA. Of the fine-needle aspirate/biopsy specimens, 59% (13/22) contained epithelial cells strongly positive for α-inhibin. When selecting for specimens that exhibited distinct strips of epithelium, the α-inhibin strong positivity rate increased to 73% (8/11). VEGF-A values were supportive of false-negative α-inhibin IHC in three cases and true-negative α-inhibin IHC in one case.

Conclusion This study postulates a diagnostic algorithm to confirm pancreatic SCA which may help to decrease unnecessary follow-up endoscopy/surgical resection and would decrease the associated morbidity, mortality and financial costs in patients with this otherwise benign condition.

  • pancreatic serous cystadenoma
  • pancreatic cystic neoplasia
  • α-inhibin immunohistochemistry
  • fine-needle aspiration
  • intracystic cea
  • intracystic vegf

Statistics from Altmetric.com

Introduction

Serous cystadenomata (SCAs) are benign entities that historically have been described to constitute 1% of neoplasia from the exocrine pancreas.1 They account for nearly 16% of surgically resected pancreatic cystic neoplasms (PCNs), which with the rising use of cross-sectional imaging techniques, have increasing clinical recognition, particularly in those over age 70.2 3 In their most classic histological appearance, glycogen-rich strips of flat to cuboidal epithelial cells and an associated extensive capillary network are separated by fibrous septa that form a multiloculated cystic architecture with entrapped serous fluid. Histotype deviation in cyst size can subclassify SCAs into microcystic (most common), oligocystic, mixed-form and solid variants, and despite some differences with regard to their primary pancreatic locations, gender distribution and gross appearances, they all share similar biological behaviour.4 More specifically, these lesions are generally considered to be clinically indolent and non-malignant, so a reliable diagnosis can be safely managed by radiological follow-up alone in small, asymptomatic lesions.5 6

Despite the generally non-threatening clinical behaviour of SCAs, these lesions to date have unreliable prospective diagnostic criteria. From a radiological perspective, be it by CT, MRI or endoscopic ultrasound (EUS), SCAs can be difficult to distinguish from other PCNs, which can often produce a wide differential diagnosis that may comprise inflammatory benign to neoplastic malignant entities as ends of the spectrum of possibilities.7 8 This situation can consequently lead to retrospectively inappropriate repeat endoscopy and/or pancreatic resection as a result of a limited ability to prospectively differentiate SCAs from higher risk lesions.

Fine-needle aspiration and core-needle biopsy, both by CT-guided percutaneous and EUS-guided routes, have been mainstays for the diagnosis of PCNs with radiological uncertainty. Unfortunately, diagnostic accuracy has been limited by a high proportion of paucicellular specimens in this group.9 Biochemical and molecular analyses of cystic fluid have attempted to alleviate the diagnostic dependence on morphology with some success. For example, cyst fluid assessment of carcinoembryonic antigen (CEA) and amylase have been of some use in differentiating SCAs from mucinous lesions and pancreatic pseudocysts10 11; SCA can be evidenced by a cyst fluid biochemical profile of CEA <5 ng/mL (50% sensitive, 95% specific for SCA/pseudocyst) and amylase <250 U/L (44% sensitive, 98% specific for SCA/mucinous cystadenoma (MCN)/mucinous cystadenocarcinoma).

Increased levels of intracystic vascular endothelial growth factor A (VEGF-A) have been described as a 100% sensitive and 97% specific marker of serous neoplasia when using a concentration cut-off of 8500 pg/mL.12 Whole-exome sequencing of PCN tissue has also shown that the application of a panel of five genes (VHL, RNF43, KRAS, GNAS and CTNNB1) allows distinction of mucinous from non-mucinous cysts with good reliability.13 Despite these advancements, current day-to-day surgical pathological practices, particularly those outside of large academic centres, have limitations to the availability of such modalities. Given the increasing incidence of this clinical dilemma, there is demand from practising community pathologists for a reliable diagnostic approach to pancreatic SCA that can, when possible, be efficiently accomplished internally, before having to resort to sending specimens to a reference laboratory for more rigorous downstream diagnostic testing.

