Background
The liver is the most frequent site of colorectal cancer metastases. Approximately 15 to 25% of patients present with synchronous, and 25 to 30% develop metachronous liver metastasis [
1]. Curative metastasectomy is possible in the setting of liver-limited metastases; however, only 15 to 20% of patients are considered candidates for hepatic resection at the time of presentation. Systemic chemotherapy in the preoperative setting is increasingly used to improve the potential benefit of surgery and to downstage initially unresectable disease, allowing for potentially curative surgery [
2,
3]. The latter strategy, referred to as “conversion chemotherapy” or “downsizing therapy,” is a major reason for the yearly increase in the number of liver resections for colorectal liver metastases (CRLM). The most commonly used CRLM chemotherapy regimens include oxaliplatin (OX) plus 5-fluorouracil (5-FU) and leucovorin (FOLFOX) and irinotecan plus 5-FU and leucovorin (FOLFIRI). As chemotherapy is often administered prior to hepatic resection, adverse effects on the background liver parenchyma of CRLM patients are increasingly recognized [
4,
5]. Recent meta-analysis studies have demonstrated that the nature of preoperative chemotherapy liver injury is a regimen-specific phenomenon [
6]. For example, OX-based regimens are associated with sinusoidal injury, whereas irinotecan-based regimens are associated with steatohepatitis. In the context of liver surgery, chemotherapy-induced liver injury could increase the risks of intra- and postoperative complications and postoperative liver insufficiency [
7].
Sinusoidal obstruction syndrome (SOS), previously termed veno-occlusive disease (VOD), is considered the result of severe toxic injury affecting hepatic sinusoidal endothelial cells (SECs). SOS has been associated with the use of OX-based systemic chemotherapy, cytoreductive therapy prior to hematopoietic stem cell transplantation, hepatic irradiation, pyrrolizidine alkaloid-containing herbal remedies, liver transplantation, and a rare autosomal recessive disorder of the liver and immune system (hepatic veno-occlusive disease with immunodeficiency) [
8]. Macroscopically, the affected liver typically has a characteristic bluish-red marbled appearance and therefore, has been called “blue liver syndrome”. Histologically, SOS is characterized by distinct areas of dilated sinusoids with congestion, which may be associated with liver cell plate atrophy. In severe cases, it can also be associated with perisinusoidal fibrosis, nodular regenerative hyperplasia (NRH), obstruction of centrilobular veins, and peliotic change [
9]. The reported incidence of OX-induced SOS varies between 8.3 and 54% [
9,
10]. Patients with OX-induced SOS in the setting of CRLM have significantly impaired functional hepatic reserve and are predisposed to increased blood transfusions and higher morbidity after the hepatic resection [
7]. Recent studies have demonstrated that SOS induced by OX-based chemotherapy might diminish the response to chemotherapy in patients with CRLM [
11,
12]. SOS may also compromise liver regeneration in patients undergoing hepatectomy [
13]. Therefore, both pathologists and clinicians should be aware of this syndrome as it has a relatively high prevalence and may affect patient outcomes.
“Sinusoidal capillarization” is the altered phenotype of SECs characterized by increased CD34 expression, basement membrane formation, and decreased fenestrae numbers, and is typically seen at the periphery of cirrhotic nodules. In cases of OX-induced liver injury, increased CD34 expression by SECs has been associated with the severity of sinusoidal injury and deterioration of hepatic functional reserve [
14,
15]. It is well known that hepatic stellate cells (HSCs) transform from a quiescent vitamin A-rich type to a myofibroblastic fibrogenic phenotype that can produce excessive amounts of extracellular matrix, leading to perisinusoidal fibrosis [
16]. While centrilobular perisinusoidal/venular fibrosis is often reported in cases with OX-induced liver injury, the expression status of SMA by HSCs has only been sporadically described in this clinical setting [
17,
18]. In the normal liver, GS is expressed exclusively in a narrow rim of pericentral hepatocytes; however, the area of GS-expressing hepatocytes has recently been shown to be significantly expanded in several pathologic conditions including focal nodular hyperplasia (FNH), liver cirrhosis of various etiologies, and idiopathic portal hypertension [
19‐
22]. Because these conditions are associated with altered intrahepatic blood flow caused by shunt vessels or the presence of aberrant vasculature, impaired sinusoidal microcirculation in SOS may also lead to altered GS expression in the hepatic lobules.
