Taxonomic reclassification of Kaposi Sarcoma identifies disease entities with distinct immunopathogenesis

The current categorisation of disease is largely based upon histopathology and clinical manifestations and includes several widely employed classifications such as: International Statistical Classification of Diseases and Related Health Problems (ICD9/10), Systematized Nomenclature of Medicine-Clinical Terms (SNOMED-CT), Unified Medical Language System (UMLS) and Medical Subject Headings (MeSH). These systems are based on the cataloguing of disease in the nineteenth century that originated with the biostatistician William Farr who grouped together “zymotic” or infectious diseases. The underlying principle behind these schemes was to assist in the diagnosis of diseases and perhaps help the choice of treatment. In recent times, the advances in molecular basis of disease and the associated identification of therapeutic targets for precision medicine has led to calls for a new taxonomy of disease [1, 2].

Kaposi Sarcoma (KS) is renowned in the medical and wider community as an acquired immune deficiency syndrome (AIDS) defining malignancy, causing highly visible pigmented cutaneous lesions. Those affected were highly stigmatised in the 1980s and early 1990s before the introduction of effective antiretroviral therapy. However, within the specialist medical community KS was already well known, having first been reported in 1872 by Moritz Kaposi [3]. He described KS as an aggressive pigmented sarcoma of the skin in a strikingly different group of elderly European men. Subsequently, KS has variously been subclassified into four [4] or five [5] subtypes based upon the population of people affected (Table 1).

The variety of KS that Kaposi described became known as ‘Classic’ or sporadic KS and occurs mostly in elderly and middle-aged men of Eastern European, Middle Eastern or Mediterranean origin. It typically presents as pigmented plaques or nodules on the lower limbs and frequently runs an indolent cause [6]. In the 1940s and 50s KS was described in children and adults from sub-Saharan Africa and labelled as endemic KS. In affected children the disease has a preponderance towards involvement of lymph nodes causing lymphoedema and may show more aggressive spread to the viscera, whilst in adults the disease presents in a similar manner to classic KS with plaques or nodules on the limbs, and less often invades locally or disseminates [6, 7]. In 1981 KS was described in young men who have sex with men (MSM) [8] shortly before the identification of human immunodeficiency virus (HIV) as the cause of severe immune deficiency in such patients [5]. This type of KS became known as AIDS related KS, epidemic KS or HIV associated KS and in patients with uncontrolled HIV often presents with widespread nodules and plaques on the limbs, trunk and face, and spread to organs especially the gut and lungs is not uncommon. The era of effective combination antiretroviral therapy (cART) has reduced the occurrence of KS in this groups, but even in those with well controlled HIV, cutaneous and more rarely visceral KS may occur [9]. KS has also been identified in patients who are immunocompromised due to iatrogenic causes, most commonly those seen in solid organ- or bone marrow-transplant recipients and this KS is known as iatrogenic KS [10, 11]. Finally, it has been recently recognised that MSM who are HIV negative also have an increased rate of KS, representing a newly defined fifth group, in whom KS presents in a similar manner to classic KS but in a younger age cohort [12].

The higher rate of KS in HIV infected MSM, as compared to other HIV infected populations, such as haemophiliacs [13], and the link of KS with immunosuppression were key indicators that KS had an infective cause. The causative virus, human herpesvirus-8 (HHV8), was first identified in 1994 using a technique called representational difference analysis, which isolated KS-specific DNA sequences that showed homology with known gammaherpes viruses and helped isolate the newly detected virus [14].

As the causative agent of KS, HHV8 infection is necessary to develop KS, although only a small proportion of those infected will develop KS. In addition to HHV8 infection a degree of impaired host immunity is required. This is clearly identified in HIV associated- and iatrogenic-KS, but is less well defined in the other KS variants. Theories include immunosenescence with ageing for classic KS and a combination of chronic infection and or malnutrition in endemic KS [5, 15].

The initial HHV8 infection is believed to be largely asymptomatic but similar to other herpesviruses infection is lifelong with latent and lytic phases of infection [16, 17]. KS is the most common disease caused by HHV8 infection, but the virus is also linked with primary effusion lymphoma and multicentric Castleman’s disease [17].