Immunohistochemical staining for α-inhibin has been described as a highly sensitive marker for pancreatic SCA on resection specimens. Salomao et al were the first to employ this association on fine-needle aspirate (FNA) cell blocks from patients with SCAs later confirmed on resection. They found that 88% of their patients with SCA were α-inhibin immunohistochemistry (IHC) positive despite approximately 75% demonstrating scant cellularity.9 This therefore offers the possibility for a more accessible adjunct in the biopsy confirmation of pancreatic SCA despite the challenges of their aspirate paucicellularity.

In order to contribute to the effort in minimising the costs and morbidities associated with an inability to rule out pancreatic malignancy in PCNs, the objective of this study is to confirm the reported use of α-inhibin IHC on FNA and biopsy specimens as a diagnostic biomarker in 14 patients with lesions clinically and radiologically suspicious for SCA. In support of the case for true-positive status, we correlate intracystic CEA and VEGF values with FNA/biopsy α-inhibin immunopositivity. We additionally postulate an algorithmic approach to a non–resection-based diagnosis of SCA as a tool to assess for suitability of radiological follow-up alone in small, asymptomatic PCNs suspicious for SCA.

Methods

Following approval by the institutional research ethics board, 14 EUS-guided FNAs and 8 EUS-guided 22-guage core needle biopsies from 14 patients with pancreatic microcystic/oligocystic lesions radiologically suspicious for SCA were retrospectively reviewed (age 47–83 years, mean 65 years; six lesions from pancreatic body, two from pancreatic head, four from pancreatic tail, one from pancreatic neck, one from the uncinate process; cyst size 2.0–7.6 cm, mean 3.7 cm; see table 1). Seven patients had repeat FNA/biopsy (table 2). All patients had their original diagnoses reviewed and were assessed for specimen quality based on the presence or absence of distinct strips of epithelium (three or more cohesive cells with flat, cuboidal or low columnar epithelioid morphology with no atypia, defined by nuclear membrane irregularity, hyperchromatism and high nuclear:cytoplasmic ratio; the same criteria were applied for biopsies and FNAs, including cell blocks). Some of the original diagnoses that were suggestive of pancreatic SCA, despite lacking strips of epithelium, were evidenced using a morphological criteria of one to two flat, cuboidal or low columnar, non-atypical, non-mucinous, glycogenated epithelioid cells; some pathologists additionally used positive α-inhibin IHC and/or cyst fluid biochemistry prospectively in the generation of their diagnoses. A ‘non-diagnostic’ assessment was given to any specimen that was without any obvious epithelium. It is worth noting that one patient (patient 7) had a biopsy and FNA that were indefinite for mild atypia and, despite its possible morphological consistency with SCA, was favoured to be an intraductal papillary mucinous neoplasm (IPMN) on the basis of cyst biochemistry.

Table 1

Demographics and clinical parameters of 14 patients retrospectively reviewed with PCNs radiologically suspicious for SCA

Table 2

Fourteen patients with PCNs radiologically suspicious for SCA with their respective FNA/biopsy diagnoses based on H&E staining, and accompanying α-inhibin IHC results

We immunohistochemically stained each representative slide (generated from CytoLyt-fixed cell blocks for FNA specimens and paraffin blocks for formalin-fixed, fine-needle biopsy specimens) for α-inhibin (clone: R1; vendor: Cell Marque; dilution: 1:100; Ab incubation 30 min; validated on formalin-fixed tissue; see online supplementary appendix A). A specimen was considered ‘strongly positive’ with strong staining demonstrated in >95% of cells (figure 1), ‘negative’ in the absence of any immunoreactivity, and ‘focally positive’ when positive and negative criteria were not fulfilled. The negative internal control for α-inhibin was taken to be gastric foveolar, intestinal, inflammatory or pancreatic stromal cells.