The primary objective of this study was to assess the value of immunohistochemical labeling for CD34, SMA, and GS in the diagnosis of SOS after chemotherapy for CRLM and to develop a scoring system that could be easily implemented by pathologists. Each case was given a histologic grade using the sinusoidal obstruction syndrome index score (SOS-I) to assess the likelihood of SOS. Cases were also given an immunohistochemical grade using the total chemotherapy-induced sinusoidal injury (CSI) score calculated as the sum of CD34, SMA, and GS scores. The relationship between the CSI score and histologic disease severity was also assessed.
Discussion
Chemotherapy for CRLM has dramatically improved over the past 15 years, with tumor response rates increasing from 20% with 5-FU alone to 60 to 70% using variable combinations of new chemotherapeutic agents [
3]. The current mainstay of chemotherapy for CRLM is FOLFIRI or FOLFOX and the standard chemotherapy backbone in first-line treatment has gradually been shifting from FOLFIRI to FOLFOX during the last decade [
23,
24]. While the use of modern chemotherapy regimens has dramatically improved the CRLM response rate, its toxicity on the background liver parenchyma is becoming more recognized. OX-based regimens are associated with SEC injury, which may manifest as SOS, centrilobular, perisinusoidal, or venular fibrosis, NRH, and peliosis-like changes in severe cases [
6,
25]. SOS has been associated with an increased need for intraoperative blood transfusions, an increase in the length of stay in the hospital after surgery [
7], a decrease in response to chemotherapy [
11,
12], early recurrence after resection and a short overall survival after resection due to liver insufficiency [
26,
27].
SOS was originally termed hepatic VOD; however, recent experimental studies in pyrrolizidine alkaloid (monocrotaline)-treated rats have clarified that the main injury occurs at the level of the hepatic sinusoids, indicating that central vein involvement is not essential for SOS development [
28]. SOS is caused by toxic injury to SECs and loss of sinusoidal wall integrity with subsequent sinusoid blockage by embolized sinusoidal lining cells. Impaired sinusoidal microcirculation leads to metabolic dysfunction and ischemic damage of adjacent liver cell plates, which may cause atrophy and/or dissociation of liver cell plates and further focal hepatocellular necrosis/dropout.
The histopathologic spectrum of OX-associated liver lesions was extensively investigated by Rubbia-Brandt et al. in a large series of surgically resected CRLM [
9]. According to their study, 54% of patients treated with OX-based regimens had moderate/severe SOS, 47% showed centrilobular perisinusoidal/venular fibrosis, 24.5% developed NRH, and 10.6% manifested peliosis hepatis, while the 111 patients treated by surgery alone had no such lesions. Although the histologic spectrum of OX-associated liver lesions has been well documented, the precise histopathologic diagnosis of OX-associated liver injury in the setting of CRLM can be potentially difficult, due to several reasons. First, surgically resected liver specimens for CRLM contain tumor nodules of varying sizes; therefore, the histologic changes of OX-associated liver injury might be obscured by or misinterpreted as local mechanical and/or hemodynamic changes caused by tumor mass effect. This is why it is critical to sample the non-neoplastic liver away from the tumor/tumors; ideally we suggest greater than 1.5 cm away from the metastatic lesions. Second, SOS lesions can be patchy and more prominent in the subcapsular region. Uneven lesion distribution may make diagnosis difficult, particularly in biopsy samples. Even in resection specimens, it may be difficult to differentiate whether focally enhanced sinusoidal dilatation/congestion is clinically an irrelevant localized change or a true manifestation of SOS. Our stains appeared to detect SOS lesions diffusely throughout the liver parenchyma and may reduce this potential sampling error. Third, the histologic detection of perisinusoidal fibrosis and NRH changes might potentially be difficult to detect by inexperienced pathologists, especially without the use of special stains such as trichrome and reticulin, which are not routinely performed on CRLM specimens. While awareness of the histologic spectrum of OX-induced liver injury, careful examination of precise areas, and appropriate use of special stains would reduce diagnostic errors, additional markers of sinusoidal endothelial injury would further improve the diagnosis and provide a more objective assessment of this unusual type of liver injury.