Given that HHV8 is required to cause KS, it is clear that the KS variants described do not have a unique pathology i.e. they are still all caused by HHV8; but rather exist on a clinico-pathological continuum caused by unique population characteristics of HHV8 prevalence, age of infection, population health and presence or absence of immune suppression. In addition, there is little evidence of significant biological differences between some of these groups in terms of clinical manifestations or histopathological features, and in many cases the therapeutic approaches are similar. We hypothesise that the most significant difference may be between KS in the immunocompetent versus KS in the immunodeficient host. To assess the medical relevance of the current classification of KS, we evaluated demographic, clinicopathological features, virological and immunological parameters at presentation, as well as overall and disease specific survival in a dataset including 676 patients all with biopsy proven KS.

At the National Centre for HIV malignancies at the Chelsea and Westminster Hospital we prospectively collected routine data on all individuals who attend. All 676 patients diagnosed with biopsy proven Kaposi sarcoma between the years 2000 and 2021 were included in the study. Data collection was approved by the national centre for HIV malignancy. Anonymised data on patient characteristics including biological variables was extracted from the database. The date from KS diagnosis until death, study censoring or loss to follow-up was used to calculate overall survival.

Comparison of variables between the groups for categorical data was by chi² test, or Fishers exact test (if one variable had a zero value), and for non-parametric continuous variables the Mann–Whitney (MW) (2 comparisons) / Kruskal–Wallis (KW) test (> 2 comparisons) were used; all P-values were two-sided. The Bonferroni correction method was used to reduce the type 1 error rate due to multiple comparisons; using the standard level of significance for a single test of P ≤ 0.05, a corrected P value of ≤ 0.0013 was required for significance (39 comparisons). Significant P-values are marked with an asterisk*. Survival was calculated from KS diagnosis until death (overall survival) or last follow-up. Survival curves were plotted according to the method of Kaplan and Meier [18]. The log rank method (Mantel-Cox proportional hazards survival analysis) was used to test for the significance of differences in survival distributions [19]. Analysis also involved disease specific (KS) survival calculated from KS diagnosis until death attributed to KS or last follow-up.

The 676 patients were assigned at presentation into the five groups according to Cesarman, Damania [5]. The largest single group being HIV positive patients, n = 572 (HIV associated KS cohort), with the second largest grouping being MSM, n = 41 (Fig. 1) followed by classic, iatrogenic and endemic. There was a predominance of cis-gendered men who made up 92% (622/676) of the patients (Table 2). Age of presentation of KS was not evenly distributed amongst the groups, with the HIV associated mean age being youngest (41), followed by MSM (52), Iatrogenic (55) Endemic (59) and Classic (74). Overall, 98.4% of patients were first seen at our centre (and data collected) at or within 30 days of diagnosis of KS (data available for 642/676 patients). Eight patients were referred with a prior diagnosis of KS a median of 23 months (range 2–77 months) after diagnosis.

There were a number of tautological findings based on the classification system, including that the MSM group were exclusively cis-male, that the classic KS patients were older and that the endemic KS patient were more likely to be Black (Table 2). There were a number of differences in clinical presentation between the 5 subgroups, with a preponderance towards disseminated presentation of skin lesions in HIV associated KS over localised presentation, with the reverse true in classic KS and MSM KS (Table 2, Additional file 1: S1A). Nodular disease was the most common presentation for all subtypes of KS except for MSM KS. In terms of sites of advanced disease, Iatrogenic KS had more visceral, pulmonary and extensive oral involvement, with extensive oral involvement and visceral involvement also more likely with HIV associated KS. Although visceral disease was frequently found in endemic KS patients at 21% the overall numbers (n = 4) were low. Tumour oedema was least common in MSM KS, and most common in endemic KS (Table 2, Additional file 1: S1B).

HHV8 bloods levels were highest in HIV associated KS followed by iatrogenic KS, and lowest in MSM KS (Table 3, Fig. 2A). CD4 levels were lowest in HIV associated KS (mean 261 cells/µl) and lower levels were also seen in iatrogenic KS (mean 543 cells/µl (Table 3, Fig. 2B)).

KS was the cause of death in 17 patients, representing 25.4% of deaths. Immunosuppression related deaths (16 deaths), and AIDS defining HIV associated malignancy deaths (16 deaths) were also common (Table 4). As the majority of deaths were directly or indirectly related to immune function (KS, Immunosuppression related and AIDS defining malignancy n = 49) overall survival was the most clinically relevant endpoint. Overall survival (OS) was similar between KS subtypes (Log rank P = 0.145) and as univariate analysis was not significant, multivariant analysis was not undertaken (Fig. 3A). Overall, the number of deaths was low with median survival not reached in all apart from the classic KS subtype (Median 191 months (95% CI 0–399 months Additional file 2: S2A), which as it was more elderly had proportionally more deaths. As the median survival was only reached in a single subtype the cohort was underpowered to detect significant differences in OS (Additional file 2: S2A). Disease specific survival was also not significantly different (Log rank P = 0.105) (Fig. 4A).