Supplemental material

Figure 1

Biopsy of PCN from patient 6 with H&E stain (A) and α-inhibin IHC (B) showing strong positivity in strips of bland epithelial cells that are suggestive of SCA (×20 magnification). FNA of PCN from patient 1 with H&E stain (C) and α-inhibin IHC (D) showing strong positivity in strips of bland epithelial cells that are suggestive of SCA (×20 magnification). FNA, fine-needle aspiration; IHC, immunohistochemistry; PCN, pancreatic cystic neoplasm; SCA, serous cystadenoma.

The α-inhibin positivity rate was generated for FNAs+biopsies, FNAs alone and biopsies alone based on the slides that were strongly positive relative to the total slides for each subset (table 3). Rates of focal positivity were similarly generated. Each of these subsets were also subdivided into those slides that showed distinct strips of epithelium and those that did not, and a corresponding α-inhibin strong positivity rate was again calculated for these groups.

Table 3

Summarised α-inhibin IHC results for all FNAs and biopsies of PCNs clinically suspicious for SCA

Additionally, α-inhibin IHC with the above protocol was employed on 10 separate formalin-fixed resection specimens with histologically confirmed SCA as an internal test of sensitivity (table 4). The sensitivity was calculated based on the formula:

Table 4

Ten additional patients with resection-confirmed pancreatic SCA with corresponding results of α-inhibin IHC on representative slides

(Inhibin strongly positive cases)/(Total histologically confirmed SCA cases)

When available, pancreatic cyst fluid samples (1:10 dilution or 1:19 dilution) were analysed for VEGF-A using the Quantikine ELISA Human VEGF Immunoassay (R&D Systems) as per manufacturer’s protocol. Available cyst fluid CEA levels were also retrospectively reviewed in the electronic medical record.

Results

Of the 14 patients with PCNs clinically suspicious for SCA, 59% (13/22) of FNA/biopsy specimens contained epithelioid cells strongly positive for α-inhibin. The α-inhibin focal positivity rate in this group was 9% (2/22). When selecting for those FNA/biopsies that exhibited distinct strips of epithelium morphologically suggestive of SCA, the α-inhibin strong positivity rate increased to 73% (8/11), compared with 45% (5/11) in those that did not exhibit distinct strips of cells. The α-inhibin focal positivity rate was 9% (1/11) in both groups (table 3).

Selecting for FNA specimens only, the α-inhibin strong positivity rate was 43% (6/14). The α-inhibin focal positivity rate was 14% (2/14) in this group. When grouping FNA specimens that exhibited distinct strips of epithelium suggestive of SCA, the α-inhibin positivity rate increased to 57% (4/7), compared with 29% (2/7) in those that did not show distinct strips of cells. The α-inhibin focal positivity rate was 14% (1/7) in both groups (table 3).

Selecting for biopsy specimens only, the α-inhibin strong positivity rate was 88% (7/8). When grouping the biopsy specimens that exhibited distinct strips of epithelium, the α-inhibin positivity rate increased to 100% (4/4), compared with 75% (3/4) in those that did not show distinct strips of epithelium. No biopsies showed focal positivity (table 3). Associated intracystic VEGF-A and CEA levels are summarised in table 2.

In the 10 additional patients with pancreatic resection–confirmed SCAs, 8 had representative slides strongly positive on α-inhibin IHC, 1 had focal positivity and 1 was negative, yielding a sensitivity of 80% (table 4).

Discussion

Diagnosis of PCNs can be challenging due to the amalgamation of clinicoradiological ambiguity, cyst fluid biochemical analyses with uncertain significance, inaccessible molecular analyses, and inadequacy with regard to paucicellular lesional FNA and/or biopsy.14 In such circumstances, clinicians and patients can be presented with the difficult decision of whether or not to proceed with costly and risky pancreatic resection secondary to an inability to rule out lesions with malignant potential; confirming a diagnosis of pancreatic SCA often presents with such a problematic clinical situation. Low cellular yield of SCA on FNA seems to be a reflection of their predominantly cystic, scantly epithelial cytohistomorphology.14 This can present pathologists with few cells from which to base their diagnosis on, and the contribution towards ‘non-diagnostic’ or false-negative diagnoses can be accentuated if they have relative inexperience with SCAs.