Hepatic SECs are highly specialized endothelial cells characterized by fenestrations and the absence of an organized basement membrane. Fenestrae are grouped into sieve plates and control the exchange of fluids, solutes, and particles between the sinusoidal blood and the space of Disse, which contains numerous protruding microvilli of hepatocytes. Unlike vascular endothelium elsewhere in the body (except for splenic sinusoids), hepatic SECs do not express CD34 other than the small areas directly adjacent to portal tracts [
29]. In normal conditions, hepatocytes and HSCs maintain the normal SEC phenotype; however, in chronic liver disease, the SECs undergo a phenotypic shift to regular vascular endothelium that is termed “sinusoidal capillarization” [
30]. In such conditions, capillarized SECs lack fenestration, develop an organized basement membrane, and express CD34. Sinusoidal capillarization by means of immunohistochemistry for CD34 was observed in 81.8% of SOS+ cases, which was statistically more frequent and more extensive than in controls. To date, immunohistochemical CD34 assessment of sinusoidal capillarization in the setting of chemotherapy-induced liver injury in patients with CRLM has only been reported in a few studies. According to Narita et al., SOS was found in 39 of 80 patients treated with hepatic resection for CRLM after chemotherapy, of which 33 (85%) showed sinusoidal CD34 overexpression [
14]. They also found that age over 70 years, male sex, and the presence of SOS were independent factors associated with CD34 overexpression. Nalbantoglu et al. demonstrated aberrant capillarization of the sinusoids by CD34 labeling in 98% (45/46) of cases treated with hepatic resection for CRLM after OX-based chemotherapy, and 41% (19/46) showed extensive staining throughout the lobule [
15]. In their series, the overall frequency of sinusoidal capillarization by CD34 was much more frequent than that of histologically recognized OX-induced liver injury (68.1%, 32/47), and extensive multifocal CD34 expression was associated with liver injury severity. As evidenced by our study and others, sinusoidal capillarization is quite frequent in patients with CSI and CRLM and appears to be a fairly early event in the SOS disease process.
While SMA upregulation in activated HSCs has been sporadically reported in VOD of the liver after allogeneic bone marrow transplantation, SMA expression status in the setting of chemotherapy-induced sinusoidal endothelial injury in CRLM has only rarely been described [
17,
18]. Sato et al. described that the area of activated HSCs significantly increased in zones 1 and 2, was more prominent in zone 3 of the hepatic lobules after bone marrow transplantation compared to normal liver tissues, and was much larger in zone 3 of liver tissues with VOD [
31]. Rubbia-Brandt et al. reported strong SMA reactivity in almost all HSCs within the lobules, both along the dilated and intact sinusoids in OX-induced SOS in patients with CRLM, while sparse HSCs were SMA-positive in a small proportion of control liver lobules [
18]. In our study, SMA expression in perisinusoidal HSCs was seen in 72.7% of SOS+ cases, which was more frequent and extensive than in the control group. These results are in agreement with previous descriptions and indicate that HSC activation is a salient feature of SOS.