We compared classic KS to endemic KS, to determine if there were clinicopathological differences between the two classifications (Additional file 3: S3). Classic KS patients were older (mean 74 vs 59 years), but this difference is inherent within the classification system. In terms of clinicopathological differences, there were the similar levels of pulmonary (9 vs 11% chi² P = 0.55), gastrointestinal (4 vs 11% chi² P = 0.36), and other visceral disease involvement (9 vs 21% chi² P = 0.24) and no oral involvement. There were non-significant differences in levels of tumour associated oedema (26 vs 42% chi² P = 0.27) and tumour ulceration (17 vs 21% chi² p = 0.76), and no difference in presenting histology (chi² p = 0.09). In addition there was no difference in the proportion of patients with positive plasma HHV8 DNA (chi² P = 0.71), absolute plasma HHV8 DNA levels (MW P = 0.60) or Log HHV8 levels (MW P = 0.81), and no difference in leukocyte count (MW P = 0.79), lymphocyte count (MW P = 0.91), CD4 count (MW P = 0.39), CD4% (MW P = 0.95), CD8 count (MW P = 0.51) or CD8% (MW P = 0.95) at diagnosis (Additional file 3: S3).

KS in HIV negative MSM was also reviewed and compared to classic and endemic KS. Although KS in MSM was shown to affect more white people and more men, the latter is certainly inherent of the current classification system (Table 2). MSM patients had low levels of visceral involvement with no gastrointestinal, pulmonary or oral involvement, with 5% having other visceral involvement. There was also less tumour ulceration (7%), and less tumour associated oedema (7%) in MSM. In addition, there was no difference in blood HHV8 detection (chi² P p = 0.20) or mean levels (KW P = 0.128) or mean Log HHV8 (KW P = 0.660), and no significant difference in leukocyte, lymphocyte, CD4 or CD8 levels at diagnosis (Additional file 4: S4).

Given the similarities between classic and endemic KS and the low level of organ involvement and tumour ulceration or oedema in MSM KS, plus similar HHV8 and cellular levels, coupled with the observation that HIV associated and iatrogenic-KS were associated with more advanced presentation, we investigated whether an immunosuppressed (HIV associated and Iatrogenic KS) versus non-immunosuppressed (classic, endemic and MSM KS) would be a clinically suitable classification system.

Using this classification system, the immunosuppressed group was younger (KW P < 0.0001*), whereas the non-immunosuppressed group had more Black people (chi² P < 0.0001*) (Table 5). KS related deaths occurred in both groups (Table 4). Skin KS was more often disseminated (chi² P < 0.0001*) and extensive oral mucosal involvement (chi² P < 0.0001*) was more common in the immunosuppressed (Table 5, Fig. 5A and B). There was a trend towards more visceral involvement in the immunocompromised patients including pulmonary involvement (chi² P = 0.07), gastrointestinal involvement (chi² P = 0.09) and other visceral involvement (chi² P = 0.08) but none reached the level of significance. Defining advanced disease as the presence of visceral disease or extensive oral involvement, showed that immunosuppressed patients were significantly more likely to have advanced disease (26% vs 8% chi² = 0.0012*) compared to non-immunosuppressed patients. The immunosuppressed group had a lower leukocyte count but similar lymphocypte count, lower CD4 count (MW P < 0.001*) and CD4% (MW P < 0.001*), and higher CD8 counts (MW P < 0.001*) and CD8% (MW P < 0.001*) (Table 6). There was a trend towards higher average HHV8 viral load (MW P = 0.007) and Log HHV8 (MW P = 0.010) in the immunosuppressed group (Table 6, Fig. 5C). Overall survival was not significantly different between immunosuppressed and non-immunosuppressed (Log rank P = 0.795) (Fig. 3B). The majority of the deaths was in the immunosuppressed group, but the median survival was only reached in the non-immunosuppressed group (median 282 months (S2B)), and therefore the cohort was underpowered to detect significant differences in OS between these groups. Disease specific survival was also not significantly different (Log rank P = 0.793) (Fig. 4B).

Exploratory analysis was completed on the duration of immunosuppression. Duration of immunosuppression was defined as the time from HIV diagnosis to KS diagnosis in HIV associated KS and the time from start of iatrogenic immunosuppression to KS in the iatrogenic group. Those in whom HIV and KS diagnosis was simultaneous were excluded because the period of immunosuppression could not be defined for these patients. Data was available for 406/593 patients with a median of 45.4 months of immunosuppression (range 1–444 months).