Immunohistochemical α-inhibin positivity has previously been described as an accessible, sensitive marker for epithelial cells of SCA, an assertion that we confirm with a sensitivity of 80% in our cohort of 10 patients with resection-confirmed SCA (table 4). Moreover, α-inhibin IHC staining on cell blocks made from the FNAs of pancreatic lesions clinically suspicious for SCA has to date showed potential as a tool in favour of the diagnosis.9 Our efforts towards the verification of such found that 59% of FNAs and biopsies of lesions clinicoradiologically suspicious for pancreatic SCA were positive for α-inhibin (table 3). However, the fact that 73% of those FNAs and biopsies stained positively when selecting for those that exhibited distinct strips of epithelium attests to the intuitive probability that paucicellularity drives a loss in α-inhibin sensitivity.

It was unexpected that in those FNAs and biopsies with previously non-diagnostic morphological diagnoses, 45% showed epithelial cells strongly positive for α-inhibin (table 3). Based on previous studies that demonstrate α-inhibin as a relatively specific marker for SCA, the implication is that α-inhibin IHC might expose SCA epithelial cells that were otherwise not apparent based on morphology alone. However, pathologists should take caution in interpreting inhibin positivity that is without a morphological correlate of SCA. A finding of single epithelioid cells that show inhibin positivity in an FNA or biopsy with no morphologically observable epithelial cells on H&E has unknown significance and should not be concluded as evidence of SCA in isolation. Alternatively, situations as in figure 2 show that morphologically non-diagnostic biopsies can show inhibin positivity in groups of cells, and correlation with the H&E-stained slide can sometimes elucidate epithelial cells retrospectively consistent with SCA. In the distinction between these two scenarios, we recommend a threshold of at least five contiguous inhibin-positive cells on otherwise non-diagnostic biopsies before features of SCA can be evidenced (figure 3). Future studies that include long-term follow-up will be required to validate the appropriateness of α-inhibin IHC utilisation and interpretation in this context.

Figure 2

FNA cell block (#2) of PCN from patient 1 with H&E stain (A) and α-inhibin IHC (B) (×20 magnification). Based on H&E alone, the aspirate is non-diagnostic. What initially seems to show non-specific blood vessel–type morphology on H&E is strongly positive on α-inhibin IHC and is retrospectively consistent with epithelial cells from an SCA. FNA, fine-needle aspirate; IHC, immunohistochemistry; PCN, pancreatic cystic neoplasm; SCA, serous cystadenoma.

Figure 3

Flowchart postulating an approach to the clinicopathological diagnosis of pancreatic SCA. A positive α-inhibin immunohistochemical result is recommended to be on the basis of strong positivity only. It is emphasised that this approach should only be used in the context of synaptophysin and chromogranin immunohistochemical negativity. CEA, carcinoembryonic antigen; FNA, fine-needle aspirate; FNB, fine-needle biopsy; IHC, immunohistochemistry; VEGF, vascular endothelial growth factor.

It is difficult to attribute significance to those FNAs/biopsies that showed focal/weak positivity for α-inhibin IHC, particularly given the prevalence of paucicellularity in FNA/biopsy of SCA suspicious lesions, and previous studies have treated such staining patterns as non-diagnostic. We consequently recommend abstaining from the allocation of an SCA diagnosis in the absence of strong immunoreactivity for α-inhibin.