Recent experimental studies have suggested that normal SECs of the liver function as a gatekeeper, preventing HSC activation; however, once the SEC capillarizes, it no longer prevents HSC activation and permits or promotes HSC activation and subsequent fibrosis [
30]. Because drugs and toxins that cause SOS are selectively more toxic to SECs than to hepatocytes, the disease is initiated by damage at the sinusoid level. If present, prolonged loss of normal SEC with capillarization permits HSC activation and perisinusoidal fibrosis and fibrous occlusion of central venules in severe cases. Thus, given the recent experimental evidence that aberrant CD34 expression by SECs and subsequent SMA overexpression by perisinusoidal HSCs are essential in SOS initiation and development, immunohistochemical labeling for CD34 and SMA appear to be a reasonable and effective tool to identify CSI in patients with CRLM. The frequent overexpression of these markers in cases with histologically evident SOS in ours and a few other observational studies further support the feasibility of these markers as diagnostic aids in this setting.
GS is exclusively expressed in a subpopulation of hepatocytes situated adjacent to the central veins, and its peculiar zonated distribution is considered to represent a “fail-safe” mechanism for ammonia detoxification in mammals [
32]. Ammonia, arriving from the intestine via the portal vein, is first metabolized by periportal hepatocytes through a low-affinity but high-capacity system involving carbamoyl phosphate synthetase and arginase 1 to generate urea. If any ammonia escapes the periportal hepatocytes, it can be scavenged and detoxified by perivenular hepatocytes through a high-affinity but low-capacity system involving GS [
33]. Recently, significant expansion of the area of GS-expressing hepatocytes in human liver has been demonstrated in certain circumstances such as FNH, cirrhosis due to various causes, and idiopathic portal hypertension [
19‐
22]. Because these conditions are well known to have altered microcirculation due to the presence of shunt vessels or abnormalities in venous and/or arterial vessels, we tested the GS expression status based on the assumption that an impaired sinusoidal microcirculation in CSI may lead to altered GS expression. In SOS, there appears to be reduced plasma access to hepatocytes due to SEC defenestration, basement membrane formation, and perisinusoidal fibrosis. In severe cases, sinusoidal perfusion might also be impaired due to microthrombi and partial or total fibrous obliteration of the central venules. Considering the fact that perivenular GS plays an important physiological role as a downstream scavenger for ammonia, it is conceivable that aberrant expression of GS in the midzonal and periportal hepatocytes in SOS may possibly represent protective recruitment of hepatocytes upstream from the perivenular area to maintain overall hepatic ammonia disposal. In our study, aberrant GS expression in the midzonal and periportal hepatocytes was observed in 31.8% of SOS+ cases. Interestingly, most GS-positive cases showed relatively high SOS-I, and none of the control cases showed aberrant GS expression, but these differences did not reach statistical significance. One limitation of our study, was a low number of control cases that had not received neoadjuvant chemotherapy prior to resection. This further supports how commonly these agents are used, and with a larger population of patients GS staining would have likely reached statistical significance.
The diagnosis of CSI should not be made based solely on the histopathologic findings. The diagnosis of CSI/SOS depends on a high index of clinical suspicion, after the clinical exclusion of other potential mimicking causes of liver injury. Several clinical features, such as hyperbilirubinemia, hepatomegaly, ascites, weight gain, and splenomegaly, have been associated with the development of SOS [
34,
35], although their sensitivity and specificity have not been well defined. In the appropriate clinical setting, histological examination of the background liver of the patients with CRLM using SOS-I and the CSI scoring system would certainly increase the diagnostic accuracy of CSI. The major limitation of our study was that it compared CRLM SOS+ cases that received chemotherapy with control CRLM cases that did not. Future studies should use our panel to compare CRLM patients that received chemotherapy and are SOS positive with those that received chemotherapy and are SOS negative. This will ensure that chemotherapy use alone does not result in a positive staining pattern with our panel and an elevated CSI score. These studies, combined with correlation of CIS score to clinical features, will provide additional valuable information for the assessment of sensitivity and specificity of CIS score for clinically significant chemotherapy-related liver injury.