Data was available for 392/597 HIV associated KS patients. Median time from HIV to KS diagnosis was 30 months (range 1–184 months) for those with advanced KS involvement at diagnosis (extensive oral and/or visceral) and 51.7 months for patients without advanced KS involvement (range 1–444 months) (P < 0.0001* MW). Data on duration of iatrogenic immunosuppression was available for 14/21 patients. Those with advanced iatrogenic KS involvement had longer periods of immunosuppression (35.9 months versus 63.3), however the difference was not significant (P = 0.11).

Multicentric Castleman’s disease (MCD) which is also HHV8 associated, was recorded as a concomitant diagnosis in 32 patients with HIV associated KS. Median viral load was significantly higher at KS diagnosis 506,500 copies/ml (range 0-554 M) in patients with concomitant or prior MCD versus a median of 0 copies/ml (range 0-49 M) in patients without MCD (P < 0.0001* MW). Two other patients had MCD as a concomitant diagnosis, one with iatrogenic KS (Viral load 0 copies/ml) and one with classic KS (viral load 81,300 copies/ml), indicating that MCD occurred almost exclusively in the immunosuppressed cohort.

The taxonomy of KS is based on a current classification system focused on the demographics and description of KS clinicopathological features in separate populations affected by KS. This system developed as KS was found to involve each group sequentially; from classic KS (1870s), to endemic KS (1940s), iatrogenic KS (1960s), HIV associated KS (1980s) and MSM KS (2010s) [4, 5]. These descriptions took no account of the causative agent, HHV8 which was only discovered in 1994, and paid only limited acknowledgment to the presence or absence of population specific modifiers including immunosuppression.

In this review of 676 consecutive patients with KS who attended the UK national HIV oncology centre at Chelsea and Westminster hospital, we have shown that a significant proportion of the variance in population demographics between the 5 subgroups of KS are tautological findings based on the characteristics of the current classification system. For example, classic KS patients were older in a group defined as ‘elderly’, similarly, there were more Black people in the endemic KS group defined geographical as patients from Africa, there were more cis-men in the HIV associated KS group (HIV remains most common in MSM in the UK [20]) and more cis-men in the MSM KS group.

Further issues with the current classification system are highlighted as follows. Endemic and classic KS classification are both based upon the geographical location of the patient. However, as migration is possible, geographical location could be based upon the place of birth or upbringing (as most HHV8 acquisition is before the age of 20 years [21]) or based on the location the patient resides at the time of KS diagnosis. Which of these definitions is preferable is not defined in the current classification system. A further issue is that classic KS patients are defined as ‘elderly’ in the classification system, but the definition of elderly will vary upon the population mean age [22] and the age of the definer. Classic KS and to a lesser extent endemic KS are also ascribed to racial groups, but is it correct to use racial group in this definition? The differences in KS presentation maybe a reflection of HHV8 seroprevalence in different geographical populations or groups rather than a true reflection of difference caused by race as identified genetically.

In this study we investigated the current classification system by comparing demographic, clinicopathological, virological and immunological parameters in the classic and endemic KS populations. Whilst demographic differences were present e.g. classic KS patients were older (mean 74 vs 59 years) KW P = 0.0004*, this difference is inherent to the current classification system. In contrast, there were no significant clinicopathological differences between these two classifications nor were there differences in blood based immune cells or HHV8 viral levels. The same was true when classic KS and endemic KS were compared to MSM KS, though overall MSM KS appeared to have a less severe clinical presentation.

Given this lack of clinicopathological or immunological differences between classic, endemic and MSM KS, which together comprise those patients defined as non-immunosuppressed, we re-interrogated the dataset based on the presence or absence of immunosuppression as it may reflect a more meaningful difference in pathogenesis of KS than the current classification system.

By reclassifying patients to a non-immunosuppressed (classic, endemic, MSM KS) or immunosuppressed (HIV associated and iatrogenic KS) system we showed that there were meaningful differences between KS clinicopathological findings at presentation. This included more disseminated skin disease and more advanced disease (including visceral and extensive oral involvement) at presentation of KS. Within the immunocompromised group, there was a trend towards more visceral involvement of each subtype, including GI, pulmonary and other visceral involvement. In terms of blood parameters there were lower CD4 counts and leuckocyte counts, higher CD8 counts and a trend towards higher HHV8 viral load in the immunocompromised group, and MCD another HHV8 associated disease was also more common. These findings show that this classification system defines a real difference in pathogenesis between the two groups, in contrast to the current classification system. This division of KS into immunosuppressed and non-immunosuppressed also has therapeutic relevance. The approach to KS in immunosuppressed individuals is to improve the immune function either with cART to treat HIV or by reducing the anti-rejection therapy in the case of allograft recipients prior to direct anti-KS therapy. In contrast in immunocompetent individuals with KS, the treatment strategies are directed against the KS lesions directly usually using surgery, radiotherapy or chemotherapy.