Biopsy results for α-inhibin showed positivity rates that were pronounced relative to those in the FNA group (table 3), even in those specimens that were otherwise morphologically non-diagnostic. The percentage of cases that had morphological features of SCA was similar between biopsies and FNAs, and controlling for those specimens that showed strips of bland epithelium, α-inhibin strong positivity rates were still higher on biopsy specimens than FNA cell blocks (100% vs 57%). This may have been secondary to differences in immunohistochemical affinity between the cell blocks and the formalin-fixed biopsies, but Salamao et al describe an 88% sensitivity of α-inhibin IHC on FNA cell blocks for resection-confirmed SCAs, so this may be unlikely.9 While our results are likely a manifestation of a limited sample size, biopsies might prove to be a more sensitive tissue sampling modality than FNAs with respect to the application of α-inhibin IHC, perhaps by the nature of their more abundant tissue and/or the benefits of their formalin fixation.

Cystic fluid biochemistry, namely CEA and amylase levels, can be helpful in the confirmation of SCA given their specificity in distinction of SCAs from mucinous cystic neoplasms and pseudocysts, respectively.10 11 Intracystic CEA levels of <5 ng/mL and amylase levels of <250 U/L favour SCA, and in the context of radiological and immunohistopathological support, the diagnosis of SCA should be confirmed. Although our study was incomplete in its analysis of these parameters, all cases either focally or strongly immunopositive for α-inhibin which had accompanying cystic biochemistry showed CEA levels <0.5 ng/mL (very likely not MCNs). The lesions with α-inhibin immunonegativity with CEA levels <0.5 ng/mL might be postulated to represent pseudocysts (patients 2 and 3), but this is difficult to conclude with certainty in the absence of amylase levels, and the radiological perspective does not support this. Furthermore, their intracystic VEGF values (>39 870 and 16 130, respectively) are highly suggestive of SCA. It is therefore more likely that patients 2 and 3 had false-negative α-inhibin results, the latter on the basis of aspirative paucicellularity. Alternatively, the lesions from patients 5 and 7 which show α-inhibin immunonegativity in the context of intracystic CEA levels of 20 000, 13 and 22 ng/mL, respectively, are on that basis not likely to be SCAs. To that point, patient 5 had a VEGF-A value of 4715 pg/mL (below the cut-off of 8500 pg/mL, which does not support SCA). Patient 7’s initial FNA and biopsy pathology interpretations evidenced the possibility of mucinous neoplasia, and an intraductal papillary mucinous neoplasm (IPMN) was in fact later confirmed after distal pancreatectomy. This finding in patient 7 supports the previously described specificity of α-inhibin for serous cystadenoma, particularly when the differential includes IPMN, which have described immunonegativity.15 This consideration is key when considering that IPMNs, which are lesions of malignant potential, have recently been described to account for 50% of all resected pancreatic cysts compared with the value of 3% given in 1990.2

Our study does not include paired resection diagnoses of SCA in 13/14 patients with FNA or biopsy, as these patients are currently in a surveillance stage, so the gold standard of SCA true-positive confirmation is not present. Although this impacts the ability to conclude with certainty that α-inhibin positivity is diagnostic of SCA, our inclusion of intracystic CEA and VEGF-A levels helps to support our hypothesis that positive α-inhibin IHC cases represent true-positive SCA status, despite significant limitations with regard to sample size and correlative α-inhibin IHC–positive cases. Moreover, we rely on the previous documentation of α-inhibin IHC specificity for SCA, with some noteworthy diagnostic considerations. Solid pseudopapillary neoplasms have showed no immunoreactivity for α-inhibin, and other α-inhibin IHC–positive mimics can be accurately differentiated from SCA with morphological correlation.16 Namely, α-inhibin positivity in mucinous cystic neoplasms is present in the stromal component, not the epithelial component.15 Acinar cell carcinomas can occasionally show α-inhibin immunoreactivity, but can be differentiated from SCA by the presence of a large central nucleolus.17 It has also been documented that neuroendocrine neoplasms have a 4.1% positivity rate for α-inhibin IHC,18 so careful consideration of the clinical suspicion for neuroendocrine neoplasms, or unusual cytomorphological features suspicious for neuroendocrine neoplasia, should warrant the application of synaptophysin and chromogranin IHC to rule out this possible diagnostic pitfall.9 For this reason, our practice currently maintains a very low threshold for the initial co-employment of these immunohistochemical stains. Knowledge of these aforementioned caveats to α-inhibin IHC interpretation in this clinical context helps to guide its appropriate utilisation and accuracy with regard to downstream decision-making.