Potential weaknesses to this proposed classification system exist. Firstly, there is no difference in OS or disease specific survival between the two groups despite the noted clinicopathological differences, and, as the number of deaths was low, the cohorts were underpowered for overall survival analyses. The event rate was even lower for disease specific survival, which was similarly underpowered, and no distinction on the recurrence rate of KS was possible between the two groups. This study was undertaken after the widespread use of cART which produced dramatic improvement in KS survival in the HIV associated KS group [23], and thus improvement in treatment of KS [4, 5, 24] and of HIV [25] over the last 40 years is likely responsible for the low death rates and the similarities in OS and disease specific survival.

Secondly, our data set is from the national HIV oncology centre and therefore a significant number of the patients in this study have HIV associated KS which subsequently comprises the majority of the immunosuppressed group. This may bias the analysis of the data particularly in the immunosuppressed group. However, this also reflects the rarity of iatrogenic immunosuppression coupled with a potential low level of KS in the iatrogenic group which may be less exposed to the HHV8 virus, and therefore is likely to remain a small component of immunosuppressed KS numbers.

Thirdly, it is apparent that immunocompetence does not exist as a binary element but rather as a continuum. This is reflected in patients with HIV associated KS who may have regression of their disease when they start anti-retroviral therapy [24] or when iatrogenic patients have KS regression with reduction of immunosuppressive medication [5] suggesting that even within the immunosuppressed group the degree of immunosuppression is important. It is known that when patients with HIV associated-KS who have well controlled HIV grow older they may develop new KS lesions most frequently on the lower limbs in a similar manner to classic KS patients, despite having undetectable HIV RNA levels and high CD4 cell counts [9]. Research has shown that such patients with well controlled HIV who develop KS show an increase in frequency of CD4+ and CD8+ T cells with an immunosenescent phenotype (CD57+ and CD28−) and are therefore susceptible to the same process of immunosenescence with ageing that contributes to KS in classic/non-immunocompromised KS patients [26]. Data on the control of HIV by cART in those patients with HIV associated KS that may help clarify this hypothesis of dual immunosuppression and immunosenescence was not available as part of this study, but would be an important area of future research. However, the study data did show that patients with HIV associated KS presenting with skin only KS (akin to classic KS) had a significantly longer period of immunosuppression compared to those who had advanced KS, highlighting variation even among those with HIV associated KS. One explanation for this finding is that patients with HIV who have a longer duration of cART, have less immune suppression from HIV (which is well controlled) and more immunosenescence, and therefore are less likely to have advanced KS at presentation, supporting the above hypothesis. In contrast, this trend which was not detectable in the iatrogenic group where immunosuppression is less likely to lessen with time.

However, as a broad classification, the immunosuppressed versus non-immunosuppressed is a logical division as those with HIV associated KS and iatrogenic KS have factors causing additional immunosuppression compared to those that maybe in action in the non-immunosuppressed (classic/endemic/MSM) group, including immunosenescence or more poorly understood factors such as chronic infection or malnutrition [5]. As we have shown immunosuppressed patients have a more advanced presentation it is clear these patients are more likely to require anti-tumour therapy, in addition to any intervention designed to bolster the immune system, and therefore our proposed reclassification is a meaningful way to define those who may require more aggressive treatment. Further investigation of the use of KS treatments using this new classification system as a basis will therefore form an important part of future research. The expanding role of immune-based therapies for KS [27] and the reliance of this tumour on the programmed-cell death 1 pathway as a mechanism of immune escape [28] further strengthens the rationale for a classification system centred on the host immune function.

In conclusion, the data we have presented shows that the current classification system is suboptimal, and poorly defines meaningful clinical and pathogenic differences between the subgroups. We propose that a more clinically relevant classification system for KS is to divide patients into immunocompromised and non-immunocompromised groups as this is closely linked with more advanced presentation in the immunocompromised patient and can be more helpful with guiding the need for more aggressive clinical therapeutic approaches.