In conclusion, our study confirms the sensitivity and, with intracystic biochemical correlation, helps to support the previously documented specificity of α-inhibin IHC for pancreatic SCA on a retrospective cohort of patients with pancreatic lesions radiologically suspicious for SCA. We found that in patients with distinct strips of bland epithelium morphologically consistent with SCA, 73% of FNAs/biopsies were positive for α-inhibin IHC. By employing knowledge of its specificity in context, α-inhibin IHC positivity can contribute to SCA confirmation, which may afford safe avoidance of downstream endoscopy or resection. Taken together, our study postulates an algorithm for pancreatic SCA confirmation that may be helpful in laboratories with limited access to intracystic VEGF levels or molecular testing. We suggest that patients with pancreatic lesions that are radiologically suspicious for SCA with FNA and/or biopsy specimens that show strips of non-atypical, cuboidal, glycogenated, α-inhibin IHC strongly positive and synaptophysin/chromogranin IHC–negative epithelium, a diagnosis of SCA can be made. In this context, radiological follow-up alone can be applied safely, particularly in those with cystic CEA levels <5 ng/mL and amylase levels <250 U/L (figure 3). Future research might include a comparison of FNA versus biopsy in the α-inhibin IHC confirmation of pancreatic SCA or could pertain to the possible use of focal positivity in α-inhibin IHC. A study which correlates α-inhibin IHC status with a greater amount of VEGF data points, including additional cases that are α-inhibin positive and suggestive of SCA, would also be beneficial. Most importantly, studies exploring the long-term appropriateness of our postulated algorithm in pancreatic SCA confirmation, with clinical follow-up, would be required to confirm its successful capacity to help rule out pancreatic lesions with malignant potential.

Take home messages

  • Pancreatic serous cystadenomata are often subject to diagnostic challenge.

  • Correlation of radiological, clinical (eg, cyst fluid biochemistry/vascular endothelial growth factor) and pathological parameters (eg, cytohistomorphology and α-inhibin immunohistochemistry status) is essential to the accurate diagnosis of pancreatic serous cystadenomata.

  • α-Inhibin immunohistochemistry is confirmed to be a sensitive ancillary test in the detection of epithelium from pancreatic serous cystadenomata. This characteristic is particularly helpful in the endoscopic ultrasound–guided biopsy/fine-needle aspiration diagnosis of pancreatic serous cystadenomata, given the challenges of their specimen hypocellularity.

  • α-Inhibin immunohistochemistry is generally specific for pancreatic serous cystadenomata, but important diagnostic pitfalls must be ruled out, which may be afforded when using the standard algorithmic approach postulated herein.

References

Footnotes

  • Handling editor Runjan Chetty.

  • Contributors MS: data curation, formal analysis, writing—original draft, writing—review and editing. SR: data curation, writing—review and editing. JH: data curation, writing—review and editing. FD: supervision, data curation, writing—review and editing. H-MY: supervision, data curation, writing—review and editing. DFS: conceptualisation, data curation, supervision, methodology, writing—review and editing. FD, H-MY and DFS co-supervised this work.

  • Funding Partial funds for this project were received by the VGH and UBC Hospital Foundation.

  • Competing interests DFS reports consulting fees from Robarts Clinical Trials Inc., and the Provincial Health Services Authority, Vancouver, BC, as well as honoraria from Amgen.

  • Patient consent for publication Not required.

  • Ethics approval This manuscript has been read and approved by all authors, has not been published and is not under consideration for publication elsewhere. The study has been approved by the institutional research ethics board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as online supplementary information.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.